Tony Pearson is a Master Inventor and Senior IT Architect for the IBM Storage product line at the
IBM Systems Client Experience Center in Tucson Arizona, and featured contributor
to IBM's developerWorks. In 2016, Tony celebrates his 30th year anniversary with IBM Storage. He is
author of the Inside System Storage series of books. This blog is for the open exchange of ideas relating to storage and storage networking hardware, software and services.
(Short URL for this blog: ibm.co/Pearson )
My books are available on Lulu.com! Order your copies today!
Safe Harbor Statement: The information on IBM products is intended to outline IBM's general product direction and it should not be relied on in making a purchasing decision. The information on the new products is for informational purposes only and may not be incorporated into any contract. The information on IBM products is not a commitment, promise, or legal obligation to deliver any material, code, or functionality. The development, release, and timing of any features or functionality described for IBM products remains at IBM's sole discretion.
Tony Pearson is a an active participant in local, regional, and industry-specific interests, and does not receive any special payments to mention them on this blog.
Tony Pearson receives part of the revenue proceeds from sales of books he has authored listed in the side panel.
Tony Pearson is not a medical doctor, and this blog does not reference any IBM product or service that is intended for use in the diagnosis, treatment, cure, prevention or monitoring of a disease or medical condition, unless otherwise specified on individual posts.
For the longest time, people thought that humans could not run a mile in less than four minutes. Then, in 1954, [Sir Roger Bannister] beat that perception, and shortly thereafter, once he showed it was possible, many other runners were able to achieve this also. The same is being said now about the IBM Watson computer which appeared this week against two human contestants on Jeopardy!
(2014 Update: A lot has happened since I originally wrote this blog post! I intended this as a fun project for college students to work on during their summer break. However, IBM is concerned that some businesses might be led to believe they could simply stand up their own systems based entirely on open source and internally developed code for business use. IBM recommends instead the [IBM InfoSphere BigInsights] which packages much of the software described below. IBM has also launched a new "Watson Group" that has [Watson-as-a-Service] capabilities in the Cloud. To raise awareness to these developments, IBM has asked me to rename this post from IBM Watson - How to build your own "Watson Jr." in your basement to the new title IBM Watson -- How to replicate Watson hardware and systems design for your own use in your basement. I also took this opportunity to improve the formatting layout.)
Often, when a company demonstrates new techology, these are prototypes not yet ready for commercial deployment until several years later. IBM Watson, however, was made mostly from commercially available hardware, software and information resources. As several have noted, the 1TB of data used to search for answers could fit on a single USB drive that you buy at your local computer store.
Take a look at the [IBM Research Team] to determine how the project was organized. Let's decide what we need, and what we don't in our version for personal use:
Do we need it for personal use?
Yes, That's you. Assuming this is a one-person project, you will act as Team Lead.
Yes, I hope you know computer programming!
No, since this version for personal use won't be appearing on Jeopardy, we won't need strategy on wager amounts for the Daily Double, or what clues to pick next. Let's focus merely on a computer that can accept a question in text, and provide an answer back, in text.
Yes, this team focused on how to wire all the hardware together. We need to do that, although this version for personal use will have fewer components.
Optional. For now, let's have this version for personal use just return its answer in plain text. Consider this Extra Credit after you get the rest of the system working. Consider using [eSpeak], [FreeTTS], or the Modular Architecture for Research on speech sYnthesis [MARY] Text-to-Speech synthesizers.
Yes, I will explain what this is, and why you need it.
Yes, we will need to get information for personal use to process
Yes, this team developed a system for parsing the question being asked, and to attach meaning to the different words involved.
No, this team focused on making IBM Watson optimized to answer in 3 seconds or less. We can accept a slower response, so we can skip this.
(Disclaimer: As with any Do-It-Yourself (DIY) project, I am not responsible if you are not happy with your version for personal use I am basing the approach on what I read from publicly available sources, and my work in Linux, supercomputers, XIV, and SONAS. For our purposes, this version for personal use is based entirely on commodity hardware, open source software, and publicly available sources of information. Your implementation will certainly not be as fast or as clever as the IBM Watson you saw on television.)
Step 1: Buy the Hardware
Supercomputers are built as a cluster of identical compute servers lashed together by a network. You will be installing Linux on them, so if you can avoid paying extra for Microsoft Windows, that would save you some money. Here is your shopping list:
Three x86 hosts, with the following:
64-bit quad-core processor, either Intel-VT or AMD-V capable,
8GB of DRAM, or larger
300GB of hard disk, or larger
CD or DVD Read/Write drive
Computer Monitor, mouse and keyboard
Ethernet 1GbE 4-port hub, and appropriate RJ45 cables
Surge protector and Power strip
Local Console Monitor (LCM) 4-port switch (formerly known as a KVM switch) and appropriate cables. This is optional, but will make it easier during the development. Once your implementation is operational, you will only need the monitor and keyboard attached to one machine. The other two machines can remain "headless" servers.
Step 2: Establish Networking
IBM Watson used Juniper switches running at 10Gbps Ethernet (10GbE) speeds, but was not connected to the Internet while playing Jeopardy! Instead, these Ethernet links were for the POWER7 servers to talk to each other, and to access files over the Network File System (NFS) protocol to the internal customized SONAS storage I/O nodes.
The implementation will be able to run "disconnected from the Internet" as well. However, you will need Internet access to download the code and information sources. For our purposes, 1GbE should be sufficient. Connect your Ethernet hub to your DSL or Cable modem. Connect all three hosts to the Ethernet switch. Connect your keyboard, video monitor and mouse to the LCM, and connect the LCM to the three hosts.
Step 3: Install Linux and Middleware
To say I use Linux on a daily basis is an understatement. Linux runs on my Android-based cell phone, my laptop at work, my personal computers at home, most of our IBM storage devices from SAN Volume Controller to XIV to SONAS, and even on my Tivo at home which recorded my televised episodes of Jeopardy!
For this project, you can use any modern Linux distribution that supports KVM. IBM Watson used Novel SUSE Linux Enterprise Server [SLES 11]. Alternatively, I can also recommend either Red Hat Enterprise Linux [RHEL 6] or Canonical [Ubuntu v10]. Each distribution of Linux comes in different orientations. Download the the 64-bit "ISO" files for each version, and burn them to CDs.
Graphical User Interface (GUI) oriented, often referred to as "Desktop" or "HPC-Head"
Command Line Interface (CLI) oriented, often referred to as "Server" or "HPC-Compute"
Guest OS oriented, to run in a Hypervisor such as KVM, Xen, or VMware. Novell calls theirs "Just Enough Operating System" [JeOS].
For this version for personal use, I have chosen a [multitier architecture], sometimes referred to as an "n-tier" or "client/server" architecture.
Host 1 - Presentation Server
For the Human-Computer Interface [HCI], the IBM Watson received categories and clues as text files via TCP/IP, had a [beautiful avatar] representing a planet with 42 circles streaking across in orbit, and text-to-speech synthesizer to respond in a computerized voice. Your implementation will not be this sophisticated. Instead, we will have a simple text-based Query Panel web interface accessible from a browser like Mozilla Firefox.
Host 1 will be your Presentation Server, the connection to your keyboard, video monitor and mouse. Install the "Desktop" or "HPC Head Node" version of Linux. Install [Apache Web Server and Tomcat] to run the Query Panel. Host 1 will also be your "programming" host. Install the [Java SDK] and the [Eclipse IDE for Java Developers]. If you always wanted to learn Java, now is your chance. There are plenty of books on Java if that is not the language you normally write code.
While three little systems doesn't constitute an "Extreme Cloud" environment, you might like to try out the "Extreme Cloud Administration Tool", called [xCat], which was used to manage the many servers in IBM Watson.
Host 2 - Business Logic Server
Host 2 will be driving most of the "thinking". Install the "Server" or "HPC Compute Node" version of Linux. This will be running a server virtualization Hypervisor. I recommend KVM, but you can probably run Xen or VMware instead if you like.
