backtop


Print 26 comment(s) - last by namechamps.. on Sep 21 at 8:57 AM


Pliant's SCSI 300 GB and 150 GB enterprise "Lightning" SSDs available in 3.5" and 2.5" form factors offer up to an incredible 180,000 IOPs, at a price that's less than that of the top players on the market.  (Source: Pliant Technology)
Pliant Technology up the bar for performance thanks to custom ASIC

Kingston, Intel, Samsung, and OCZ are all names you may have heard of in the solid state drive (SSD) industry.  Pliant Technologies is one that you likely have not.  Founded in 2006 Pliant Technologies describes itself as a proprietor of "Green IT" solutions.  With a management team composed of former Fujitsu, IBM, Seagate, Maxtor, Quantum, and Conner executives, the company has veteran experience for such a young firm.  And it just dropped a bombshell of a product on the performance SSD market.

The company has just released two "Lightning" high performance enterprise SSDs that threaten to blow away the competition.  The drives uses proprietary ASICs to deliver an incredible input-output performance per second (IOPS) that close to doubles the fastest of its competitors.  The Enterprise Flash Drive (EFD) LS offers 180,000 IOPS in a 3.5" form factor, while the 2.5" form factor EFD LB claims 140,000 IOPS of performance. 

If that's not enough to sate the appetite of even the most die-hard flash drive enthusiast, this will be --  the drives also offer 500MB/sec and 320 MB/sec reads and 420MB/sec read and 220MB/sec write rates for the 3.5" and 2.5", respectively.  This compares favorably to leading competitors, such as the OCZ Vertex Turbo, an overclocked SSD which offers 270 MB/sec reads and 210 MB/sec writes.

The 3.5" is available in 300 GB and 150 GB varieties, while the 2.5" is exclusively available in 150 GB form.

Greg Goelz, vice president of marketing at Pliant says the drives are also extremely hardy.  He brags, "Put it on a log application and write to it as hard as you want for five years -- it will run 24/7 for at least that long."

He claims the drives have no write limits, and will experience no performance degradation even when used continuously for up to five years.  He says his company is working with EMC Corp., Hewlett-Packard Co., Hitachi Data Systems and Sun Microsystems Inc. to offer the drives in integrated solutions for businesses.

Joseph Unsworth, research director for NAND flash semi-conductors at Gartner Inc. thinks that Pliant may indeed shake up the market.  He states, "They're able to claim some pretty solid performance numbers on read and writes and they're also able to claim unlimited program and erase [write/erase] cycles.  That's big. In an enterprise environment, that's one of the major concerns: The wear out of the SSD."

The SSD moves up from the Fibre Channel interface used by many of its competitors, instead using a serial-attached SCSI (SAS) interface, which supports up to 6 Gbit/sec transfer speeds (compared to 4 Gbit/sec for most Fibre drives).  While 8 Gbit/sec Fibre products are on their way, Pliant has a trump card up its sleeve -- 12 Gbit/sec SCSI drives, set to drop in 2012.  Some competitors are jumping on the SAS bandwagon, but they are unable to match Pliant's lofty IOPS.

On an architecture level, the drives are pretty standard, using 12 I/O channels to interleaved single level cell (SLC) NAND flash chips from Samsung Corp.  What is special is its controller, an ASIC.  Most of the industry uses FPGAs controllers.  Also unique is the fact that Pliant has no DRAM cache, achieving its massive performance without the aid of caching.  And the drive also supports a special triple redundancy error correction code algorithm that makes sure data is correctly written, even if two copies of it are corrupted.

Eager to buy one?  Unfortunately for enthusiasts, the drives are enterprise only and will not be hitting consumer retail channels.  The cost?  Pliant remains mum, but it would comment that its drives would fall between the price of Intel X-25-E SSD 64GB SATA drive ($780) and the STECs Zeus SSD 73GB Fibre Channel drive ($6,000) -- so in theory one could conjecture electrifying drives might be had for around $3,000, if you're lucky.  Let's hope this initial business offering allows this new kid on the block to drop its prices a bit and enter the consumer market, as enthusiasts are surely thirsty for that kind of performance, if it can be had at a reasonable price.



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

well
By jay401 on 9/15/2009 9:17:51 AM , Rating: 5
The "most die-hard flash drive enthusiast" would care about the random read/write speeds as an indicator of the drive's true performance, particularly after it has been filled with data. :)




RE: well
By Flunk on 9/15/2009 12:46:55 PM , Rating: 2
That's only true for certain applications E.G. most desktop apps and database applications. If you needed a drive store a single enormous sequential log file or another very linear job sequential write performance is what matters. Remember, this is an enterprise product.


