And there was me thinking solid-state drives were just a flash in the pan
EMC is heading towards a FAST future, with transparent and automated data movement between NAND flash and spinning disk in its drive arrays. In a Tech Talk yesterday - reported by Stifel Nicolaus analyst Aaron Rakers - Barry Burke, Symmetrix chief strategy officer, said that FAST (Fully-Automated Storage Tiering) will be …
Not that my array has(or will get) FAST(v2) like technology, but I guess I'm ahead of the curve in that using 100% sata in my enterprise storage system to provide both good performance and high capacity when all disks are striped together. It doesn't scale low all that well but at a certain point(I'd say for 3PAR architecture at least 100 disks) it works well. Our system has 200 disks, going to 300 next year. Previous array had more than 500 disks, half of them being 10k RPM, new system is faster overall and has more net usable storage because of the added performance.
"We might expect other enterprise array suppliers with SSD storage tiers announced or coming, such as 3PAR, Dell, Fujitsu, Hitachi Data Systems, HP, IBM, NetApp, Oracle/Sun, Pillar and Xiotech, to all follow Compellent and EMC's lead and introduce their own automated data movement across storage tiers between now and the end of next year."
Sun and NetApp already have a variation of this functionality with Amber Road and PAM.
Pillar has led the idea of tiered storage management with its QoS and Application-Aware profiles to manage storage resources in line with application IO need. It's been in every Axiom shipped over the last four years. In fact, Pillar takes it a step further through automated tiering across 3 essential resources - CPU, Cache, and Capacity - to achieve IO service levels.
The only hard data out there shows there is no ROI for SSD as Tier-zero, and there won't be any ROI until SSD cost/GB drops below about $5/Gbyte. And no, there's no ROI not even when energy costs are included.
Now, considering that EMC is selling STEC today for about $200/Gbyte, where the is the value proposition (except in gross margins for EMC and STEC)?
If you mean a 4MB block then yes. If you are running a 4K block size at the application layer, major coalescing occurs when it is written to disk, so unless your application has VERY localized data sets, this model isn't as efficient as you'd expect, however, that said, they do it at a more granular level then most.
"Sun and NetApp already have a variation of this functionality with Amber Road and PAM"
no they don't. They just use "SSD" as level 2 cache. With Amber Road their "write acceleration" cache (or whatever they call it) simply "accelerates" the writes to that of SATA drives. Impressive.
Biting the hand that feeds IT © 1998–2022