Host 3 - File and Database Server
Host 3 will hold your information sources, indices, and databases. Install the "Server" or "HPC Compute Node" version of Linux. This will be your NFS server, which might come up as a question during the installation process.
Technically, you could run different Linux distributions on different machines. For example, you could run "Ubuntu Desktop" for host 1, "RHEL 6 Server" for host 2, and "SLES 11" for host 3. In general, Red Hat tries to be the best "Server" platform, and Novell tries to make SLES be the best "Guest OS".
My advice is to pick a single distribution and use it for everything, Desktop, Server, and Guest OS. If you are new to Linux, choose Ubuntu. There are plenty of books on Linux in general, and Ubuntu in particular, and Ubuntu has a helpful community of volunteers to answer your questions.
Step 4: Download Information Sources
You will need some documents for your implementation to process.
IBM Watson used a modified SONAS to provide a highly-available clustered NFS server. For this version, we won't need that level of sophistication. Configure Host 3 as the NFS server, and Hosts 1 and 2 as NFS clients. See the [Linux-NFS-HOWTO] for details. To optimize performance, host 3 will be the "official master copy", but we will use a Linux utility called rsync to copy the information sources over to the hosts 1 and 2. This allows the task engines on those hosts to access local disk resources during question-answer processing.
We will also need a relational database. You won't need a high-powered IBM DB2. Your implementation can do fine with something like [Apache Derby] which is the open source version of IBM CloudScape from its Informix acquisition. Set up Host 3 as the Derby Network Server, and Hosts 1 and 2 as Derby Network Clients. For more about structured content in relational databases, see my post [IBM Watson - Business Intelligence, Data Retrieval and Text Mining].
Linux includes a utility called wget which allows you to download content from the Internet to your system. What documents you decide to download is up to you, based on what types of questions you want answered. For example, if you like Literature, check out the vast resources at [FullBooks.com]. You can automate the download by writing a shell script or program to invoke wget to all the places you want to fetch data from. Rename the downloaded files to something unique, as often they are just "index.html". For more on wget utility, see [IBM Developerworks].
Step 5: The Query Panel - Parsing the Question
Next, we need to parse the question and have some sense of what is being asked for. For this we will use [OpenNLP] for Natural Language Processing, and [OpenCyc] for the conceptual logic reasoning. See Doug Lenat presenting this 75-minute video [Computers versus Common Sense]. To learn more, see the [CYC 101 Tutorial].
Unlike Jeopardy! where Alex Trebek provides the answer and contestants must respond with the correct question, we will do normal Question-and-Answer processing. To keep things simple, we will limit questions to the following formats:
Who is ...?
Where is ...?
When did ... happen?
What is ...?
Host 1 will have a simple Query Panel web interface. At the top, a place to enter your question, and a "submit" button, and a place at the bottom for the answer to be shown. When "submit" is pressed, this will pass the question to "main.jsp", the Java servlet program that will start the Question-answering analysis. Limiting the types of questions that can be posed will simplify hypothesis generation, reduce the candidate set and evidence evaluation, allowing the analytics processing to continue in reasonable time.
Step 6: Unstructured Information Management Architecture
The "heart and soul" of IBM Watson is Unstructured Information Management Architecture [UIMA]. IBM developed this, then made it available to the world as open source. It is maintained by the [Apache Software Foundation], and overseen by the Organization for the Advancement of Structured Information Standards [OASIS].
Basically, UIMA lets you scan unstructured documents, gleam the important points, and put that into a database for later retrieval. In the graph above, DBs means 'databases' and KBs means 'knowledge bases'. See the 4-minute YouTube video of [IBM Content Analytics], the commercial version of UIMA.
Starting from the left, the Collection Reader selects each document to process, and creates an empty Common Analysis Structure (CAS) which serves as a standardized container for information. This CAS is passed to Analysis Engines , composed of one or more Annotators which analyze the text and fill the CAS with the information found. The CAS are passed to CAS Consumers which do something with the information found, such as enter an entry into a database, update an index, or update a vote count.
(Note: This point requires, what we in the industry call a small matter of programming, or [SMOP]. If you've always wanted to learn Java programming, XML, and JDBC, you will get to do plenty here. )
If you are not familiar with UIMA, consider this [UIMA Tutorial].
Step 7: Parallel Processing
People have asked me why IBM Watson is so big. Did we really need 2,880 cores of processing power? As a supercomputer, the 80 TeraFLOPs of IBM Watson would place it only in 94th place on the [Top 500 Supercomputers]. While IBM Watson may be the [Smartest Machine on Earth], the most powerful supercomputer at this time is the Tianhe-1A with more than 186,000 cores, capable of 2,566 TeraFLOPs.
To determine how big IBM Watson needed to be, the IBM Research team ran the DeepQA algorithm on a single core. It took 2 hours to answer a single Jeopardy question! Let's look at the performance data:
Number of cores
Time to answer one Jeopardy question
Single IBM Power750 server
< 4 minutes
Single rack (10 servers)
< 30 seconds
IBM Watson (90 servers)
< 3 seconds
The old adage applies, [many hands make for light work]. The idea is to divide-and-conquer. For example, if you wanted to find a particular street address in the Manhattan phone book, you could dispatch fifty pages to each friend and they could all scan those pages at the same time. This is known as "Parallel Processing" and is how supercomputers are able to work so well. However, not all algorithms lend well to parallel processing, and the phrase [nine women can't have a baby in one month] is often used to remind us of this.
Fortuantely, UIMA is designed for parallel processing. You need to install UIMA-AS for Asynchronous Scale-out processing, an add-on to the base UIMA Java framework, supporting a very flexible scale-out capability based on JMS (Java Messaging Services) and ActiveMQ. We will also need Apache Hadoop, an open source implementation used by Yahoo Search engine. Hadoop has a "MapReduce" engine that allows you to divide the work, dispatch pieces to different "task engines", and the combine the results afterwards.
Host 2 will run Hadoop and drive the MapReduce process. Plan to have three KVM guests on Host 1, four on Host 2, and three on Host 3. That means you have 10 task engines to work with. These task engines can be deployed for Content Readers, Analysis Engines, and CAS Consumers. When all processing is done, the resulting votes will be tabulated and the top answer displayed on the Query Panel on Host 1.
Step 8: Testing
To simplify testing, use a batch processing approach. Rather than entering questions by hand in the Query Panel, generate a long list of questions in a file, and submit for processing. This will allow you to fine-tune the environment, optimize for performance, and validate the answers returned.
There you have it. By the time you get your implementation fully operational, you will have learned a lot of useful skills, including Linux administration, Ethernet networking, NFS file system configuration, Java programming, UIMA text mining analysis, and MapReduce parallel processing. Hopefully, you will also gain an appreciation for how difficult it was for the IBM Research team to accomplish what they had for the Grand Challenge on Jeopardy! Not surprisingly, IBM Watson is making IBM [as sexy to work for as Apple, Google or Facebook], all of which started their business in a garage or a basement with a system as small as this version for personal use.
As we get to larger and larger flash and spinning disk drives, a common question I get is whether to use RAID-5 versus RAID-6. Here is my take on the matter.
A quick review of basic probability statistics
Failure rates are based on probabilities. Take for example a traditional six-sided die, with numbers one through six represented as dots on each face. What are the chances that we can roll the die several times in a row, that we will have no sixes ever rolled? You might think that if there is a 1/6 (16.6 percent) chance to roll a six, then you would guarantee hit a six after six rolls. That is not the case.
# of Rolls
Probability of no sixes (percent)
So, even after 24 rolls, there is more than 1 percent chance of not rolling a six at all. The formula is (1-1/6) to the 24th power.
Let's say that rolling one to five is success, and rolling a six is a failure. Being successful requires that no sixes appear in a sequence of events. This is the concept I will use for the rest of this post. If you don't care for the math, jump down to the "Summary of Results" section below.