RE: well
By Flunk on 9/15/2009 12:50:36 PM , Rating: 2
Forgot to mention that for massively parallel apps the number of IOPS is what matters and if this thing actually performs as they claim it would be a spectacular device for that sort of use.


RE: well
By Continuation on 9/15/2009 9:33:27 PM , Rating: 5
If all you want is sequential write, then SSD is not what you want.

An array of plain old 15k SAS disk will give you higher sequential performance/cost than any SSD ever could.

Here, a $280 15k SAS gives you 3G/s transfer rate:
http://www.pcmall.com/pcmall/shop/detail.asp?dpno=...

People pay for SSD for its IOPS, not sequential performance, which is very comparable to SAS.


RE: well
By jimhsu on 9/17/09, Rating: -1
RE: well
By namechamps on 9/21/2009 8:57:58 AM , Rating: 2
The drive you linked is 3Gb (little b as in bits) per second.

That is 375MB/sec which is less than these drives.
However that is BURST speed.

Per the ad "Sustained data transfer rates vary with drive speed, capacity and model but are typically 50 to 90 MB/s in sequential."

50-90MB/sec sustained.


RE: well
By Janooo on 9/15/2009 1:31:17 PM , Rating: 2
So check the latest Anand's article about SSD's.

The best Intel: Up to 35K IOPS reads, writes - up to 6.6K IOPS (80GB), up to 8.6K IOPS (160GB).

I guess 180K is for reads. Who knows about writes.


RE: well
By Assimilator87 on 9/15/2009 2:13:41 PM , Rating: 2
The only drives I've seen with comparable IOPS are the Fusion-io ioDrive Duo, which come up as $14k+ on Google. If these guys are really selling their drive for under $6k, they could make a serious dent in super high end SSDs.


RE: well
By wifiwolf on 9/15/2009 2:20:10 PM , Rating: 2
Was wondering the same. Too good to be true either.
Not to mention : no wearing and no performance hit on usage...
We can only say something after a review.


I'll believe it....
By Ristogod on 9/15/2009 9:17:05 AM , Rating: 2
I'll believe it when I see it. Some of it seems like some lofty claims.




RE: I'll believe it....
By drq on 9/15/2009 9:33:18 AM , Rating: 2
Yep, like Bitboys Oy...


RE: I'll believe it....
By XZerg on 9/15/2009 10:00:45 AM , Rating: 2
darn bitboys... they really hyped a paperjob real good...


RE: I'll believe it....
By CommodoreVic20 on 9/15/2009 12:52:50 PM , Rating: 2
Wow, are you referring to the bit boys of the mid nineties with the vaporware video card designs?


RE: I'll believe it....
By FaaR on 9/15/2009 5:44:40 PM , Rating: 3
Perhaps you missed the bit of the piece where it mentioned staff coming from a half-dozen world-class storage product makers (past and present)...

Doesn't exactly compare to Bitboys, I'd say.


RE: I'll believe it....
By afkrotch on 9/17/2009 6:02:36 AM , Rating: 2
Ya, staff coming from a half-dozen world-class storage product makers. They are more like 3D Realms.


RE: I'll believe it....
By clovell on 9/15/2009 11:01:33 AM , Rating: 5
I'd like to see Anand put this one on the bench.


What makes the ASIC special
By mmcdonalataocdotgov on 9/15/2009 10:45:27 AM , Rating: 2
is what I want to know. All ASICs start out as FPGAs. Do other SSD manufacturers really leave their controllers in a true field programmable state, or are they all set to application specific in the fab? Every piece of hardware like this has an ASIC (like a CPU or a GPU) but not many are left true FPGA, except your CMOS, or other flash programmable chips. And not many are FP on the fly (n x per second).

This is the real meat of the article and only glossed over. Let me check out MIT Technology Review where this probably came from, or IEEE...




RE: What makes the ASIC special
By splint on 9/15/2009 12:29:23 PM , Rating: 1
No, an ASIC, by definition, (Application-Specific) is a custom chip. It may have been an FPGA at one point in a lab somewhere during initial product development.


RE: What makes the ASIC special
By emboss on 9/15/2009 2:10:57 PM , Rating: 5
Mtron used to use actual FPGAs (a Xilinx Spartan 3 IIRC) in their drives for a short time. Once they ramped up quantities they transitioned to an ASIC. JMicron, Indilinx, Intel controllers are all ASICs. So there's nothing really special happening here.

Apart from that, most of their claims are just marketing fluff. Their "unlimited" erase cycles aren't really unlimited - it's just that since they're using SLC, they've got 100K erase cycles which works out to be some stupidly long length of time. They're probably also over-provisioning the flash to a greater degree than consumer dives - a quick calculation suggests 27% free space compared to 6.9% in for example the X25-M. That massively helps with dropping the write amplification, at the obvious cost of needing more flash.

