This week, I am in beautiful Sao Paulo, Brazil, teaching Top Gun class to IBM Business Partners and sales reps. Traditionally, we have "Tape Thursday" where we focus on our tape systems, from tape drives, to physical and virtual tape libraries. IBM is the number #1 tape vendor, and has been for the past eight years.
(The alliteration doesn't translate well here in Brazil. The Portuguese word for tape is "fita", and Thursday here is "quinta-feira", but "fita-quinta-feira" just doesn't have the same ring to it.)
In the class, we discussed how to handle common misperceptions and myths about tape. Here are a few examples:
- Myth 1: Tape processing is manually intensive
In my July 2007 blog post [Times a Million], I coined the phrase "Laptop Mentality" to describe the problem most people have dealing with data center decisions. Many folks extend linearly their experiences using their PCs, workstations or laptops to apply to the data center, unable to comprehend large numbers or solutions that take advantage of the economies of scale.
For many, the only experience dealing with tape was manual. In the 1980s, we made "mix tapes" on little cassettes, and in the 1990s we recorded our favorite television shows on VHS tapes in the VCR. Today, we have playlists on flash or disk-based music players, and record TV shows on disk-based video recorders like Tivo. The conclusion is that tapes are manual, and disk are not.
Manual processing of tapes ended in 1987, with the introduction of a silo-like tape library from StorageTek. IBM quickly responded with its own IBM 3495 Tape Library Data Server in 1992. Today, clients have many tape automation choices, from the smallest IBM TS2900 Tape Autoloader that has one drive and nine cartridges, all the way to the largest IBM TS3500 multiple-library shuttle complex that can hold exabytes of data. These tape automation systems eliminate most of the manual handling of cartridges in day-to-day operations.
- Myth 2: Tape media is less reliable than disk media
For any storage media to be unreliable is to return the wrong information that is different than what was originally stored. There are only two ways for this to happen: if you write a "zero" but read back a "one", or write a "one" and read a "zero". This is called a bit error. Every storage media has a "bit error rate" that is the average likelihood for some large amount of data written.
According to the latest [LTO Bit Error rates, 2012 March], today's tape expects only 1 bit error per 10E17 bits written (about 100 Petabytes). This is 10 times more reliable than Enterprise SAS disk (1 bit per 10E16), and 100 times more reliable than Enterprise-class SATA disk (1 bit per 10E15).
Tape is the media used in "black boxes" for airplanes. When an airplane crashes, the black box is retrieved and used to investigate the causes of the crash. In 1986, the Space Shuttle Challenger exploded 73 seconds after take-off. The tapes in the black box sat on the ocean floor for six weeks before being recovered. Amazingly, IBM was able to successfully restore [90 percent of the block data, and 100 percent of voice data].
- Myth 3: Most tape restores fail
Why do people still believe that most tape restores fail? Curtis Preston, on his Backup Central blog, has a great post [Gartner Never Said 71 percent of Tape Restores Fail].
Analysts are quite upset when they are quoted out of context, but in this case, Gartner never said anything closely similar to this. Nor did the other analysts that Curtis investigated for similar claims. What Garnter did say was that disk provides an attractive alternative storage media for backup which can increase the performance of the recovery process.
Back in the 1990s, Savur Rao and I developed a patent to help backup DB2 for z/OS by using the FlashCopy feature of IBM's high-end disk system. The software method to coordinate the FlashCopy snapshots with the database application and maintain multiple versions was implemented in the DFSMShsm component of DFSMS. A few years later, this was part of a set of patents IBM cross-licensed to Microsoft for them to implement a similar software for Windows called Data Protection Manager (DPM). IBM has since introduced its own version for distributed systems called IBM Tivoli FlashCopy Manager that runs not just on Windows, but also AIX, Linux, HP-UX and Solaris operating systems.
Curtis suspects the "71 percent" citation may have been propogated by an ambitious product manager of Microsoft's Data Protection Manager, back in 2006, perhaps to help drive up business to their new disk-based backup product. Certainly, Microsoft was not the only vendor to disparage tape in this manner.
A few years ago, an [EMC failure brought down the State of Virginia] due to not just a component failure it its production disk system, but then made it worse by failing to recover from the disk-based remote mirror copy. Fortunately, the data was able to be restored from tape over the next four days. If you wonder why nobody at EMC says "Tape is Dead" anymore, perhaps it is because tape saved their butts that week.