Error Correcting Codes (ECC) and Unreadable Read Errors (URE)
When I speak to my travel agent, I have to provide my six-character [Record Locator] code. Pronouncing individual letters can be error prone, so we use a "spelling alphabet".
The International Radiotelephony Spelling Alphabet, sometimes known as the [NATO phonetic alphabet], has 26 code words assigned to the 26 letters of the English alphabet in alphabetical order as follows: Alfa, Bravo, Charlie, Delta, Echo, Foxtrot, Golf, Hotel, India, Juliett, Kilo, Lima, Mike, November, Oscar, Papa, Quebec, Romeo, Sierra, Tango, Uniform, Victor, Whiskey, X-ray, Yankee, Zulu.
Foxtrot Golf Mike Oscar Victor Whiskey
Foxtrot Gold Mine Oscar Vector Whisker
Boxcart Golf Miko Boxcart Victor Whiskey
Having five or so characters to represent a single character may seem excessive, but you can see that this can be helpful when communications link has static, or background noise is loud, as is often the case at the airport!
If spelling words are misheard, either (a) they are close enough like "Gold" for "Golf", or "Whisker" for "Whiskey", that the correct word is known, or (b) not close enough, such that "Boxcart" could refer to either "Foxtrot" or "Oscar" that we can at least detect that the failure occurred.
For data transfers, or data that is written, and later read back, the functional equivalent is an Error Correcting Code [ECC], used in transmission and storage of data. Some basic ECC can correct a single bit error, and detect double bit errors as failures. More sophisticated ECC can correct multiple bit errors up to a certain number of bits, and detect most anything worse.
When reading a block, sector or page of data from a storage device, if the ECC detects an error, but is unable to correct the bits involved, we call this an "Unrecoverable Read Error", or URE for short.
Bit Error Rate (BER)
Different storage devices have different block, sector or page sizes. Some use 512 bytes, 4096 bytes or 8192 bytes, for example. To normalize likelihood of errors, the industry has simplified this to a single bit error rate or BER, represented often as a power of 10.
Bit Error Rate per read (BER)
Consumer HDD (PC/Laptops)
Enterprise 15k/10k/7200 rpm
Solid-State and Flash
IBM TS1150 tape
In other words, the chance that a bit is unreadable on optical media is 1 in 10 trillion (1E13), on enterprise 15k drives is 1 in 10 quadrillion, and on LTO-7 tape is 1 in 10 quintillion.
There are eight bits per byte, so reading 1 GB of data is like rolling the die eight billion times. The chance of successfully reading 1GB on DVD, then would be (1 - 1/1E13) to the 8 billionth power, or 99.92 percent, or conversely a 0.08 percent chance of failure.
In this paper, Google had studied drive failure using an "Annual Failure Rate" or AFR. Here are two graphs from this paper:
This first graph shows AFR by age. Some drives fail in their first 3-6 months, often called "infant mortality". Then they are fairly reliable for a few years, down to 1.7 percent, then as they get older, they start to fail more often, up to 8.3 percent.
This second graph factors in how busy the drives are. Dividing the drive set into quartiles, "Low" represents the least busy drives (the bottom quartile), "Medium" represents the median two quartiles, and "High" represents the busiest drives, the top quartile. Not surprisingly, the busiest drives tend to fail more often than medium-busy drives.
Given an AFR, what are the chances a drive will fail in the next hour? There are 8,766 hours per year, so the success of a drive over the course of a year is like rolling the die 8,766 times. This allows us to calculate a "Drive Error Rate" or DER:
Drive Error Rate per hour (DER)
For example, an AFR=3 drive has a 1 in 287,800 chance of failing in a particular hour. The probability this drive will fail in the next 24 hours would be like rolling the die 24 times. The formula is (1-1/287,800) to the 24th power, resulting in a failure rate of roughly 0.008 percent.
Let's take a typical RAID-5 rank with 600GB drives at 15K rpm, in a 7+P RAID-5 configuration.
During normal processing, if a URE occurs on a individual drive, RAID comes to the rescue. The system can rebuild the data from parity, and correct the broken block of data.
When a drive fails, however, we don't have this rescue, so a URE that occurs during the rebuild process is catastrophic. How likely is this? Data is read from the other seven drives, and written to a spare empty drive. At 8 bits per byte, reading 4200 GB of data is rolling the die 33.6 trillion times. The formula is then (1-1/E16) to the 33.6 trillionth power, or approximately 0.372 percent chance of URE during the rebuild process.
The time to perform the rebuild depends heavily on the speed of the drive, and how busy the RAID rank is doing other work. Under heavy load, the rebuild might only run at 25 MB/sec, and under no workload perhaps 90 MB/sec. If we take a 60 MB/sec moderate rebuild rate, then it would take 10,000 seconds or nearly 3 hours. The chance that any of the seven drives fail during these three hours, at AFR=10 rolling the DER die (7 x 3) 21 times, results in a 0.025 percent chance of failure.
It is nearly 15 times more likely to get a URE failure than a second drive failure. A rebuild failure would happen with either of these, with a probability of 0.397 percent.
The situation gets worse with higher capacity Nearline drives. Let's do a RAID-5 rank with 6TB Nearline drives at 7200 rpm, in a 7+P configuration. The likelihood of URE reading 42 TB of data, is rolling the die 336 trillion times, or approximately 3.66 percent chance of URE failure. Yikes!
The time to rebuild is also going to take longer. A moderate rebuild rate might only be 30 MB/sec, so that rebuilding a 6TB drive would take 55 hours. The chance that one of the other seven drives fail, assuming again AFR=10, during these 55 hours results in a 0.462 percent.
This time, a URE failure is nearly eight times more likely than a double drive failure. The chance of a rebuild failure is 4.12 percent. Good thing you backed up to tape or object storage!
The math can be done easily using modern spreadsheet software. The URE failure rate is based on the quantity of data read from the remaining drives, so a 4+P with 600GB drives is the same as 8+P with 300GB drives. Both read 2.4 TB of data to recalculate from parity. The Double Drive failure rate is based on the number of drives being read times the number of hours during the rebuild. Slower, higher capacity drives take longer to rebuild. However, in both the 15K and 7200rpm examples, the chance of a URE failure was 8 to 15 times more likely than double drive failure.
Many of the problems associated with RAID-5 above can be mitigated with RAID-6.
After a single drive fails, any URE during rebuild can be corrected from parity. However, if a second drive fails during the rebuild process, then a URE on the remaining drives would be a problem.
Let's start with the 600GB 15k drives in a 6+P+Q RAID-6 configuration. The chance of a second drive failing is 0.0252 percent, as we calculated above. The likelihood of a URE is then based on the remaining six drives, 3600 GB of data. Doing the math, that is 0.0319 percent chance. So, the change of a URE during RAID-6 failure is the probability of both occurring, roughly 0.0000806 percent. Far more reliable than RAID-5!
Likewise, we can calculate the probability of a triple drive failure. After the second drive fails, the likelihood of a third drive at AFR=10, results in 0.00000546 percent.
Combining these, the chance of failure of rebuild is 0.000861 percent.
Switching to 6 TB Nearline drives, in a 6+P+Q RAID-6 configuration, we can do the math in the same manner. The likelihood of URE and two drives failing is 0.0145 percent, and for triple drive failure is 0.00183 percent. Chance of rebuild failure is 0.0163 percent.
Summary of Results
Putting all the results in a table, we have the following:
RAID-5 rebuild failure (percent)
RAID-6 rebuild failure (percent)
600GB 15K rpm
6 TB 7200rpm
Hopefully, I have shown you how to calculate these yourself, so that you can plug in your own drive sizes, rebuild rates, and other parameters to convince yourself of this.
In all cases, RAID-6 drastically reduced the probability of rebuild failure. With modern cache-based systems, the write-penalty associated with additional parity generally does not impact application performance. As clients transition from faster 15K drives to slower, higher capacity 10K and 7200 rpm drives, I highly recommend using RAID-6 instead of RAID-5 in all cases.
