The Year in Storage: Reflection on 2012 and Predictions for 2013

It is that time of year again when we look back to see where we’ve been and look forward to see where the storage industry is going. Every year, I try to predict technology winners, losers and those stuck in the middle.

Both 2010 and 2011 saw significant stagnation in the storage industry, but in my opinion, 2012 was a breakout year. I expect that, unless something bad happens to the economy, 2013 will continue with that trend.

“Consolidation” is going to be next year’s buzz word. Large vendors are flush with cash and the folks in the investment community are applying significant pressure for growth and profits.

2012 Review

So how well did I do with my 2012 predictions?

1. I said that declustered RAID would be available from most major vendors, a prediction I based on 4 TB drives becoming mainstream. Though some vendors do have 4 TB drives, not all of them do. Some vendors, such as NetApp and IBM, do offer declustered RAID, but it surely is not ubiquitous in the industry.
2. I also predicted that SSDs would achieve 2,000,000 IOPS performance. I got this one right, but not until October. I am not saying in any way, shape or form that doing 2 million IOPS is useful, but the SSD IOPS arms race will continue no matter what I think.
3. In addition, I said that SAS RAID/JBOD cards would hit 1.2 million IOPs by the end of 2012 because these faster cards are needed for SSD RAID devices. Well, I was off by a little. Back in September, an LSI press release announced a single 8-port 12Gb/s SAS ROC IC achieving over 1.2 million IOPS (I/O operations per second) and said that “Volume shipments of 12Gb/s SAS-enabled servers is estimated to occur by early 2013, followed by the availability of external storage systems by mid-to-late 2013.” So I missed this one by a few months. Darn.
4. I predicted that the appliance market would grow significantly for everything from home PCs to large enterprises to HPCs. I believed there would be significant growth in this area given that low-end CPUs were getting cheaper and faster. I do not think that anyone can argue about the growth in this market from big data, NAS, archive appliances, etc. The market for simplified storage is growing and growing at a high rate.
5. Also, I forecast that in 2012 there would be significant improvements in parallel file system performance for metadata scaling data. This is happening for sure with new appliances from IBM with GPFS and appliances from DDN, NetApp, Xyratex and others for Lustre. This is a large growth market because everyone wants a large common namespace and current NAS products are not meeting the challenge.
6. I predicted that SSD marketing consolidation would continue. I do not think that anyone can argue this is not happening. Texas Memory becoming part of IBM is just one example.
7. Next, I said that enterprise SSDs density would grow at best by 50 percent for 2.5 inch drives. Last year we had 600 GB enterprise SSDs at the top end. Today, we have 800 GB enterprise SSDs at the top end for 33 percent growth.
8. I also stated that 16 Gbit Fibre Channel would be released long before the end of 2012. But I said the per-port cost for an HBA and switch ports for, say, a 16-port switch would be more than four times the cost of 10 GbE ports. The cost of 10GbE ports would drop quickly, given that the volume is dramatically increasing. All of this happened exactly how I thought it would. Both QLogic and Emulex released 16 Gb Fibre Channel. 10 GbE ports are finally starting to drop in price, and the cost of a NIC is just tens of dollars.
9. In addition, I predicted that cloud storage would continue to have problems, and I forecast additional data losses or non-availability (e.g., data would need to be restored from tape) from at least one major provider. How many outages did we have this year? Azure, Amazon, Google and others all reported downtime.
10. Finally, I said that LTFS (linear tape file system) products would be released into the market by more vendors and that this would allow tapes to be shipped around the world without requiring backup or HSM applications to read the data. Today, we have products in the market from Crossroads, IBM, HP, Oracle and others. LTFS has moved to SNIA for standardization, and proprietary tape formats are being seriously threatened.

Results: My final tally for 2012 was 8 correct and 2 wrong, with 1 of those 2 being delayed till early 2013.

2013 Predictions

As I said last year, I wanted to get my 2013 predictions out before December 21, 2012, in case the Mayans are correct.

Also, one other thing that I said last year holds true: All of what I have written was true the day I wrote it, but nothing in our industry stays the same for long.

1. My first prediction is that sometime in 2013 we will find that PCIe-4 is going to be delayed–probably till 2016. This means that technologies such as 100 GbE will have to run in 16-lane PCIe slots. Without 32 lanes you cannot have a dual port card. The latest thing I can find from the PCI SIG is this slide:

   

2. There will be continued market consolidation among SSD vendors. The market has too many vendors with limited technology differentiation and with high marketing and sales costs. This model does not work nor can it be sustained given the development costs.
3. 10 GbE per port switch prices will finally be priced reasonably. The number of 10 GbE chipsets is growing fast. Soon, the market will be flooded with new chips and this will drive the prices of switch ports down.
4. In the enterprise, we will finally see a significant movement to 2.5 inch disks rather than 3.5 inch drives. There are lots of advantages of 2.5 inch drives over 3.5 inch drives in terms of watts per IOP and watts per MB/sec. This is not to say that I am predicting the death of 3.5 inch drives–just movement to reduce the market share for 3.5 inch drives.
5. We will see file system appliance vendors becoming application appliance vendors with specific highly tuned appliances embedded into the architecture.
6. We will see a significant reduction in the number of LTO tape library vendors and major cutbacks from some vendors. The reason is that LTO-6 only increased density to 2.5 TB, and enterprise vendors are already way beyond that in terms of density and performance. For large tape library configurations, LTO is more expensive than enterprise tape given the cost of robot slots and number of tape drives needed to meet performance requirements. Though LTFS will help, it is not going to stop this trend.
7. We will see the T10 PI (host to disk data protection) standard finally become mainstream. There have been a few technologies that have caused this to lag, but I think we will finally get everything in place. If you are a regular reader, you know that I have made this prediction before and been wrong. We really need this technology.
8. More consolidation in the HPC community will impact us all. Last year Intel purchased the QLogic Infiniband business, Cray’s network interconnect group, and WhamCloud, the Lustre file system company. Intel has clearly started a trend, and other vendors will have to respond. I predict this will begin in 2013.
9. A vendor will demonstrate new non-volatile memory technology that could be used for things like database index tables. I am not saying who will demonstrate it or what the technology will be, but I believe we will see something next year.
10. Finally, this last item is more of a hope than a prediction, but I will make it anyway. The industry really needs more than POSIX (open/fopen, read/fread,write/fwrite) and more than simple REST put/get interfaces for data in the future. Neither has the richness to address the myriad of polices that are needed in our future world. I predict that there will finally be some honest discussion about this amongst the customers that need it and the vendors that could create it. Maybe this should be my request to Santa. I have tried to encourage this discussion for years and I have gotten no traction.

Have a happy, safe holiday season and a prosperous 2013!