The 2016 Flash Summit concluded last week in Santa Clara and although I was not able to attend this year, all the proceedings are available online. There is a wealth of information. If there is anything you want to know about flash technology, flash futures, and flash use cases you can probably find it in these proceedings. A common message in this summit is that flash is ready for prime time and you are wasting your money buying disks. Marc Staimer’s presentation on Why Flash Should Be Used For Every Workload showed that from every perspective; performance, reliability, availability, infrastructure costs, capacity, density, power consumption, productivity, and end user experience, flash had much better TCO and ROI than hard disks.
While the street price of the most flash media may not be lower than SATA disks, our 7TB FMDs with embedded compression is already lower in cost than 15K RPM hard drives on a per bit bases. Storage Switzerland’s George Crump interviewed our Director of Flash Storage, Matt Pujol, on the technology behind our 7TB FMD. Matt hinted at a 14TB flash FMD in the near future. Seagate unveiled a 60 TB SSD at this year’s Summit, so we can expect to see the capacities of flash modules accelerate way beyond what disk technology will be able to deliver. Disk vendors like Seagate understand this and are changing their product mix.
Dan King, Director of Information Service and Support for Wellmark, presented How to Fund Flash Storage where he showed that even when the cost of flash was higher when he went to an all flash VSP 1000, the savings in OPEX more than offset the CAPEX costs. While most people focus on CAPEX, his philosophy was to go against the grain and use CAPEX to reduce OPEX which is usually 80% of the total cost.
Instead of performance, the primary reason for buying flash over disks now is higher capacity and lower cost.
With the movement to all flash data centers will there be enough fab plants to generate the DIMMs required to meet this demand? (fab is a semiconductor fabrication plant which costs several billions of USD to build). After a period of time when no new fabs were being built it looks like the semiconductor vendors are starting to crank up the fabs again to meet the demands of the memory market. At the Flash Summit there was a session devoted to the China Flash market. In this session Cui Hao from WatchStor a leading technology media company in China, presented Flash Anytime Everywhere in which he said that Institutions in China predict that by 2016-2017, China will build 10 fabs, with a total investment of one hundred billion! Press reports show Intel and Samsung are spending billions for new fabs and SK Hynix plans to invest $38.9 B to build 3 new fabs by 2024. In order to recover their billions of dollars in fab investments these vendors must be planning on selling a lot of DIMMs. So it looks like we will have enough supply to meet demand.
We used to say that the future of the data center is Flash. That future is today