(FTC Disclosure: I work for IBM and this post can be considered a paid, celebrity endorsement for all of the IBM tape and software products mentioned on this post. I own shares of stock in both IBM and Google, and use Google's Gmail for my personal email, as well as many other Google services. While IBM, Google and Microsoft can be considered competitors to each other in some areas, IBM has working relationships with both companies on various projects. References in this post to other companies like EMC are merely to provide illustrative examples only, based on publicly available information. IBM is part of the Linear Tape Open (LTO) consortium.)
Last year, Google lost the email data for half a million Gmail accounts due to a software error. Once again, tape came to the rescue, with [Google restoring lost Gmail data from tape backups].
- Myth 4: Vendors and Manufacturers are no longer investing in tape technology
IBM and others are still investing Research and Development (R&D) dollars to improve tape technology. What people don't realize is that much of the R&D spent on magnetic media can be applied across both disk and tape, such as IBM's development of the Giant Magnetoresistance read/write head, or [GMR] for short.
Most recently, IBM made another major advancement with tape with the introduction of the Linear Tape File Systems (LTFS). This allows greater portability to share data between users, and between companies, but treating tape cartridges much like USB memory sticks or pen drives. You can read more in my post [IBM and Fox win an Emmy for LTFS technology]!
Next month, IBM celebrates the 60th anniversary for tape. It is good to see that tape continues to be a vibrant part of the IT industry, and to IBM's storage business!
technorati tags: IBM, Google, Microsoft, EMC, Brazil, LTO, TS2900, TS3500, Space Shuttle, Challenger
On my last blog post [Is this what HDS tells our mainframe clients?], I poked fun at Hu Yoshida's blog post that contained a graphic with questionable results. Suddenly, the blog post disappeared altogether. Poof! Gone!
Just so that I am not accused of taking a graph out of context, here is Hu's original post, in its entirety:
"Since my last post on Storage Performance Efficiency, Claus wrote on the use of HDP, Hitachi Dynamic Provisioning and HDT, Hitachi Dynamic Tiering for mainframes on Virtual Storage Platform (VSP). Naturally, this prompted me to think of the specific performance efficiency implications for mainframes.
HDP brings the performance benefits of automated wide striping and HDT automatically keeps the hot pages of data on the highest performance tier of storage for mainframes, just as it does for open systems. There are differences between open systems and mainframe implementation due to mainframe CKD and CCHHR formats for instance, the page size is optimized for mainframe storage formats and storage reclamation must be host initiate. For more information check out our website: http://www.hds.com/assets/pdf/how-to-apply-latest-advances-in-hitachi-mainframe-storage.pdf
There are also additional performance efficiencies specific for mainframes.
Mainframe HDP is the foundation for Extended Addressable Volumes, which increases the size of 3390 volumes from 65,520 cylinders to 262,668 cylinders. This, along with HyperPAV--which facilitates multiple accesses to a volume, addressing the problem of queuing on a very large volume with a single UCB--enhances throughput with many more concurrent I/O operations.
The thin provisioning of HDP also increases the performance of mainframe functions that move, copy, or replicate these thin volumes like Concurrent Copy, FlashCopy V02, and HUR, since the actual volumes are smaller.
If you have mainframes, check out the capacity and performance efficiency of VSP with HDP and HDT.
For other posts on maximizing storage and capacity efficiencies, check these out:
At this point, you might be wondering: "If Hu Yoshida deleted his blog post, how did Tony get a copy of it? Did Tony save a copy of the HTML source before Hu deleted it?" No. I should have, in retrospect, in case lawyers got involved. It turns out that deleting a blog post does not clear the various copies in various RSS Feed Reader caches. I was able to dig out the previous version from the vast Google repository. (Many thanks to my friends at Google!!!).
The graph itself was hosted separately has been deleted, but it was just taken from slide 10 of the HDS presentation [How to Apply the Latest Advances in Hitachi Mainframe Storage], so it was easy to recreate.
(Lesson to all bloggers: If you write a blog post, and later decide to remove it for whatever legal, ethical, moral reasons, it is better to edit the post to remove offending content, and add a comment that the post was edited, and why. Shrinking a 700-word article down to 'Sorry Folks - I decided to remove this blog post because...' would do the trick. This new edited version will then slowly propagate across to all of the RSS Feed Reader caches, eliminating most traces to the original. Of course, the original may have been saved by any number of your readers, but at least if you have an edited version, it can serve as the official or canonical version.)
Perhaps there was a reason why HDS did not want to make public the FUD its sales team use in private meetings with IBM mainframe clients. Whatever it was, this appears to be another case where the cover-up is worse than the original crime!
technorati tags: HDS, Hu Yoshida, VSP, EAV
Five years ago, I sprayed coffee all over my screen from something I read on a blog post from fellow blogger Hu Yoshida from HDS. You can read what cased my reaction in my now infamous post [Hu Yoshida should know better]. Subsequently, over the years, I have disagreed with Hu on a variety of of topics, as documented in my 2010 blog post [Hu Yoshida Does It Again].
(Apparently, I am not alone, as the process of spraying one's coffee onto one's computer screen while reading other blog posts has been referred to as "Pulling a Tony" or "Doing a Tony" by other bloggers!)
Fortunately, my IBM colleague David Sacks doesn't drink coffee. Last month, David noticed that Hu had posted a graph in a recent blog entry titled [Additional Storage Performance Efficiencies for Mainframes], comparing the performance of HDS's Virtual Storage Platform (VSP) to IBM's DS8000.
For those not familiar with disk performance graphs, flatter is better, lower response time and larger IOPS are always desired. This graph implies that the HDS disk system is astonishingly faster than IBM's DS8000 series disk system. Certainly, the HDS VSP qualifies as a member of the elite [Super High-End club] with impressive SPC benchmark numbers, and is generally recognized as a device that works in IBM mainframe environments. But this new comparison graph is just ridiculous!
(Note: While SPC benchmarks are useful for making purchase decisions, different disk systems respond differently to different workloads. As the former lead architect of DFSMS for z/OS, I am often brought in to consult on mainframe performance issues in complex situations. Several times, we have fixed performance problems for our mainframe clients by replacing their HDS systems with IBM DS8000 series!)
Since Hu's blog entry contained very little information about the performance test used to generate the graph, David submitted a comment directly to Hu's blog asking a few simple questions to help IBM and Hu's readers determine whether the test was fair. Here is David's comment as submitted:
(Disclosure: I work for IBM. This comment is my own.)
I was quite surprised by the performance shown for the IBM DS8000 in the graph in your blog. Unfortunately, you provided very little detail about the benchmark. That makes it rather difficult (to say the least) to identify factors behind the results shown and to determine whether the comparison was a fair one.
Of the little information provided, an attribute that somewhat stands out is that the test appears to be limited to a single volume at least, that's my interpretation of "LDEV: 1*3390-3"? IBM's internal tests for this kind of case show far better response time and I/Os per second than the graph you published.
Here are a few examples of details you could provide to help readers determine whether the benchmark was fair and whether the results have any relevance to their environment.
- What DS8000 model was the test run on? (the DS8000 is a family of systems with generations going back 8 years. The latest and fastest model is the DS8800.)
- What were the hardware and software configurations of the DS8000 and VSP systems, including the number and speed of performance-related components?
- What were the I/O workload characteristics (e.g., read:write ratio and block size(s))?
- What was the data capacity of each volume? (Allocated and used capacity.)
- What were the cache sizes and cache hit ratios for each system? (The average I/O response times under 1.5 milliseconds for each system imply the cache hit ratios were relatively high.)
- How many physical drives were volumes striped across in each system?"
Unlike my blog on IBM, HDS bloggers like Hu are allowed to reject or deny comments before they appear on his blog post. We were disappointed that HDS never posted David's comment nor responded to it. That certainly raises questions about the quality of the comparison.
So, perhaps this is yet another case of [Hitachi Math], a phrase coined by fellow blogger Barry Burke from EMC back in 2007 in reference to outlandish HDS claims. My earliest mention was in my blog post [Not letting the Wookie Win].
By the way, since the test was about z/OS Extended Address Volumes (EAV), it is worth mentioning that IBM's DS8700 and DS8800 support 3390 volume capacities up to 1 TB each, while the HDS VSP is limited to only 223 GB per volume. Larger volume capacities help support ease-of-growth and help reduce the number of volumes storage administrators need to manage; that's just one example of how the DS8000 series continues to provide the best storage system support for z/OS environments.
Personally, I am all for running both IBM and HDS boxes side-by-side and publishing the methodology, the workload characteristics, the configuration details, and the results. Sunshine is always the best disinfectant!
technorati tags: IBM, DS8000, DS8800, HDS, Hu Yoshida, USP, VSP, mainframe, EAV
Well, it's Tuesday, and you know what that means... IBM announcements!
In today's environment, clients expect more from their storage, and from their storage provider. The announcements span the gamut, from helping to use Business Analytics to analyze Big Data for trends, insights and patterns, to managing private, public and hybrid cloud environments, all with systems that are optimized for their particular workloads.
There are over a dozen different announcements, so I will split these up into separate posts. Here is part 1.
- IBM Scale Out Network Attach Storage (SONAS) R1.3
I have covered [IBM SONAS] for quite some time now. Based on IBM's General Parallel File System (GPFS), this integrated system combines servers, storage and software into a fully functional scale-out NAS solution that support NFS, CIFS, FTP/SFTP, HTTP/HTTPS, and SCP protocols. IBM continues its technical leadership in the scale-out NAS marketplace with new hardware and software features.
The hardware adds new disk options, with 900GB SAS 15K RPM drives, and 3TB NL-SAS 7200 RPM drives. These come in 4U drawers of 60 drives each, six ranks of ten drives each. So, with the high-performance SAS drives that would be about 43TB usable capacity per drawer, and with the high-capacity NL-SAS drives about 144TB usable. You can have any mix of high-performance drawers and high-capacity drawers, up to 7200 drives, for a maximum usable capacity of 17PB usable (21PB for those who prefer it raw). This makes it the largest commercial scale-out NAS in the industry. This capacity can be made into one big file system, or divided up to 256 smaller file systems.
In addition to snapshots of each file system, you can divide the file system up into smaller tree branches and snapshot these independently as well. The tree branches are called fileset containers. Furthermore, you can now make writeable clones of individual files, which provides a space-efficient way to create copies for testing, training or whatever.
Performance is improved in many areas. The interface nodes now can support a second dual-port 10GbE, and replication performance is improved by 10x.
SONAS supports access-based enumeration, which means that if there are 100 different subdirectories, but you only have authority to access five of them, then that's all you see, those five directories. You don't even know the other 95 directories exist.
I saved the coolest feature for last, it is called Active Cloud Engine™ that offers both local and global file management. Locally, Active Cloud Engine placement rules to decide what type of disk a new file should be placed on. Management rules that will move the files from one disk type to another, or even migrates the data to tape or other externally-managed storage! A high-speed scan engine can rip through 10 million files per node, to identify files that need to be moved, backed up or expired.
Globally, Active Cloud Engine makes the global namespace truly global, allowing the file system to span multiple geographic locations. Built-in intelligence moves individual files to where they are closest to the users that use them most. This includes an intelligent push-over-WAN write cache, on-demand pull-from-WAN cache for reads, and will even pre-fetch subsets of files.
No other scale-out NAS solution from any other storage vendor offers this amazing and awesome capability!
- IBM® Storwize® V7000
Last year, we introduced the [IBM Storwize V7000], a midrange disk system with block-level access via FCP and iSCSI protocols. The 2U-high control enclosure held two cannister nodes, a 12-drive or 24-drive bay, and a pair of power-supply/battery UPS modules. The controller could attach up to nine expansion enclosures for more capacity, as well as virtualize other storage systems. This has been one of our most successful products ever, selling over 100PB in the past 12 months to over 2,500 delighted customers.
The 12-drive enclosure now supports both 2TB and 3TB NL-SAS drives. The 24-drive enclosures support 200/300/400GB Solid-State Drives (SSD), 146 and 300GB 15K RPM drives, 300/450/600GB 10K RPM drives, and a new 1TB NL-SAS drive option. For those who want to set up "Flash-and-Stash" in a single 2U drawer, now you can combine SSD and NL-SAS in the 24-drive enclosure! This is the perfect platform for IBM's Easy Tier sub-LUN automated tiering. IBM's Easy Tier is substantially more powerful and easier to use than EMC's FAST-VP or HDS's Dynamic Tiering.
Last week, at Oracle OpenWorld, there were various vendors hawking their DRAM/SSD-only disk systems, including my friends at Texas Memory Systems, Pure Storage, and Violin Memory Systems. When people came to the IBM booth to ask what IBM offers, I explained that both the IBM DS8000 and the Storwize V7000 can be outfitted in this manner. With the Storwize V7000, you can buy as much or little SSD as you like. You do not have to buy these drives in groups of 8 or 16 at a time.