Also, they "cheat" a bit on the benchmarks. The "single" Lightning drive is actually just two drives stuck inside the same case - they're not even RAIDed, it physically requires two SATA connections. So there's an immediate halving of their numbers if you want to compare it with a single X25-M for example. Nor do they actually support the just-released 6 Gb/s SATA/SAS - the 6 Gb/s is the total bandwidth from two regular 3 Gb/s SAS ports.

Finally, the "all done with no cache" thing is a bit of a semantic cheat too. They use DRAM in the same way that the Indilinx, Intel, and Samsung controllers do - to hold mapping information so that it doesn't have to hit the flash all the time to figure out where to write (this is what was the killer for JMicron drives). I suspect in reality they do exactly the same as the Indilinx/Intel/Samsung drives, which is defer any partial logical page writes until there is a write to a different logical page.

So, in summary, it looks like Just Another SSD Controller. They throw in a heap of very over-provisioned SLC flash, stick two drives in a single casing, and as expected get some big numbers. I'm more than willing to be surprised, but given what can be dug up so far it really doesn't look to earth-shattering.


RE: What makes the ASIC special
By bhigh on 9/15/2009 6:51:46 PM , Rating: 2
quote:
The "single" Lightning drive is actually just two drives stuck inside the same case - they're not even RAIDed, it physically requires two SATA connections.

It's probably a dual-ported SAS drive. It allows for multipathing for more bandwidth or to survive a single link failing. It's pretty common in high-end SAS drives.


RE: What makes the ASIC special
By emboss on 9/16/2009 10:32:22 AM , Rating: 2
Ah, good point. Too used to spinning rust, where dual-port isn't usually done for performance reasons :)


Misleading Marketing or Poor Reporting?
By GourdFreeMan on 9/15/2009 7:15:01 PM , Rating: 4
This article sounds like deliberately misleading marketing fluff. Using an ASIC rather than an FPGA means a faster controller, but it doesn't magically change the properties of flash memory! The claim of "unlimited program and erase [write/erase] cycles" can be nothing more than pure hyperbole. Using triple redundancy error correction code and reserving extra space on the drive will significantly increase the lifespan of the drive, but eventually enough cells will fail and the drive will die. Likewise, AFAIK triple redundancy ECC will not automatically correct the error in the event two of the three drives fail; it will only be likely to detect it (the two failed drives would have a small probability of agreeing and thus returning the wrong result). The claim of "no performance degradation even when used continuously for up to five years" doesn't explicitly mention small random writes, and probably doesn't include them. Unless your filesystem is designed around flash's limitations (i.e. internal 512k cluster size), you are eventually going to run into some performance problems with internal fragmentation if you do nothing but small writes on the drives IO limit 24/7.

There have been drives that made similar claims of huge IO speeds only to be found to suffer from the same problems as other SSDs once the drive had been filled once. Fusion IO, for example lost 90% of its random IO performance after being filled. See http://www.xtremesystems.org/forums/showthread.php...




By GourdFreeMan on 9/15/2009 7:21:55 PM , Rating: 2
Sorry, typo. "on" should read "at" in the last sentence of my first paragraph.


By mindless1 on 9/15/2009 10:33:01 PM , Rating: 2
The key to the statement seems to be "for 5 years". That means if you use the drive and have a problem, they will warranty replace it for 5 years. They're not signing a contract to give you their first born if it fails.

Pretty easy for anyone to say unlimited writes for 5 years isn't it? I could say the same to market any flash drive on earth right now, and just charge enough to cover the eventual RMAs but frankly I doubt many will fail within 5 years if it's designed well. After that it's anybody's guess till we have some real data.


By murphyslabrat on 9/15/2009 9:32:22 AM , Rating: 4
So, the only bit that wasn't almost directly lifted from the source article, was that last bit about price. It was a vague claim, but even then, was not cited. For all we know, he could have made it up.




Boom
By xKeGSx on 9/15/2009 9:12:39 AM , Rating: 2
Now we just need a real world review.. and BAM they get bought up and the owner makes a sweet profit.




"This is from the DailyTech.com. It's a science website." -- Rush Limbaugh




Latest Headlines
2/10/2012 Daily Hardware Reviews
February 10, 2012, 5:50 PM
2/9/2012 Daily Hardware Reviews
February 9, 2012, 11:54 AM
2/8/2012 Daily Hardware Reviews
February 8, 2012, 1:11 PM
2/7/2012 Daily Hardware Reviews
February 7, 2012, 12:23 PM










botimage
Copyright 2012 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki