Pliant pops out SASsy SSD
Competition for STEC and Intel
Startup Pliant is launching its Lightning solid state drive (SSD) today, with SAS interfaces only and STEC-beating performance claims.
A fair amount of Pliant's marketing literature speaks of SSDs being much faster than hard disk drives and compares an all-HDD configuration with a hybrid SSD/HDD configuration to satisfy a TPC-C benchmark-style 640K transactions/minute, 320,000 IOs per second (IOPS), 18TB database application task. In the HDD case there are three racks of 1,000 short-stroked 15,000rpm drives, and the overall cost is more than $450,000.
The hybrid Pliant SSD/HDD configuration has six shelves with just 21 10K hard drives and costs $225,000, with concommittant lower $/IOPS, $/GB numbers and much higher IOPS/watt figures.
Currently STEC has the lion's share of enterprise storage array design wins for SSDs that replace fast hard drives. It is quite widely thought that fast SAS drives will start replacing Fibre Channel drives, especially with the 6Gbit/s SAS interface coming into play. That's why Pliant is concentrating on SAS interfaces only. These are dual-port and full duplex SAS interfaces, by the way.
Greg Goeltz, Pliant's marketing VP, says it's 3Gbit/s SAS today. We should expect 6Gbit/s SAS deployments in mid-2010.
It says its Lightning products come in 2.5-inch and 3.5-inch form factors. The LB 2.5-inch version has 150GB capacity and runs at more than 140,00 sustained IOPS with sustained read and write data transfers of 420MB/sec and 220 MB/sec respectively.
The LS 3.5-inch products come in either 150G or 300GB capacities, and is rated at more than 180,000 sustained IOPS, and sustained read and write bandwidth of 500MB/sec and 320MB/sec respectively.
STEC's ZeusIOPS drive also comes in both 2.5-inch and 3.5-inch form factors and has SAS interface versions as well as Fibre Channel ones. Capacities are higher with up to 400GB (2.5-inch) or 800GB (3.5-inch). The bandwidth numbers are 350MB/sec for reads and 300MB/sec for writes; Pliant offers faster reads but slower writes.
STEC says the product can do up to 80,000 read IOPS and 40,000 write IOPS, ceding a lead to Pliant in the IOPS area. STEC has cheaper multi-level cell (MLC) ZeusIOPS product. Pliant is ignoring MLC and going all-out for performance.
A Pliant pitch is that its products perform better in varying mixes of read and write I/Os, and across varying block sizes; performance, it says, is more consistent than other products, ie essentially flat and not tailing off.
It says it offers unlimited writes for the lifetime of the device and it does not use a DRAM write cache for fast write performance. Autonomous background processes in the controller patrol memory, to check data integrity, and reclaim deleted space, with no impact on read and write access speed when they are active. It says reads and writes can take place concurrently.
Pliant claims its data reliability is far higher than the competition with data reliability seeing arrors in less than 1 sector in 10 to the power 17 bits. It cites competing product figures of 10 to the power 15 or 16 bits.
The Pliant product uses a Pliant-designed ASIC (Application-specific integrated circuit) in its controller with single level cell (SLC) NAND chips sourced from flash foundries. The ASIC approach was used, Goeltz said, because FPGAs (field-programmable gate arrays) wouldn't be fast enough.
Pliant already has qualification units with OEMS and beta product running at customer sites. There is a whisper that EMC is an investor in Pliant. It's not announcing any design wins and won't say whether it has any or not.
It doesn't matter how good its claimed numbers are unless OEM suppliers of enterprise storage arrays view them as real and qualify their drives for use. Most of all Pliant has to gain SAS SSD design wins with OEMs that are currently using STEC SSDs. It may well also send its sales reps calling on Dell/EqualLogic and Pillar Data who have adopted Intel SSDs. There are fewer than ten top tier OEMs using enterprise flash drives; it's a very limited market and the sales targets and competition are both crystal clear.
Consequently Pliant has a polished competitive story ready to run. Goelz says: "STEC's next-generation product is about half the speed of the Lightning product. They have a very good starting point. (But) after one hour or so performance drops 40 per cent or so and maintains that level. It's about half of their published speed. (For reliability) they rely heavily on supercapacitors and a write cache for performance.
"Intel's X25-E is really the X25-M-class product with SLC flash and heavily relies on write cacheing (but) there's no protection for the cache. If you lose power you lose cache data. In mixed read/write situations it loses more than 90 per cent of its advertised performance.
"SandForce, like us, believe performance should be flat. It needs to get a customer first, though. Seagate is already a formidable HDD player but this is a different technology. Not-invented-here syndrome can creep in to companies in this field. It's still not clear that Seagate will make its end-09 committed date."
So Pliant is already saying its technology is faster than the competition's, performs with less drop-off in mixed environments, and is more reliable. It has product with potential OEMs and with customers who can help persuade OEMs that, yes, this is good flash, and they should adopt it.
This is a very high stakes game. Not winning a significant proportion of fewer than 10 tier one OEMs could break Pliant. Flash buyers and influencers in 3PAR, Compellent, Dell, EMC, Fujitsu, HDS, HP, IBM, NetApp, Sun, Xiotech and some other companies are going to get the most carefully prepared, well though-out, and sustained sales pitches from Pliant they could imagine.
It's got to be a text book case of competitive selling against an incumbent. This is going to be intense. ®