In the prank, I indicated that I had submitted my video to the [Arizona International Film Festival], of AIFF for short, which coincidently was running April 1-20, and that it had won an award. I invited everyone who read my blog to see me accept the award at a ceremony at 6:00pm on April 1 at the Fox Theater, followed by the 8:00pm showing of another award-winning film.
I didn't submit the video, the video didn't win any award, and I was not invited to the award ceremony. I did, however, plan to see the movie at 8:00pm.
When I got there, I learned that a dozen of my friends, not realizing it was a prank, showed up, asking for me. The AIFF was quite amused, and invited me to award ceremony still going on. The other filmmakers were impressed I had concocted such an elaborate social media campaign!
A slideshow is another style of video, animating still images to music. The [Ken Burns effect] was named after the technique fellow filmmaker Ken Burns used in his documentaries.
In 2010, I worked with the XIV team to address FUD that our competitors were flinging about double drive failures. My blog post [Double Drive Failure Debunked: XIV Two Years Later] set the record straight and put this issue to rest once and for all. XIV sales shot up dramatically after this post went public!
Are you tired of hearing about Cloud Computing without having any hands-on experience? Here's your chance. IBM has recently launched its IBM Development and Test Cloud beta. This gives you a "sandbox" to play in. Here's a few steps to get started:
Generate a "key pair". There are two keys. A "public" key that will reside in the cloud, and a "private" key that you download to your personal computer. Don't lose this key.
Request an IP address. This step is optional, but I went ahead and got a static IP, so I don't have to type in long hostnames like "vm353.developer.ihost.com".
Request storage space. Again, this step is optional, but you can request a 50GB, 100GB and 200GB LUN. I picked a 200GB LUN. Note that each instance comes with some 10 to 30GB storage already. The advantage to a storage LUN is that it is persistent, and you can mount it to different instances.
Start an "instance". An "instance" is a virtual machine, pre-installed with whatever software you chose from the "asset catalog". These are Linux images running under Red Hat Enterprise Virtualization (RHEV) which is based on Linux's kernel virtual machine (KVM). When you start an instance, you get to decide its size (small, medium, or large), whether to use your static IP address, and where to mount your storage LUN. On the examples below, I had each instance with a static IP and mounted the storage LUN to /media/storage subdirectory. The process takes a few minutes.
So, now that you are ready to go, what instance should you pick from the catalog? Here are three examples to get you started:
IBM WebSphere sMASH Application Builder
Base OS server to run LAMP stack
Next, I decided to try out one of the base OS images. There are a lot of books on Linux, Apache, MySQL and PHP (LAMP) which represents nearly 70 percent of the web sites on the internet. This instance let's you install all the software from scratch. Between Red Hat and Novell SUSE distributions of Linux, Red Hat is focused on being the Hypervisor of choice, and SUSE is focusing on being the Guest OS of choice. Most of the images on the "asset catalog" are based on SLES 10 SP2. However, there was a base OS image of Red Hat Enterprise Linux (RHEL) 5.4, so I chose that.
To install software, you either have to find the appropriate RPM package, or download a tarball and compile from source. To try both methods out, I downloaded tarballs of Apache Web Server and PHP, and got the RPM packages for MySQL. If you just want to learn SQL, there are instances on the asset catalog with DB2 and DB2 Express-C already pre-installed. However, if you are already an expert in MySQL, or are following a tutorial or examples based on MySQL from a classroom textbook, or just want a development and test environment that matches what your company uses in production, then by all means install MySQL.
This is where my SSH client comes in handy. I am able to login to my instance and use "wget" to fetch the appropriate files. An alternative is to use "SCP" (also part of PuTTY) to do a secure copy from your personal computer up to the instance. You will need to do everything via command line interface, including editing files, so I found this [VI cheat sheet] useful. I copied all of the tarballs and RPMs on my storage LUN ( /media/storage ) so as not to have to download them again.
Compiling and configuring them is a different matter. By default, you login as an end user, "idcuser" (which stands for IBM Developer Cloud user). However, sometimes you need "root" level access. Use "sudo bash" to get into root level mode, and this allows you to put the files where they need to be. If you haven't done a configure/make/make install in awhile, here's your chance to relive those "glory days".
In the end, I was able to confirm that Apache, MySQL and PHP were all running correctly. I wrote a simple index.php that invoked phpinfo() to show all the settings were set correctly. I rebooted the instance to ensure that all of the services started at boot time.
Rational Application Developer over VDI
This last example, I started an instance pre-installed with Rational Application Developer (RAD), which is a full Integrated Development Environment (IDE) for Java and J2EE applications. I used the "NX Client" to launch a virtual desktop image (VDI) which in this case was Gnome on SLES 10 SP2. You might want to increase the screen resolution on your personal computer so that the VDI does not take up the entire screen.
From this VDI, you can launch any of the programs, just as if it were your own personal computer. Launch RAD, and you get the familiar environment. I created a short Java program and launched it on the internal WebSphere Application Server test image to confirm it was working correctly.
If you are thinking, "This is too good to be true!" there is a small catch. The instances are only up and running for 7 days. After that, they go away, and you have to start up another one. This includes any files you had on the local disk drive. You have a few options to save your work:
Copy the files you want to save to your storage LUN. This storage LUN appears persistent, and continues to exist after the instance goes away.
Take an "image" of your "instance", a function provided in the IBM Developer and Test Cloud. If you start a project Monday morning, work on it all week, then on Friday afternoon, take an "image". This will shutdown your instance, and backup all of the files to your own personal "asset catalog" so that the next time you request an instance, you can chose that "image" as the starting point.
Another option is to request an "extension" which gives you another 7 days for that instance. You can request up to five unique instances running at the same time, so if you wanted to develop and test a multi-host application, perhaps one host that acts as the front-end web server, another host that does some kind of processing, and a third host that manages the database, this is all possible. As far as I can tell, you can do all the above from either a Windows, Mac or Linux personal computer.
Getting hands-on access to Cloud Computing really helps to understand this technology!
Back in June, I mentioned this blog was [Moving to MyDeveloperWorks] which is based on IBM Lotus Connections.
Finally, the move is complete for all bloggers. If you are having problems with the redirects, you might need to unsubscribe and re-subscribe in your RSS feed reader. Here are the new links for several IBM bloggers that have moved over:
Well, it's Tuesday again, and you know what that means? IBM Announcements!
I just got back from my vacation, so this is a guest post from my colleagues Moshe Weiss, Senior Manager, Development and Design, IBM Storage; and Diane Benjuya, Portfolio Marketing Manager for IBM Spectrum Accelerate.
1. What is IBM announcing?
Today IBM announces another leap forward in storage management, with the availability of IBM Hyper-Scale Manager version 5.1. In April 2016, when IBM announced IBM FlashSystem A9000 and A9000R, they also introduced a fully revamped GUI: IBM Hyper-Scale Manager 5.0. That version brought FlashSystem A9000/A9000R clients a terrific new storage management experience, with advanced look and feel, analytics tools, and other enhancements for managing smarter, with greater simplicity and in less time.
Hyper-Scale Manager 5.x dramatically reduces task time -- by 45% for this task
With Hyper-Scale Manager 5.1, IBM is bringing this exceptional GUI and unified user management experience across the entire set of Spectrum Accelerate-based products, which IBMers internally refer to as the "A family":
IBM Spectrum Accelerate software
IBM XIV Storage System
IBM Hyper-Scale Manager lets you view and move quickly across software-defined, disk based, and all-flash storage in seconds, equipping you with the information you need to ensure every application is performing at its peak.
2. What is innovative about the new GUI -- how does it help clients?
IBM Hyper-Scale Manager 5 makes storage management more insightful and easier in multiple ways, helping clients find info, act and troubleshoot faster. Concepts implemented include: web application with tablet-ready design, single page application, strong navigation scheme, smart filter with analytics, capacity trend/forecast, call for action, better communication using social media. All this helps users make fast, informed decisions while being able to see at a glance the impact of any change on the environment, including into the future. IBM team has designed it over the past three years working closely with clients and using Design Thinking methodology.