The Storwize V7000 is the sister product of the IBM SAN Volume Controller, so you can replicate between one and the other. I see two use cases for this. First, you might have a SVC at a primary location, and decide to replicate just the subset of mission-critical production data to a remote location, and use the Storwize V7000 as the target device. Secondly, you could have three remote or branch offices (ROBO) that replicate to a centralized data center SAN Volume Controller.
Lastly, like the SVC, the Storwize V7000 now supports clustering so that you can now combine multiple control enclosures together to make a single system.
- IBM® Storwize® V7000 Unified
Do you remember how IBM combined the best of SAN Volume Controller, XIV and DS8000 RAID into the Storwize V7000? Well, IBM did it again, combining the best of the Storwize V7000 with the common NAS software base developed for SONAS into the new "Storwize V7000 Unified".
You can upgrade your block-only Storwize V7000 into a file-and-block "Storwize V7000 Unified" storage system. This is a 6U-high system, consisting of a pair of 2U-high file modules connected to a standard 2U-high control enclosure. Like the block-only version, the control enclosure can attach up to nine expansion enclosures, as well as all the same support to virtualize external disk systems. The file modules combine the management node, interface node and storage node functionality that SONAS R1.3 offers.
What exactly does that mean for you? In addition to FCP and iSCSI for block-level LUNs, you can carve out file systems that support NFS, CIFS, FTP/SFTP, HTTP/HTTPS, and SCP protocols. All the same support as SONAS for anti-virus checking, access-based enumeration, integrated TSM backup and HSM functionality to migrate data to tape, NDMP backup support for other backup software, and Active Cloud Engine's local file management are all included!
- IBM SAN Volume Controller V6.3
The SAN Volume Controller [SVC] increases its stretched cluster to distances up to 300km. This is 3x further than EMC's VPLEX offering. This allows identical copies of data to be kept identical in both locations, and allows for Live Partition Mobility or VMware vMotion to move workloads seamlessly from one data center to another. Combining two data centers with an SVC stretch cluster is often referred to as "Data Center Federation".
The SVC also introduces a low-bandwidth option for Global Mirror. We actually borrowed this concept from our XIV disk system. Normally, SVC's Global Mirror will consume all the bandwidth it can to keep the destination copy of the data within a few seconds of currency behind the source copy. But do you always need to be that current? Can you afford the bandwidth requirements needed to keep up with that? If you answered "No!" to either of these, then the low-bandwidth option is you. Basically, a FlashCopy is done on the source copy, this copy is then sent over to the destination, and a FlashCopy is made of that. The process is then repeated on a scheduled basis, like every four hours. This greatly reduces the amount of bandwidth required, and for many workloads, having currency in hours, rather than seconds, is good enough.
I am very excited about all these announcements! It is a good time to be working for IBM, and look forward to sharing these exciting enhancements with clients at the Tucson EBC.
technorati tags: IBM, SONAS, GPFS, SAS, NL-SAS, Active Cloud Engine, Global+Namespace, Storwize+V7000, V7000U, V7000 Unified, block-only, block-and-file, SVC, SSD, Easy Tier, Flash-and-Stash, Texas Memory Systems, Pure Storage, Violin Memory
I gotten several emails expressing worry that I have fallen off the face of th earth. The last two weeks have been educational and eye-opening for me. I can't provide details in my blog, so I will just say that it involved government agencies that IBM refers to as "dark accounts", and that I am now back safely in the USA. Between adjusting to time zone differences, ridiculously long hours, and restricted access to the internet, I was unable to blog lately.
Instead, I will resume my coverage of the [IBM System Storage Technical University 2011]. The "Solutions Expo" runs Monday evening through Wednesday lunch. This is a chance for people to explore all the solutions that are part of IBM's large "eco-system" for IBM System storage and System x products. There were several sponsors for this event.
As is often the case at these conferences, the various booths hand out fun items. The hot items this year were tie-dyed tee-shirts from Qlogic, and propeller beanies from the IBM rack and power systems team. Here is Amanda, one of the bartenders showing off the latter.
After the expo on Tuesday night, my friends at [Texas Memory Systems] held an after-party. Unlike the pens, tee-shirts and keychains at the Expo, these guys had a raffle for real storage products. Here is Erik Eyberg handing out a RamSan PCIe card, valued at $14,000 or so. IBM recently certified the TMS RamSan as External SSD storage for the IBM SAN Volume Controller (SVC). The SVC can optimize performance using this for automated sub-LUN tiering with the IBM System Storage Easy Tier feature.
technorati tags: IBM, Storage, University, Qlogic, TMS, RamSan, SVC, Easy Tier