Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
It is another year and time for yet another try at predicting the future in our complex storage industry with myriad variables. The year in review for 2011 can be summarized with the same two words I used for 2010 -- stagnation and consolidation.
Stagnation, as there has been limited innovation. This is totally understandable given the continued economic situation around the world, but things are not quite as bad as they were in 2009 and 2010, and there are significant signs of new investment. Consolidation continues for Tier-2 and Tier-3 storage players, which have been acquired right and left, as the large vendors prepare for the economic upturn. As usual, I was a bit more optimistic than I should have been, and some of the things that I predicted would happen in 2011 have not. However, many, I believe, will happen by the end of 2Q12, albeit a day late and $1 short as the saying goes.
Remember, all of what I have written was true the day I wrote it, but nothing in our industry stays the same for long.
2011 Storage Predictions
- Declustered RAID will become mainstream because of 4 TB drives: This did not happen. A few vendors have recently released some products, but this is not in any way close to mainstream and 4 TB drives are not available on the market yet. I get a 0.
- Large RAID vendors will support FCoE targets: A number of RAID and appliance vendors have finally gotten on board with this technology. This is true for enough vendors to earn me a 1.
- PCIe 3.0 doubles performance over PCIe 2.0 per lane, and today even midrange motherboards have PCIe 2.0 16-lane slots: There are PCIe 3.0 x16 motherboards, but no x16 cards, at least that I know about. Since there are no x16 cards, who knows if the x16 slots run a rate? To be fair in scoring, I get a 0.
- SSDs will achieve 2,000,000 IOPS performance: Nope, the fastest claim I could find at press time was 1.4M IOPS. A far cry from 2 million. More on this later. I get a 0.
- SAS RAID/JBOD cards will hit 1,000,000 IOPS: Another swing and a miss. The highest I could find was 700,000 IOPS. A good increase over last year, but not 1 million. I get a 0.
- pNFS products will finally be released: It has been a long time coming. The final code with the needed changes in the Linux kernel and layout drivers are coming from a number of vendors. Although layout drivers for some file systems are released, many, especially parallel file systems, are not. However, all of these are coming, as we now have crossed the boundary with pNFS in the Linux kernel. This is closer to 1 than 0, so I am going to give myself a 1.
- Silent data corruption in file systems will be addressed in hardware with T10 DIF/PI: The disks are here, and vendors are rolling out products. A number of vendors have released product with more to come very soon. I get a 1 here.
- The dramatic declines in tape unit sales will significantly slow: Every tape drive vendor and media vendor I talk with says that the growth in the archive market has improved sales, and media is selling very well and growing. I get a 1.
- SSD density growth will slow: Both Jeff Layton and I have written about the expected slowdown in SSD density growth. Last year, the biggest 2.5-inch SAS SSD I found was around 500 GB, and this year it is the same. I therefore get a 1, and should have considered this for prediction No. 4 on SSD IOPS. Bad Henry.
- 2.5-inch enterprise drives will surpass 3.5-inch drives in sales in Q4 2011. The disk drive market is changing, and 2.5-inch drives are the future for IOPS performance, watts per GB, watts per IOP for disk and other reasons. I had expected 2.5-inch drives to dominate much sooner, but the economy caused RAID vendors to delay developing new backend drive trays for 2.5-inch drives. From the information I have, 2.5-inch enterprise drives have exceeded the sales of 3.5-inch enterprise drives. I get a 1 here.
In summary, I got 4 wrong and 6 correct -- not great, but exactly the same as last year. If I had only a few more months, I would have gotten a few more correct. Which brings us to ...
Storage Predictions for 2012
Since I got some things wrong last year, I will start with them.
- Declustered RAID will be available most major vendors. If this does not happen, large storage environments will have significant data loss. I have been writing about this for a few years, and it is time the vendors got on board. I believe in 2012 they will be on board.
- SSDs will achieve 2,000,000 IOPS performance. I should have known this was not going to change last year, as both Jeff Layton and I have predicted the slowing of SSD density. I think that we can hit the 2 million number, but as I mentioned, name an application that could take advantage of this feature, as it would have to seriously rework how it processes I/O requests. We will hit this macho number in 2012, but it will be a stretch, and it will require new flash technology and processors to do it.
- Last year, I said SAS RAID/JBOD cards will hit 1,000,000, and we are at 700,000. I believe that we will be about 1.2 million by the end of 2012. These faster cards are needed for SSD RAID devices.
- The appliance market will grow significantly from the home PC to large enterprise to HPC. There will be significant growth in this area, given that low-end CPUs are getting cheaper and faster. At the larger scale, there are not many scalable file systems, and few people to manage the file systems that are out there. The embedded file system market will also grow as pNFS grows. By the way, I have a home NAS and love it. My OS disk and data are now separated. I see this type of technology becoming standard throughout the market.
- I do not see much, if any, improvement in local file systems on Linux or Windows. The scalability problems of data and metadata will continue. This is one of the reasons for the above predication. I predict in 2012 there will be significant improvements in parallel file system performance for metadata scaling data.
- SSD market consolidation will continue as I have said, and there will be fewer SSD vendors in 2012 than there were 1in 2011, which is exactly what happened to the disk drive market.
- Enterprise SSDs density will grow at best 50 percent for 2.5 inch drives. We have gone from 2 TB to 3 TB drives in 2011 for enterprise SATA, and 600 GB to 900 GB for enterprise 2.5-inch SAS drives. These SSD drives did not grow in 2011, so it will be about time for some density increase. I am pretty sure I know why drives did not grow this year, and time will tell if I am correct about them growing in 2012.
- 16 Gbit Fibre Channel will be released long before the end of 2012, but the per-port cost for an HBA and switch ports for, say, a 16-port switch will be more than 4X the cost of 10 GbE port costs. The cost of 10GbE ports is dropping quickly, given that the volume is dramatically increasing. Even BestBuy is now selling 10 GbE connectivity. Other things that will put price pressure on Fibre Channel include 6 Gbit SAS, which will also take some market share for connectivity that does not need to be long distance.
- Cloud storage will continue to have problems, and I predict additional data losses or non-availability (e.g., data will need to be restored from tape) from at least one major provider.
- LTFS (linear tape file system) products will be released into the market by more vendors (I am aware of only two that have a generally available product) and allow tapes to be shipped around the world without requiring backup or HSM applications to read the data. I believe LTFS is starting to catch on, and I expect to see more vendors releasing products or significant market penetration by the end of 2012. Proprietary tape formats will start to become a thing of the past.
I could predict other simple things like POSIX I/O will not change, local file systems will not show much improvement and a variety of obvious predictions to bump my score up. I, of course, would never do that. Many of these predictions will depend on the economy improving to make them come true. That, of course, I cannot predict, but like everyone else, I am hopeful.
Happy Holidays and best wishes for a more prosperous 2012 that will allow me to get more of the predictions correct. I will do my best to ensure my 2013 predictions come out before Dec. 21, 2012, just in case the Mayans are correct.
Henry Newman is CEO and CTO of Instrumental Inc. and has worked in HPC and large storage environments for more than 29 years. The outspoken Mr. Newman initially went to school to become a diplomat, but was firmly told during his first year that he might be better suited for a career that didn't require diplomatic skills. Diplomacy's loss was HPC's gain.