Get a holistic view of your storage
Provisioning, Monitoring and Troubleshooting
Find everything, get anywhere
Call for action!
The IBM team applied an "emotional design" approach that makes users feel emotionally attached to GUI for its coolness and elegance -- making the experience not just more productive but also more pleasant.
Version 5.1 brings many exciting and important new features to ease the client's day to day activities. Here are some key ones:
Managing your "A Family" in one UI
Instantly gain insights, spot problematic areas
Integrated Capacity Analytics
4. Any unique features that will be focused on?
The IT industry is entering a cognitive era, right? So IBM has brought cognitive into the GUI. The GUI actually learns each user's habits and preferences over time and adapts the experience to the specific user.
5. How does 5.1 add value to the family of products based on Spectrum Accelerate software?
Hyper-Scale manager makes this powerful family for private, public, hybrid block storage clouds that much more attractive and relevant. Just imagine yourself:
Waking up, driving to the office, opening the UI and seeing that your FlashSystem A9000 systems are doing worse than your XIV in terms of IOPS. Scary, but no worries.
You drill down to the specific FlashSystem A9000 by comparing IOPS. You find that a QoS performance class is deliberately reducing performance for the host. A quick analysis, and you find that it is due to the contract with the host. After a short chat with the host admin, you establish better terms, and decide to stop the IO limitation on the volumes and move them to a disk-based XIV to reduce dollar-per-TB cost.
You look for the best candidate by looking at the capacity trend/forecast charts for each XIV and at growth rate per month. You compare performance metrics and chose the preferred XIV to move the volumes to.
You migrate the volumes from the A9000 to the chosen XIV using the same interface, creating connectivity in one click. You then add the same host configuration as for the A9000 to the XIV in a second click. Then just map and monitor the new IO statistics with a third click. Easy!
Imagine carrying out your daily work and decisions -- creating volumes, monitoring, mirroring, troubleshooting and configuring -- across different systems of different types within the family in single clicks -- without the need to move between user interfaces. You can think of Hyper-Scale Manager 5.1 as a GUI come alive: a dynamic, breathing, thinking work enhancer that simplifies and helps you make the most of your investment.
Come see it in action! Register now for the [Live demo webinar], scheduled for Wednesday, November 9, 2016, from 10am to 11:30am MST!
Download the software from [IBM Fix Central], installation is one click and takes just seconds!
Here is an infographic!
Comments? Feedback? Enter them below. Both Moshe and Diane would be pleased to hear from you!
Today, I met with Teresa Ferraro and Mike Buttrum from FirstRain in their Manhattan office in downtown New York City. IBM recently contracted FirstRain to provide IBMers like myself with analytics on publicly-available news to keep us informed for business meetings. Here's how IBMers can get the most out of this service.
Basically, FirstRain takes a list and generates the best summaries of publicly-available news that are most relevant. You can organize into different channels. Here I have seven channels.
Companies to watch refer to existing or prospective clients that I plan to be talking with soon. Some of my colleagues are assigned to specific clients, so they can set this up once and enjoy the news for the rest of the year. I, on the other hand, meet with different clients every week, so I will be updating this list on a frequent basis.
I have divided the Competitors between major ones, and smaller startups. Since I am often working with business partners and distributors, I made that a separate channel as well.
For product lines, I picked three: Data migration, Data storage solutions, and Software defined storage.
For conferences where I don't know which companies will attend, such as the IBM Technical University, I can set up information by territory. Here is one for Brazil.
I also attend industry-oriented events, so I can pick those vertical markets that might be helpful with dinner conversations. In this example, I chose Energy, Electric Utilities and Gas Utilities.
Once you have your channels configured, you get your results in various sections:
Management Changes lists any changes in top C-level positions, who left the company, who got recently hired.
Key Developments indicates news like mergers and acquisitions and government regulations.
First Reads prioritizes the top six articles for your channel. You can access more, but these six will get you started as you have your morning coffee.
First Tweets gives you the six most relevant tweets, if those articles above were just "TL;DR"
A section on Business Influencers and Market Drivers is interesting to see who the big players are, and what topics are driving the most conversation. Here's an example from my Energy/Electric/Gas channel:
The Most Talked About section covers quotes and commentary about the most talked about companies in your channel.
With most news sources focused on politics, weather and celebrity gossip, it is nice to have a quicker, more focused approach to get the news I need to prepare for my client briefings. Special thanks to my hosts Teresa and Mike for their hospitality!
IBM has chosen three particular Software Defined Environments. At one end, IBM is a platinum sponsor of OpenStack which supports x86 servers, POWER systems and z System mainframes. A problem with open source projects like this, however, is that they can be a bit like putting together IKEA furniture from pieces in a box: "Some assembly required."
At the other end, highly proprietary environments from VMware and Microsoft bring enterprise-ready out-of-the-box solutions. However, nobody wants to be limited to just x86-based solutions. IBM offers the best of both worlds, basing its IBM Cloud and SmartCloud software on OpenStack standards, but providing enterprise-ready solutions for x86, POWER Systems and z System mainframes. This includes IBM Cloud Manager with OpenStack, IBM Cloud Orchestrator, and IBM SmartCloud Cost Management software products.
(Analogy: If open source solutions were vanilla ice cream, and proprietary solutions were chocolate ice cream, then IBM Cloud and SmartCloud is vanilla ice cream with chocolate sauce on top! This is the same approach IBM used for WebSphere Application Server, based on Apache web server, and IBM BigInsights, based on Hadoop analytics.)
For some people, software defined can also refer to how the resources are deployed. Rather than using specialized hardware, solutions based on industry-standard hardware can be delivered either as pre-built appliances, services in the Cloud, or as software-only products.
Back in the 1990s, IBM came up with the [Seascape Storage Enterprise Architecture], deciding to focus the design of its storage systems to be based, where possible and practical, on industry-standard components.
Let's review a few products:
IBM SAN Volume Controller (SVC) and Storwize V7000: IBM storage hypervisors were originally designed to run on industry-standard x86 servers. The IBM scientists at Almaden Research Center referred to this as the "COMmodity PArts Storage System" (COMPASS) architecture.
That is still mostly true 12 years later, but SVC and Storwize V7000 does have specialized hardware, including host bus adapter cards and the [Intel® QuickAssist] chip for Real-time Compression.
IBM DS8000 disk system: The DS8000 is based on off-the-shelf IBM POWER servers. Originally, you could only purchase POWER-based servers from IBM, but now thanks to the [OpenPOWER Foundation], you now have more options.
The DS8000 does use some specialized hardware for its host and device adapters, taking advantage of ASICs and FPGAs to optimize performance.
IBM XIV storage system: IBM acquired XIV back in 2008, but its design is very similar to Seascape architecture. All of the Intellectual Property was in the software, installed on industry-standard x86 servers, cache memory, host bus adapters and 7200 RPM nearline disk drives. I joked that the entire hardware bill-of-materials could be ordered directly from the CDW catalog!
IBM FlashSystem: IBM is #1 rank in the All-Flash Array market. Rather than using off-the-shelf commodity Solid-State drives (SSD), the IBM FlashSystem employs specialized hardware based on FPGAs to optimize performance.
IBM FlashSystem came from the recent acquisition of Texas Memory Systems, and was not designed under the IBM Seascape architecture.
Combining the method the resources are controlled and managed with the way storage is deployed results in a quadrant. Let's take a look at this from a storage perspective:
Traditional storage products that are based on specialized hardware that do not support Software Defined Environment APIs.
Storage products that are based on specialized hardware, but have been enhanced to support Software Defined Environment APIs. For OpenStack, this refers to Cinder and Swift interfaces. For VMware, this would include VAAI, VASA and VADP interfaces and vCenter Console plug-ins.
