the fastest SSDs - click to read article
the fastest SSDs
SSD SoCs controllers
SSD controller chips ..
2010 - year 1 of the SSD Market Bubble - article on storagesearch.com
fizzing SSD bubble? ..

storagesearch.com

by Zsolt Kerekes, editor, April 2008

storage search
10 years - "leading the way to the new storage frontier"

the Flash SSD Performance Roadmap

Predicting and Reporting Key Trends in Flash SSD Performance - upto 2012

I've often asked where I think the flash SSD market is heading in terms of performance.

Although my earlier article RAM SSDs versus Flash SSDs - which is Best? looked at the state of the art (in 2007) it didn't include a performance roadmap.

This article does. And it has a bunch of simple to remember predictions or "laws". For convenience you can think of them as "Zsolt's Laws". Or maybe "Z's Laws" is easier to spell. The first law is the only one you need to remember.

Overall this article tells you the market model assumptions I would work to if I were setting up a new SSD company to design the fastest flash SSDs.

Obviously I'm not doing that - because otherwise the many people I talk to in SSD companies around the world wouldn't be talking to me.

I've got a good crystal ball and the lucky thing about my many past predictions related to the storage market - is that if they weren't going to be true at first - then suggesting to companies that they ought to spend more time looking at things in a particular kind of way becomes a self fulfilling prophecy. My safety net is that as many designers in the industry also read these articles - if I say something that's too outrageously wrong - they correct me - before I look too ridiculous. A quick delete, cut and paste, and it's even more accurate than it was before.

The rate of speedup predicted is much faster than would be predicted by Moore's Laws - and that's the significant point. Some explanations why - are included at the foot of the article.


  • Prediction 1 - Flash SSD throughput and IOPs (in traditional HDD form factors) will more than double every year in the period from 2007 to 2012.

    This predicts (in effect) that in 2011 a single 3.5" form factor flash SSD will be able to deliver similar throughput to some of the fastest RAM SSDs available in 2007, with over 2,000MB/s sustainable reads and writes.
The single most useful thing to take away from this article is the assertion above. But if you want to have some more to think about read on.

click for more info

  • Prediction 2 - Rackmount flash SSD throughput and IOPS performance will be a multiple of the performance for a single disk. These factors have already been shown to be scalable in SSD RAID arrays.

    This needs little explanation as some of the results are intuitive and we've already published plenty of articles on this subject. However, some of the architectural features which are now used in SSD RAID systems - such as MFT technology - can also be designed into individual SSD disk modules.

  • Prediction 3 - The asymmetry of sustained read to write IOPs will improve from 10 to 1 (the fastest devices available in 2007) - but will never achieve parity (1 to 1).

    As this change occurs in the market flash SSD arrays will become viable choices in many enterprise server speedup applications which hitherto had been the exclusive domain of RAM SSDs.

    In (typical) database applications with Read:Write ratios of 4:1, an ideal flash SSD with 10:1 R/W IOPS is approximately 3x (2.8x to be exact) slower in overall applications performance than an ideal RAM SSD with similar MB/s throughput.

    ...Later:- I was wrong about this one.

    Violin Memory designed a flash SSD with balanced random R/W IOPS in Q4 2008 using a combination of factors.
    .
    • over-provisioning and fast garbage collection ensured a constant flow of pre-erased flash.
      .
    • their non blocking RAID RAM cache architecture enables a read to be done from a virtual flash block while an erase-write is still in operation on the same flash block.


    And in April 2009 - SandForce published - Fact or Fiction - all flash SSDs have unbalanced R/W performance . This is one of a set of articles which discuss design factors inside their small form factor SSD controller.

  • Prediction 4 - Latency in flash SSDs will not scale in the same way as throughput, and will always be significantly worse than that in ideal RAM SSDs.

    The ratio of read access times for RAM SSDs compared to flash SSDs may improve for a few years (as the gap gets smaller) but then it will hit a brick wall - and may in fact get worse again.

    The reason is - that flash SSDs have not yet been optimized for latency - so there is some scope to reduce the latency gap with RAM systems (which have already been highly optimized).

    But in future product generations as flash SSDs increase in density - a read or write cycle becomes an increasingly complicated on-chip process - which includes calibration, error correction and address translation all being done by controllers between the memory array and the host interface controller or card data bus.

    This series of steps (to do a simple read) will diverge from what happens in a typical RAM to the point where flash and RAM look like completely different species. That's unlike earlier generations of flash in which the read cycle looked the same as a static RAM - but simply took longer.

Here are some other preliminary notes of explanation.

