As a consultant, I am often asked to help design the architecture for the information infrastructure. A usefulanalogy to gather requirements and preferences is the difference between area rugs
and wall-to-wall carpeting
. Arearugs are not secured to the floor and cover only a portion of the floor area. Carpets are generally tacked or cemented to the floor, often with an underlay of cushion padding, stretched across the entire floor surface, out to all four walls of each room.
Each has its pros and cons, and often is a matter of preference. Some people like area rugs because they can choosea different style for each room, match the decor and color scheme of furniture, and use these to define each livingspace. Ever since paleolithic man put animal skins on the floor of their cave, people recognize that cold, hard andugly floors could be covered up with something soft and more attractive.Others prefer wall-to-wall carpeting because they want to walk around the house barefoot, have their young children crawl on their hands and knees, and give the entire house a unified look and feel. This is often an inexpensive option when compared against the cost of individual rugs.
The same is true for an information infrastructure. For some, they prefer the "area rug" approach: this style ofstorage for their email, this other type of storage for their databases, and perhaps a third for their unstructuredfile systems. When customers ask what storage would I recommend for their SAP application, or their Microsoft Exchangeemail environment, or their Business Intelligence (BI) software, I recognize they are taking this "area rug" approach.
Like area rugs, having different storage can focus on specific attributes of the workload characteristics. It alsoinsulates against company-wide changes, the dreaded "rip-and-replace" of replacing all of your storage with somethingfrom a different vendor. With "area rug" storage, you can support a dual-vendor or multi-vendor strategy, and upgrade or replace each on its own schedule.
Thanks to open standards and industry-standard benchmarks, changing out one storage solution for another is assimple as rolling up an area rug, and putting another one in its place that is similar in size dimensions.
|Others may prefer "wall-to-wall carpeting" approach: one disk system type, one tape library type,one network type, that provides unified management and minimizes the needs for unique skills. Generally, the choice of NAS, SAN or iSCSI infrastrucutre is done company-wide, and might strongly influence the set of products that will support that decision. For example, those with a mix of mainframe and distributed servers looking for SAN-attached storage may look at an [IBM System Storage DS8000] and [TS3500 tape library] that can provide support for FICON and FCP.|
Those looking at NAS or iSCSI might consider the IBM System Storage N series products, "unified storage" supporting iSCSI, FCP and NAS protocols. If you want the "wall-to-wall" to stretch across all the sites in your globally integrated enterprise, IBM's scalable NAS product, Scale-Out File Services[SoFS], provides a global name spacein combination with a clustered file system that provides incredible scalability and performance based on field-proven technology used by the majority of the [Top 100 supercomputer] deployments.
IBM can help you design an information infrastructure that fits either approach.
technorati tags: IBM, DS8000, TS3500, NAS, SAN, iSCSI, FCP, FICON, mainframe, distributed, SoFS, supercomputer
I got some interesting queries about IBM's Scale-Out File Services [SoFS
] that I mentioned in my post yesterday [Area rugs versus Wall-to-Wall carpeting
]. I thought I would provide some additional details of the product.
SoFS combines three key features: a global namespace, a clustered file system, and Information LifecycleManagement (ILM). Let's tackle each one.
- Global Name Space
A long time ago, IBM acquired a company called Transarc that developed Andrew File System (AFS) and DistributedFile System (DFS). These both provided global namespace capability, meaning that all of your files could beaccessible from a single URL file tree. Imagine if you have data centers in Tucson, Austin, Raleigh and Chicago.Normally, to access files from each city, you would have to mount a unique IP address for that location, and thento get to files in a different city, you'd have to mount a second, and so on. But with a global namespace, you could mount a single drive letter Z: and access files simply by using Z:/Tucson/abc or Z:/Austin/xyz. IBM uses its DFS to make this happen.
Just because you have access to a global namespace doesn't give you read/write authority to every file. IBM SoFS has full NTFS Access Control List (ACL) support, so that only those who can read or write data can access the files. A "hide unreadable" feature provideswhat I like to call "parental controls": you don't even get to see on your directly list any file or subdirectory that you don't have access to. For example, if there is a directory with 50 projects, but you only have authority tothree projects, then you only see the three subdirectories related to those projects, and nothing else.