Storage products that are basically software, either installed on pre-built hardware appliances, offered as services in the Cloud, or software you deploy on your own industry-standard hardware. Unfortunately, this category does not support software defined environment APIs, and so proprietary interfaces require administrator-intensive involvement instead.
Storage software for industry-standard hardware. You purchase the appropriate server, cache memory, flash and disk drives as needed. This category could also extend to pre-built appliance versions of this software, or as services in the Cloud. APIs for software defined environments are available to deploy this with self-service automation.
IBM Spectrum Storage is a family of Category IV software offerings. Here are the products announced:
Based on technology from...
IBM Spectrum Control™
Simplified control and optimization of storage and data infrastructure
SmartCloud Virtual Storage Center, Tivoli Storage Productivity Center
IBM Spectrum Protect™
Single point of administration for data backup and recovery
Tivoli Storage Manager
IBM Spectrum Accelerate™
Accelerating speed of deployment and access to data for new workloads
XIV storage system
IBM Spectrum Virtualize™
Storage virtualization that frees client data from IT boundaries
SAN Volume Controller
IBM Spectrum Scale™
High-performance, scalable storage manages exabytes of unstructured data
GPFS and codename:Elastic Storage
IBM Spectrum Archive™
Enables easy access to long term storage of low activity data
Linear Tape File System (LTFS)
Last year, IDC recognized IBM as #1 in this new emerging software defined storage market. This announcement reinforces IBM's lead in this area. See the [Press Release] for details.
We have a lot to cover, so I will do the quick recap today, and then go in-depth on subsequent posts.
IBM FlashSystem 840 and V840
The FlashSystem now offers a high-voltage 1300W power supply. There are two supplies providing redundancy. In the unlikely event that you are doing maintenance on one of them, the other supply handles all the workload. With the original power supply, the system slowed down the clock speeds to reduce electrical demand. The new power supplies can handle full performance.
Also, the Graphical User Interface (GUI) now holds 300 days of performance data with pan-and-zoom capability. Five predefined graphs showing key performance metrics with additional user-defined metrics available for visualization.
The new v7.4 level of microcode combines features from v7.2.7 and v7.3 into a single code base.
In previous 3-site mirroring implementations, you had A-to-B-to-C cascading. Metro Mirror would get the data from A-to-B, then Global Mirror would copy B-to-C. Multiple Target Peer-to-Peer Remote Copy (PPRC) feature number 7025 allows you to have two separate paths of the data: A-to-B and separately A-to-C. Some folks refer to this as a "star" configuration.
For System z mainframe clients, the new v7.4 introduces new zHyperWrite for DB2 database logs, enhances zGM (XRC) write pacing, and extends Easy Tier automated-tiering API to allow z/OS applications to influence placement on different tiers of storage.
The High Performance Flash Enclosures (HPFE) that IBM introduced last May for the "A" frames are now available for "B" frames. You can have four HPFE in A, and another 4 in B.
DS8870 now offers 600 GB 15K rpm SAS and 1.6 TB 2.5-inch SSD encryption drives for additional capacity and cost performance options to meet data growth demands within the same space. Both support data-at-rest encryption.
Lastly, we have upgraded the OpenStack Cinder driver to the latest Juno release, including features like volume replication and volume retype.
The latest SAN switch is a slim 1U high box that can be configured with 12 or 24 ports. These are 16Bps ports that can auto-negotiate down to 8Gbps, 4Gbps and 2Gbps. These are easy to set up, and can be managed with the IBM Network Advisor management software.
GPFS is the core technology for IBM's "Codename: Elastic Storage" initiative.
You have several options. First, you can purchase just the GPFS software itself. It runs natively on AIX, Windows and Linux, and can be extended to support other operating systems through the use of NAS protocols like NFS or CIFS. Today, the Linux support which was previously just x86 and POWER has been extended to include Linux on System z mainframes as well.
GPFS v4.1 offers "Native RAID" support, with de-clustered RAID in 8+2P and 8+3P configurations. Like the IBM XIV Storage System, this scatters the data across many drives, and can tolerate drive failures better than traditional RAID-5 configurations.
Another option is to get a pre-configured "Converged" appliance that combines servers, storage and hardware. We already offer SONAS and the Storwize V7000 Unified, but IBM now offers the "GPFS Storage Server" running on the new P822L Linux-on-Power servers, RHEL v7, and and GPFS v4.1 with Native RAID to twin-tailed attached DCS3700 expansion drawers. Since GPFS provides the RAID, no need for DCS37000 controllers, saving clients substantial costs.
The IBM Storwize family includes SAN Volume Controller, Storwize V7000, Storwize V7000 Unified, Storwize V5000, Storwize V3700 and Storwize V3500.
The big announcement is that IBM now offers data-at-rest encryption for block data on internal drives in the new generation of Storwize V7000 and V7000 Unified models. There is no performance impact, and no need to purchase new SED-capable drives.
Data-at-rest encryption helps in several ways. First, it protects data if a drive is pulled out and taken away maliciously. Second, it protects data if the drive fails and you want to send it back to the manufacturer for replacement. Third, it allows you to perform a "secure erase" so that the data can be sold or re-purposed without fear of anyone reading previous data.
Initially, the encryption key management is built-in, with the keys stored on a USB memory stick physically attached to the model. In the future, IBM will extend this support to SVC, extend this support to external virtualized drives, and extend this support to IBM Security Key Lifecycle Manager (SKLM).
Other announcements include 16Gbps adapters for SVC, Storwize V7000 and V7000 Unified. The entire Storwize family will also enjoy both 1.8TB 10K RPM 2.5-inch drives, and 6TB 7200RPM 3.5-inch drives
See the Announcement Letter (available later this month) for details.
New TS1150 enterprise tape drives
The anticipation is over! The new TS1150 tape drive has been announced, with 10TB raw un-compressed "JD" media cartridge capacity and 360 MB/sec throughput performance. The new drive is read/write compatible with TS1140 on JC, JY and JK media cartridges.
For the virtual tape libraries for the System z platform, IBM offers two models. The TS7740 had a small amount of disk front ending tape library of physical tape. The TS7720 had a large amount of disk with no tape library.
But then the person carrying the chocolate bar bumped into the person carrying the jar of peanut butter, and the rest is history. IBM will now allow tape attach on TS7720, best of both worlds! Large disk cache plus tape library attach.
Tape-attached TS7720 configurations can have up to eight partitions, with different partitions have different policies. Some might move data from disk cache to tape more aggressively, while other partitions may keep data on disk for longer periods, or indefinitely if needed.
Logical tape volumes can now be up to 25GB in size.
The DCS3700 is IBM's entry-level disk system for sequential-oriented workloads. Today, IBM announced new disk drive options: 400GB 2.5-inch SSD, 800 GB 2.5-inch SSD, and 1.2TB 10K RPM 2.5-inch drive. All of these offer T10 Protection Information (PI) data integrity.
Recently, a client asked how to backup their IBM PureData System for Analytics devices. IBM had [acquired Netezza in November 2010], and later renamed their TwinFin devices as the IBM PureData for Analytics, powered by Netezza.
The [IBM PureData System for Analytics] is incredibly fast for performing deep, ad-hoc analytics. However, the people who use them are "data scientists", not backup experts.
Likewise, there are backup administrators who may not be familiar with the unique characteristics of this expert-integrated system to know what backup options are available.
As with the rest of the IBM PureSystems line, the IBM PureData System for Analytics (or, PDA for short) has a combination of servers, storage and switches inside.
In a full-frame PDA, there are two servers in Active/Passive mode, these coordinate activity to FPGA-based blade servers, which have parallel access to hundreds of disk drives, storing nearly 200 TB of compressed database data. A system can span up to four frames.
But what do you backup? And why? You don't need to worry about backing up the Linux operating system or NPS server code, that is considered firmware and if anything every got corrupted, IBM would help restore it for you. System-wide metadata, such as the host catalog and global users, groups, and permissions should be backed up periodically to protect against data corruption.
There are a number of reasons to backup your user databases:
As part of firmware upgrade/downgrade
To transfer data to another system
Protect against hardware failure / disaster
Protect against data corruption
The PDA has three backup formats. You can backup the entire user database in compressed format, backup individual tables in compressed format, or export to a text-format file.
Compressed format is faster, but can only be restored to the same PDA, or a PDA that has the same or higher level of NPS firmware. The text-format is slower, but can be used to restore to lower levels of NPS firmware, or to other database systems.
There are basically two methods to backup your PDA. The first is called the "Filesystem" method. Basically, you can attach an external storage device to the NPS server, and use the built-in command line interface (CLI) to store the backups onto its file system.
On NPS version 6, the nzhostbackup will backup the /nz/data directory which stores the system tables, database catalogs, configuration files, query plans, and cached executable code for the SPU blade servers.
(I have heard that the nzhostbackup will get deprecated in NPS version 7, but I only have access to version 6. As always, [RTFM] for your specific NPS code level.)
The nzbackup with the users parameter will backup the global users, groups and permissions. This is included in the /nz/data backup contents from the nzhostbackup command, but you may want to backup and restore these separately.
The nzbackup with the db parameter will backup a user database in compressed format. To backup individual tables, use the CREATE EXTERNAL TABLE command, which can create compressed or text-format exports.
You may find that your databases are so large, they will exceed the limits of the filesystem on the external storage device. For SAN or NAS deployments, I recommend the IBM Storwize V7000 Unified with IBM General Parallel File System (GPFS). However, if you are using something else, you may need to use the "nz_backup" scripts provided which split up the backup images into smaller pieces that most other filesystems can handle.
The PDA comes with 10GbE Ethernet ports that you can attach a NAS storage device over a Local Area Network (LAN), or add Fibre Channel Protocol (FCP) ports and connect over a Storage Area Network (SAN). To keep things simple, I will refer to whichever network you decide as the "Backup Network" in the drawings.
The second method for backup is called the "External Backup Software" method. As you have probably guessed, it involves sending the backups to a supported software product like IBM Tivoli Storage Manager (or, TSM for short).
In this case, the PDA acts as a client node, similar to a laptop, desktop, or application server with internal disk. Backup data is sent over the LAN to the designated TSM server, and the TSM server in turn writes over the SAN to its storage hierarchy of disk, virtual tape and/or physical tape resources.
Backups can be done by command "on demand", or automated on a schedule. For the /nz/data directory, direct the nzhostbackup command to send the backup copy to local disk, then use TSM's dsmc archive command to transfer this backup copy to the TSM server.
For nzbackup with the users or db parameters, you can send the data directly to the appropriate TSM server by specifying the connector and connectorArgs parameters.
To reduce traffic on the TSM Server, an intermediary "TSM Proxy Node" can be put in between. In this case, the PDA sends the backup to the Proxy Node, the Proxy Node uses a "LAN Free Storage Agent" to send the backups directly to the virtual tape and/or physical tape, and then notifies the TSM Server to updates its system catalog to record which tape holds these new backups.
Another configuration involves installing the TSM LAN Free storage agent directly on the PDA. While this will require FCP ports to be added and consume more CPU resources on the NPS server, it eliminates most of the LAN traffic, allowing the PDA to send its backups directly to virtual or physical tape.
If Eskimos have 37 words for "snow", then EMC has perhaps a similar number of names for "failure". I have already covered a few of their past attempts, including [ATMOS], [Invista], and [VPLEX]. Last week, EMC introduced its latest, called XtremeIO.
But rather than focus on XtremeIO's many shortcomings, I thought it would be better to point out the highlights of IBM's All-Flash array, IBM FlashSystem.
But first, a quick story.
Two years ago, I worked the booth at [Oracle OpenWorld 2011]. After a conference attendee had visited the booths of Violin Memory and Pure Storage, he asked me why IBM did not have an all-Flash array.
Of course IBM did, and I showed him the [Storwize V7000]. For example, a 2U model with 18 SSD drives of 400GB each, configured in two RAID-5 ranks 7+P+S could offer 5.6 TB of space, running up to 250,000 IOPS at sub-millisecond response times.
Why didn't IBM advertise the Storwize V7000 as an all-Flash array? I though the question was silly at the time, since the Storwize V7000 supported SSD, 15K, 10K and 7200 RPM spinning disk, it seemed obvious that it could be configured with only SSD if you chose.
Since then, IBM has added 800GB support to the Storwize V7000, doubling the capacity. More importantly, IBM acquired Texas Memory Systems, and offers a much better all-Flash array.
Flash can be deployed in three levels. The first is in the server itself, such as with PCiE cards containing Flash chips, limited to applications running on that server only.
The second option is a hybrid disk system, that can intermix Flash-based Solid State Drives (SSD) with regular spinning hard disk drives (HDD). These can be attached to many servers.
The problem with this approach is that when Flash is packaged to pretend to be spinning disk, it undermines some of the performance benefits. Traditional disk system architectures using SCSI commands over Device adapter loops can introduce added latency.
The third fits snuggly in the middle: all-Flash arrays designed from the ground up to be only Flash.
Whereas SSD can typically achieve an I/O latency in the 300 to 1000 microseconds range, IBM FlashSystem can process I/O in the 25 to 110 microsecond range. That is a huge difference!
(FTC Disclosure: The U.S. Federal Trade Commission requires that I mention that I am an IBM employee, and that this post may be considered a paid, celebrity endorsement of both the IBM FlashSystem and IBM Storwize family of products. I have no financial interest in EMC, do not endorse the XtremeIO mentioned here, and was not paid to mention their company or products in any manner.)
Fellow blogger and IBM Master Inventor Barry Whyte has a great comparison table in his blog post [Extreme Blogging]. I thought I would add an added column for the Storwize V7000 with 18 Solid State drives.
IBM FlashSystem 820
IBM Storwize V7000 with SSD
20 Terabytes: 1U
11 Terabytes: 2U
7 Terabytes: 6U
I/O latency (microseconds)
110us (~5x faster)
Maximum I/O per second
NAND Flash type
While it is easy to show that EMC's XtremeIO does not hold a candle to IBM FlashSystems, I think it is more amusing that it is not even as good as a Storwize V7000 with SSD that IBM offered two years ago, long before [EMC acquired XtremeIO company] back in May 2012.
The first day of the residency started with introductions. Our emcee and project leader is Vasfi Gucer from IBM Austin lab. There are 17 participants (referred to as "residents") from the USA and various countries including Brazil, Canada and Sweden.
Michael Fork presenting. I am sitting on the far left side
in the pink shirt. Photo taken by Tina Williams.
To set the right expectations, Tina Williams (IBM Social Media ITSO Projects Program Manager) explained what was going to happen this week.
In a typical "residency", residents are brought together for 4-6 weeks to write an [IBM Redbook] which are often how-to guides written in a very conversational tone.
This residency is different. A bunch of social media and Cloud experts have been brought together to share experiences and to build up skills to write individual blog posts about IBM Cloud offerings. I was invited as both a world-reknown blogger as well as a Cloud expert. Everyone who signed up for this commits to write at least six blog posts about Cloud sometime in the next 90 days.
(Residents who do not have their own blogs can post to the IBM [Thoughts on Cloud] group blog Publishing is part of our promotion process, and writing blogs consistently over a period of time counts!)
Jennifer Turner (IBM Worldwide Cloud Marketing Manager) explained IBM Cloud Social Media Initiative. Five years ago, IBM was one of the top 5 Cloud service providers, then a whole bunch of things happened, and we fell out of the top 5 list, and now with the recent [IBM acquisition of SoftLayer], we are in the top 5 again!
Michael Fork (IBM CloudFirst Lead Architect), presented the latest about SoftLayer. Wow! He did a great job, and am glad to have him as a contact in case I have future questions from clients at the Tucson Executive Briefing Center.