Why Z Predicts a Faster Roadmap Acceleration than Moore's Laws

The main reason that the flash SSD market will deliver faster products much sooner than predicted by Moore's Law is that 2 additive factors are at work in this phase of the market:- architecture and semiconductor process technology.

In the past there was little point in manufacturers integrating very fast architectural features into flash SSDs - because they added to the cost - and there wasn't a big enough established market to buy them.

The architectural technologies that can speed up performance and IOPS were originally independently developed by various oems to suit particular products or markets. Until the SSD market reached a critical mass where enough users signalled they would buy faster products if they were available - there was no point in developing them.

the key architectural features which will increase throughput and shrink the asymmetry gap in read / write IOPS are:-
  • parallelization of the internal media arrays
  • improved media management technology.

    MFT from EasyCo is a software solution which has already been mentioned above. But the same algorithms could be run in the SSD hardware.

    Another licensable solution - which is already being developed at the chip level is IOP Buster architecture from Link_A_Media Devices
  • faster media controllers
  • faster host interface controllers (and faster interfaces driven by the needs of the SSD market rather than adapted from the HDD market)
  • hybridizing on board memory technologies - for example using faster RAM-like non volatile memory in some parts of the device and slower flash-like memory in the bulk storage arrays
Scaling any one of the factors above requires significant investment in IP. There is also significant risk that the overall balance of the product specs which results doesn't match the market's expectations for price and performance at the time.

A lot of trial and error will be involved as oems throw products at the market which tweak the technologies they understand best - and see which products stick.

Underlying all the architectural improvements - there will also be evolutionary and revolutionary changes in semiconductor processes occurring at the same time. Some of these will enhance currently known architectures, while others may make some architectural features obsolete.

By around 2013 - the flash SSD tornado should have reached a point where the architecture of an ideal SSD is well established - and the ongoing developments will be drive more by process changes than anything else.

We're going to publish hundreds of articles about this subject as the roadmap takes shape. So don't expect to see all the answers in a simple note like this. But I'll add more notes and links to this article as time goes on.


Here are some related articles.
  • the Fastest SSDs - updated regulalry - this gives the state of the art in popular form factors.
  • Understanding Flash SSD Performance (pdf) - by Douglas Dumitru, CTO EasyCo LLC is a reference for systems engineers who want to understand the dynamics of flash versus hard disk performance running real applications.
  • Flash in the Enterprise (pdf) by Jamon Bowen, Texas Memory Systems describes some the properties of Flash memory and then explains how their RamSan-500 product leverages its strengths and compensates for its weaknesses to offer the fastest enterprise ready rackmount Flash SSD system.
  • Design Tradeoffs for (SLC flash) SSD Performance (pdf) by Nitin Agrawal, Vijayan Prabhakaran, Ted Wobber, John D. Davis, Mark Manasse and Rina Panigrahy looks in detail at the internal architecture of a typical commercially available SLC flash SSD and the various points where choice of software algorithms can impact performance.
  • Why I Tire of - "Tier Zero Storage" - You don't need to waste any of your precious brain cells by investing "tier 0 storage" with an importance this travesty of storage jargon really doesn't deserve.
  • SSD Market History - lists key market, business and technology milestones in the 30 year history of the SSD market.
.
Spellerbyte's ScryWare utility
downloaded data from his crystal ball
directly into his spreadsheet.

Fusion-io fast SSDs - click for more info
world's fastest production PCIe SSD
from Fusion-io

2.5"   flash SSDs  from Memoright
2.5" 128GB flash SSDs
low cost per IOPS - Enterprise Storage Solution
from Memoright
.
the Problem with Write IOPS
Editor:- December 16, 2009 - StorageSearch.com today published a new article - the Problem with Write IOPS - in flash SSDs.

Flash SSD "random write IOPS" are now similar to "read IOPS" in many of the fastest SSDs. So why are they such a poor predictor of application performance?

And why are users still buying RAM SSDs which cost 9x more than SLC? - even when the IOPS specs look similar.
the problem with flash SSD  write IOPS This new article tells you why the specs got faster - but the applications didn't. And why competing SSDs with apparently identical benchmark results can perform completely differently. ...read the article
.
The New Skinny on flash SSDs
Editor:- July 28, 2009 - StorageSearch.com today published an article - RAM Cache Ratios in flash SSDs - which proposes new terms to describe and differentiate products in the flash SSD market.
read the article RAM Cache Ratios in flash SSDs It is hoped that the new classification jargon will be useful to users who have to evaluate lots of products, and will be useful to vendors as a shorthand when communicating about different segments within their flash SSD product lines. ...read the article
.
the Fastest Solid State Disks
Speed isn't everything, and it comes at a price.