There are other ways to get a global namespace. IBM also offers the IBM System Storage N series Virtual FileManager, Brocade offers Storage/X, and F5 acquired Acopia. These all work by putting a box in front of a set ofindependent NAS storage units, and giving you a single mount point to represent all of the file systems managedbehind the scenes. This however can sometimes be a bottleneck for performance.
- Clustered File System
Often, when you have a lot of data in one place, you are also expected to deliver that data to lots of clientswith relatively good performance. Otherwise, end users revolt and get their own internal direct attach storage.To solve this, you need a clustered architecture that provides access in parallel to the data.
First, we start with a node that is optimized for CIFS and NFS access. We have clocked our node to run CIFS at577 MB/sec, and NFS at 880 MB/sec, through a 10GbE pipe between a single client and a single SoFS node. Comparethat to the 400 MB/sec you get today with 4Gbps FCP, or the 800 MB/sec you will get if you upgrade to 8 GbpsFCP, and quickly you recognize that this is comparable performance for demanding workloads.
Then, you combine multiple nodes together, and have them all be able to read/write any file in the file system, andfront-end that with a load-balancing Virtual IP address (VIPA) that spreads the requests around, and you've gotyourself a lean and mean machine for accessing data.
In 2005, IBM delivered[ASC Purple] with the world's fastest file system. 1536 nodeswere able to access billions of files in the 2 Petabyte of data. The record of 126 GB/sec access to a single filewas set, and has yet to be beaten by any other vendor since.This same file system is used in SoFS, as well as a variety of other IBM storage offerings.
The back-end storage can be SAS or FC-attached, from the DS3200 to our mighty DS8300 Turbo, as well as ourIBM System Storage DCS9550 and SAN Volume Controller (SVC), and a variety of tape libraries.
- Information Lifecycle Management
Lastly, we get to ILM. With SoFS, you can have different tiers of storage, high-speed SAS or FC disk, low-speedFATA and SATA disk, and even tape. Policy-based automation allows you to place any file onto any disk tier whencreated, and other policies can migrate or delete the data trigged by certain threshold, age, or other criteria.The advantage is that this is on a file by file basis, so Z:/Tucson/Project could have a bunch of files, some ofthem on my FC disk, some of them on my SATA, and some on tape. The file path doesn't change when they move, anddifferent files in the same directory can be on different tiers.
Data movement is bi-directional. If you know you will be using a set of files for an upcoming job, say perhapsquarter-end or year-end processing, you can pre-fetch those files from tape and move them to your fastest disk pool.
There is also integrated backup support. Typically, a large NAS environment is difficult to backup. Traditionalmethods take days to scan the directory tree looking for files in need of backup. A single SoFS node can scana billion files in 95 minutes, and 8 nodes in a cluster can scan a billion files in under 15 minutes.
Recovery is even more impressive. When you recover, SoFS brings back the entire directory structure first, withall the file names in place. This would make it appear that all the data is restored, but actually it is still on tape.When you access individual files, it will then drive the recovery of that file, so your applications and end usersbasically determine the priority of the recovery. Traditional methods would wait until every file was restoredbefore letting anyone access the system.
SoFS is part of IBM's [Blue Cloud] initiativethat was launched last November 2007. Of course, IBM isn't the only one competing in this space. HDS has partneredwith BlueArc, HP has acquired PolyServe, and Sun acquired CFS for their Lustre file system. Isilon and Exanet arestart-up companies with some offerings. EMC acquired Rainfinity,and have hinted at a Hulk/Maui project that they might deliver later this year or perhaps in 2009, but by thenmight be a dollar-short and a day-late.
But why wait? IBM SoFS is available today and is orders of magnitude more scalable!
technorati tags: IBM, SoFS, Acopia, VFM, Brocade, ILM, global namespace, clustered, file system, disk, tape, storage, system, CIFS, NFS, NAS, NTFS, ACL, DFS, AFS, Transarc, ASC Purple, DS3200, SAS, FC, FCP, DS8300, Turbo, DCS9550, SVC, FATA, SATA, nodes, backup, restore, recovery, Blue Cloud, cloud computing, PolyServe, HDS, BlueArc, HP, Sun, CFS, Lustre, Isilon, Exanet, EMC, Rainfinity, Hulk, Maui
No, this is not an announcement about myself moving to Nepal.
My friends over at OLE Nepal are [looking for a Super SysAdmin]willing to live in Nepal for five months and help out with their project to help the students in the localschools there. I think this might be a great opportunity for someone to help changethe world. Those of you who have read my past blog posts about the One Laptop per Child [OLPC], such as [Understanding the LAMP platform] and [Supporting OLPC Schools with LAMP stacks] may understand the type of work involved.
- You dream in Bash
- IPv4, IPv6, Wireless Mesh networking? No problem! You know linux networking inside and out
- Extensive knowledge of BIND, DHCPD, Squid, Apache, security, etc.
- Experience working with [Moodle] would be most excellent (it is basically a PHP web application that maintains MySQL databases for lesson plans, homework assignments and other school related information)
- Adept with Python scripting or could learn it quickly. OLPC has standardized on Python for scripting (although knowledge in Perl and PHP won't hurt either)
- You look to implement a practical solution that less skilled sysadmins can easily maintain over a cooler but more complicated solution.
- You play well with others. You don’t alienate collaborators with rude e-mails that assert your technical superiority (even though you are)
- Your primary concern is meeting the educational needs of kids and teachers. Your rate technical awesomeness a distant second to meeting those critical needs.
I've been working with Dev, Bryan and Sulochan for the past three months (remotely here from Tucson, AZ)but we've come to a point where we need on-site expertise. I will continue to provide remote support.
Given the number of readers who have contacted me over the past year looking for an IT job (or a different job because they are not happy where they are), this could be an amazing experience.
technorati tags: OLE Nepal, OLPC, Bash, Linux, IPv6, Mesh, networking, Squid, Apache, security, Moodle, LAMP, PHP, Perl, Python
I am always amused in the manner the IT industry tries to solve problems. Take, for example, theprocess of backups. The simplest approach is to backup everything, and keep "n" versions of that.Simple enough for a small customer who has only a handful of machines, but does not scale well. Inmy post [Times a Million
],I coined the phrase "laptop mentality", referring to people's inability to think through solutions in large scale.
Apparently, I am not alone.Steve Duplessie (ESG) wrote in his post[Random Thoughts]:
"I may even get to stop yelling at people to stop doing full backups every week on non-changing data (which is 80 %+) just because that's how they used to do it. They won't have a choice. You can't back up 5X your current data the way you do (or don't) today."
Hu Yoshida (HDS) does a great job explaining that thereare three ways to perform deduplication for backups:
- Pre-processing. Have the backup software not backup unchanged data.
- Inline processing. Have an index to filter the output of the backup as it sends data to storage.
- Post-processing. Have the receiving storage detect duplicates and handle them accordingly.
Here's an excerpt from his post[Deduplication Ratios]:
"A full backup of 1TB data base tablespace is taken on day one. The next day another full backup is taken and only 2GB of that backup has any changes.
Using traditional full backup approaches after 2 nights, the backup capacity required is 2 x 1TB = 2TB
One method of calculating de-duplication ratios could yield a low ratio:
- Total de-duplicated backup capacity used = 1TB + 2GB = 1.002TB
- If the de-duplication ratio compares the amount of total physical storage used to the total amount that would have been used by traditional backup methods, the ratio = 2TB / 1.002TB = approximately 2:1
Another method of calculating de-duplication ratios could yield a high ratio:
- Total de-duplicated backup capacity used still = 1.002TB
- If the de-duplication ratio compares the amount of data stored in the most recent (second) backup to the amount that would have been used by traditional backup methods, the ratio 1TB / 2GB = 1000GB / 2GB = 500:1"
While IBM also offers deduplication in the IBM System Storage N series disk systems, I find that for backup, itis often more effective to apply best practices via IBM Tivoli Storage Manager (TSM). Let's take a look at some:
- Exclude Operating System files
Why take full backups of your operating system every day? Yes, deduplication will find a lot to reduce fromthis, but best practices would exclude these. TSM has an include/exclude list, and the default version excludesall the operating system files that would be recovered from "bare machine recovery" or "new system install"procedures. Often, if the replacement machine has different gear inside, your OS backups aren't what you need,and a fresh OS install may determine this and install different drivers or different settings.
- Exclude Application programs
Again, yes if there are several machines running the same application, you probably have opportunity for deduplication. However, unless you match these up with the appropriate registry or settings buried down in theoperating system, recovering just application program files may render an unusable system. Applications are bestinstalled from a common source that are either "pushed" through software distribution, or "pulled" from an application installation space.
If you have TB-sized databases, and are only doing full backups daily to protect it, have I got a solution for you.IBM and others have software that are "application-aware" and "database-aware" enough to determine what haschanged since the last backup and copy only that delta. Taking advantage of the TSM Application ProgrammingInterface (API) allows for both IBM and third party tools to take these delta backups correctly.
- User Files
Which leaves us with user files, which are often unique enough on their own from the files of other users,that would not benefit from file-level deduplication. Backing up changed data only, as TSM does with its patented ["progressive incremental backup"] method, generally gets most of the benefits described by deduplication, without having to purchase storage hardware features.
Of course, if two or more users have identical files, the question might be why these are not stored on acommon file share. NAS file share repositories can greatly reduce each user keeping their own set of duplicates.It is interesting that some block-oriented deduplication,such as that found in the IBM System Storage N series, can get some benefit because some user files are oftenderivatives of other files, and there might be some 4 KB blocks of data in common.
Last November, I visited a customer in Canada. All of their problems were a direct result of taking full backupsevery weekend. It put a strain on their network; it used up too many disk and tape resources; and it took too long tocomplete. They asked about virtual tape libraries, deduplication, and anything else that could help them. The answer was simple: switch to IBM Tivoli Storage Manager and apply best practices.
technorati tags: Steve Duplessie, ESG, Hu Yoshida, HDS, deduplication, N series, application-aware, database-aware, database, tablespace, best practice, Tivoli, Storage Manager, TSM, progressive, incremental, backup
It's been a while since I've talked about [Second Life
The latest post on eightbar[Spimes, Motes and Data centers]discusses IBM's use of virtual world technology to analyze data centers in three dimensions.New World Note asks[What's The Point Of 3D Data Centers?]One would think that a simple monitoring tool based on a two-dimensional floor plan would be enough to evaluate a data center.
Enter Michael Osias, IBM (a.k.a Illuminous Beltran in Second Life). Some of the leading news sites havebegun to notice some 3D data centers that he has helped pioneer. UgoTrade writes up an article aboutMichael and the media attention in [The Wizard of IBM's 3DData Centers].
Of course, in presenting these "Real Life/Second Life" (RL/SL) interactive technologies, IBM is sometimes the target of ridicule. Why? Because IBM is 10 years ahead of everyone else. So, are there aspects of a data center where 3D interfaces makes sense? I think there is.
- Topology Viewer
IBM TotalStorage Productivity Center has an awesome "topology viewer" that shows what servers are connectedto which switches, to which disk systems and tape libraries. This is all done in a 2D diagram, generated dynamicallywith data discovered through open standard interfaces, similar to what you might draw manually with toolslike Visio. Imagine, however, howmore powerful if it were a 3D viewer, with virtual equipment mapped to the physical location of each pieceof hardware on the data center floor, including the position on the rack and location on the data center floor.
- Temperature Flow
Designing computer room air conditioning (CRAC) systems is actually a three dimensional problem. Cold air isfed underneath the raised floor, comes up through strategically placed "vent" tiles, taken in the front ofeach rack. Hot air comes out the back of each rack, and hopefully finds ceiling duct intake to get cooled again.The temperature six inches off the floor is different than the temperature six feet off the floor, and 3Dmonitor tools could be helpful in identifying "hot spots" that need attention. In this case "spimes" representsensors in the 3D virtual world, able to report back information to help diagnose problems or monitor events.
- Server consolidation
After many people left the mainframe in favor of running a single application per distributed server, the pendulumhas finally swung back. Companies are discovering the many benefits of changing this behavior. "Re-centralization" is the task at hand. Thanks to virtualization of servers, networks and storage, sharing common resources canonce again claim the benefits of economies of scale. In many cases, servers work together in collective unitsfor specific applications that might benefit better if consolidated together onto the same equipment.
IBM's "New Enterprise Data Center" vision recognizes that people will need to focus on the management aspectsof their IT infrastructure, and 3D virtual world technologies might be an effective way to getthe job done.
technorati tags: secondlife, eightbar, spimes, motes, 3D, data center, virtual world, IBM, TotalStorage, Productivity Center, CRAC, re-centralization, New Enterprise Data Center