Mohsin Syed [@mohsinusyed], IBM Development Manager, presented [IBM Social Media Analytics], combining Hadoop-style analytics using IBM BigInsights, DB2 database and Cognos reporting. IBM can do [sentiment analysis] to determine positive and negative comments in various languages. This product was formerly known as Cognos Consumer Insight.
I was the last speaker of the day. As one of the top bloggers in both the IT Storage Industry, and company-wide within IBM, I was invited to provide a few tips on blogging to the newbies in the audience. Jeff Antley, the "co-owner" of my blog [Inside System Storage] who works on the IBM developerWorks team, was there on hand to help answer questions.
(IBM requires all highly-visible corporate blogs like mine to have at least two owners. Jeff is an expert at HTML, CSS and other web design and has been immensely helpful in getting my blog looking nicer.)
Well it's Tuesday again, and you know what that means.. IBM announcements! Today, IBM announces that next Monday marks the 60th anniversary of first commercial digital tape storage system! I am on the East coast this week visiting clients, but plan to be back in Tucson in time for the cake and fireworks next Monday.
1925 - masking tape (which 3M sold under its newly announced Scotch® brand)
1930 - clear cellulose-based tape (today, when people say Scotch tape, they usually are referring to the cellulose version)
1935 - Allgemeine Elektrizitatsgesellschaft (AEG) presents Magnetophon K1, audio recording on analog tape
1942 - Duct tape
1947 - Bing Crosby adopts audio recording for his radio program. This eliminated him doing the same program live twice per day, perhaps the first example of using technology for "deduplication".
According to the IBM Archives the [IBM 726 tape drive was formally announced May 21, 1952]. It was the size of a refrigerator, and the tape reel was the size of a large pizza. The next time you pull a frozen pizza from your fridge, you can remember this month's celebration!
When I first joined IBM in 1986, there were three kinds of IBM tape. The round reel called 3420, and the square cartridge called 3480, and the tubes that contained a wide swath of tape stored in honeycomb shelves called the [IBM 3850 Mass Storage System].
My first job at IBM was to work on DFHSM, which was specifically started in 1977 to manage the IBM 3850, and later renamed to the DFSMShsm component of the DFSMS element of the z/OS operating system. This software was instrumental in keeping disk and tape at high 80-95 percent utilization rates on mainframe servers.
While visiting a client in Detroit, the client loved their StorageTek tape automation silo, but didn't care for the StorageTek drives inside were incompatible with IBM formats. They wanted to put IBM drives into the StorageTek silos. I agreed it was a good idea, and brought this back to the attention of development. In a contentious meeting with management and engineers, I presented this feedback from the client.
Everyone in the room said IBM couldn't do that. I asked "Why not?" The software engineers I spoke to already said they could support it. With StorageTek at the brink of Chapter 11 bankruptcy, I argued that IBM drives in their tape automation would ease the transition of our mainframe customers to an all-IBM environment.
Was the reason related to business/legal concerns, or was their a hardware issue? It turned out to be a little of both. On the business side, IBM had to agree to work with StorageTek on service and support to its mutual clients in mixed environments. On the technical side, the drive had to be tilted 12 degrees to line up with the robotic hand. A few years later, the IBM silo-compatible 3592 drive was commercially available.
Rather than put StorageTek completely out of business, it had the opposite effect. Now that IBM drives can be put in StorageTek libraries, everyone wanted one, basically bringing StorageTek back to life. This forced IBM to offer its own tape automation libraries.
In 1993, I filed my first patent. It was for the RECYCLE function in DFHSM to consolidate valid data from partial tapes to fresh new tapes. Before my patent, the RECYCLE function selected tapes alphabetically, by volume serial (VOLSER). My patent evaluated all tapes based on how full they were, and sorted them least-full to most-full, to maximize the return of cartridges.
Different tape cartridges can hold different amounts of data, especially with different formats on the same media type, with or without compression, so calculating the percentage full turned out to be a tricky algorithm that continues to be used in mainframe environments today.
The patent was popular for cross-licensing, and IBM has since filed additional patents for this invention in other countries to further increase its license revenue for intellectual property.
In 1997, IBM launched the IBM 3494 Virtual Tape Server (VTS), the first virtual tape storage device, blending disk and tape to optimal effect. This was based off the IBM 3850 Mass Storage Systems, which was the first virtual disk system, that used 3380 disk and tape to emulate the older 3350 disk systems.
In the VTS, tape volume images would be emulated as files on a disk system, then later moved to physical tape. We would call the disk the "Tape Volume Cache", and use caching algorithms to decide how long to keep data in cache, versus destage to tape. However, there were only a few tape drives, and sometimes when the VTS was busy, there were no tape drives available to destage the older images, and the cache would fill up.
I had already solved this problem in DFHSM, with a function called pre-migration. The idea was to pre-emptively copy data to tape, but leave it also on disk, so that when it needed to be destaged, all we had to do was delete the disk copy and activate the tape copy. We patented using this idea for the VTS, and it is still used in the successor models of IBM Sysem Storage TS7740 virtual tape libraries today.
Today, tape continues to be the least expensive storage medium, about 15 to 25 times less expensive, dollar-per-GB, than disk technologies. A dollar of today's LTO-5 tape can hold 22 days worth of MP3 music at 192 Kbps recording. A full TS1140 tape cartridge can hold 2 million copies of the book "War and Peace".
(If you have not read the book, Woody Allen took a speed reading course and read the entire novel in just 20 minutes. He summed up the novel in three words: "It involves Russia." By comparison, in the same 20 minutes, at 650MB/sec, the TS1140 drive can read this novel over and over 390,000 times.)
If you have your own "war stories" about tape, I would love to hear them, please consider posting a comment below.
Tonight PBS plans to air Season 38, Episode 6 of NOVA, titled [Smartest Machine On Earth]. Here is an excerpt from the station listing:
"What's so special about human intelligence and will scientists ever build a computer that rivals the flexibility and power of a human brain? In "Artificial Intelligence," NOVA takes viewers inside an IBM lab where a crack team has been working for nearly three years to perfect a machine that can answer any question. The scientists hope their machine will be able to beat expert contestants in one of the USA's most challenging TV quiz shows -- Jeopardy, which has entertained viewers for over four decades. "Artificial Intelligence" presents the exclusive inside story of how the IBM team developed the world's smartest computer from scratch. Now they're racing to finish it for a special Jeopardy airdate in February 2011. They've built an exact replica of the studio at its research lab near New York and invited past champions to compete against the machine, a big black box code -- named Watson after IBM's founder, Thomas J. Watson. But will Watson be able to beat out its human competition?"
Like most supercomputers, Watson runs the Linux operating system. The system runs 2,880 cores (90 IBM Power 750 servers, four sockets each, eight cores per socket) to achieve 80 [TeraFlops]. TeraFlops is the unit of measure for supercomputers, representing a trillion floating point operations. By comparison, Hans Morvec, principal research scientist at the Robotics Institute of Carnegie Mellon University (CMU) estimates that the [human brain is about 100 TeraFlops]. So, in the three seconds that Watson gets to calculate its response, it would have processed 240 trillion operations.
Several readers of my blog have asked for details on the storage aspects of Watson. Basically, it is a modified version of IBM Scale-Out NAS [SONAS] that IBM offers commercially, but running Linux on POWER instead of Linux-x86. System p expansion drawers of SAS 15K RPM 450GB drives, 12 drives each, are dual-connected to two storage nodes, for a total of 21.6TB of raw disk capacity. The storage nodes use IBM's General Parallel File System (GPFS) to provide clustered NFS access to the rest of the system. Each Power 750 has minimal internal storage mostly to hold the Linux operating system and programs.
When Watson is booted up, the 15TB of total RAM are loaded up, and thereafter the DeepQA processing is all done from memory. According to IBM Research, "The actual size of the data (analyzed and indexed text, knowledge bases, etc.) used for candidate answer generation and evidence evaluation is under 1TB." For performance reasons, various subsets of the data are replicated in RAM on different functional groups of cluster nodes. The entire system is self-contained, Watson is NOT going to the internet searching for answers.