But if you do need the speediest SSD then wading through the web sites of over 100 current SSD oems to find a suitable candidate slows you down.

And the SSD search problem will get even worse.
the Fastest Solid State Disks

I've done the research for you to save you time. And this page is updated daily from storage news and direct inputs from oems. ...read the article,
.
Unity Semiconductor Unveils Flash's Successor
Editor:- May 19, 2009 - Unity Semiconductor exited stealth mode and stated its aim to have the lowest manufacturing cost per bit in the non volatile memory industry with a new breakthrough technology called CMOx.

The company said it will ship 64Gb devices in volume in 2011. Unity Semiconductor says it will develop and produce NAND flash successor technologies and products that, in time, will extend into high performance embedded and enterprise applications.

"It's a Technology for Terabits that will challenge high volume rotating magnetic media" said Unity Semiconductor Chairman, President & CEO Darrell Rinerson a former executive at Micron Technology and at AMD.
Flash Memory The company, also announced today it has closed a Series C funding round for $22 million. This brings to nearly $75M the total funding to date in Unity Semiconductor.
.
SandForce Unveils New flash SSD Controller
Editor:- April 13, 2009 - SandForce today emerged from stealth mode and unveiled its SF-1000 family of SSD Processors - aimed at oems building SATA flash SSDs.

Its 2.5" SSD reference design kit is the fastest 2.5" SATA flash SSD on the market - with 250MB/s symmetric R/W throughput and 30,000 R/W IOPS.

Leading OEMs are expected to release both SLC and MLC flash-based SSDs using SandForce single-chip SSD Processors later this year

"With a deep understanding of both system- and silicon-level issues, we've integrated the right balance of reliability, performance, power, cost, and time-to-market in our SSD Processors while supporting multiple flash vendors' technology," said Alex Naqvi, President and CEO of SandForce. "Our products combine key processing elements with hardware automation to efficiently address the traditional shortcomings of flash memory. This allows OEMs to provide enterprise-class SSDs to the mass-market using both SLC and lower-cost MLC flash devices while delivering peak read and write performance throughout the drive's lifecycle."

Editor's comments:- I asked SandForce's President & CEO, Alex Naqvi, for more details about the various package of technologies which are bundled in the company's "DuraClass Technology" - which achieves impressively high IOPS without relying on over-provisioning or large external RAM caches. In particular I wondered what part, if any its choice of processor SoC (from Tensilica) had to play.

Alex Naqvi explained - DuraClass performance doesn't come from the choice of processor - but in the way that they have integrated various design techniques with very fast hardware (proprietary chips) which the company has designed to accelerate the core bottleneck functions of a flash SSD controller.
SSD SoCs controllers In concert with other techniques, such as the ability to reorder data before it is written to flash (thereby attenuating write endurance by 2 orders of magnitude), RAID like internal protection and very fast garbage collection SandForce's DuraClass Technology results in small form factor enterprise class flash SSDs which have no daily write limits for MLC flash and symmetric R/W IOPS.
.
HP oems Fusion-io's SSD Accelerator Technology
Editor:- March 3, 2009 - Fusion-io announced an oem deal with HP whose new PCIe based StorageWorks IO Accelerator for for HP BladeSystem c-Class servers is based on Fusion's ioMemory SSD technology.

A low level formatting tool for the HP SSD enables users to choose what level of over-provisioning is used - as a performance tweaking option.
.
Toshiba Announces Groundbreaking 2.5" SAS SSDs
IRVINE, Calif. - January 8, 2009 - Toshiba announced it will start volume production of dual port SAS interface SLC flash SSDs in Q2 2009.

The 2.5" SSDs will have 100GB capacity, and 25,000 read IOPS, and 20,000 write IOPS. ...Toshiba profile

Editor's comments:- One of the enabling factors for the high write IOPS is the use of a non-volatile cache - which was predicted in StorageSearch.com's article - the Flash SSD Performance Roadmap (this article you're reading now - left hand side of your screen).

This brings the number of oems who have announced SAS SSDs to 6. See SSD Buyers Guide table for the full list. We'll publish a dedicated SAS SSD guide later this month.
.
Experiment Confirms flash SSD Performance Uncertainty Principle
Editor:- January 4, 2009 - Benchmark Reviews has recently published an article - "SSD Benchmark Performance Testing."

This provides experimental confirmation of something I warned about in an earlier article - Can you trust flash SSD specs & benchmarks? - namely that hard disk based performance software provides unreliable results for SSDs.

The author, Executive Editor, Olin Coles (who has published many disk benchmarks) concludes "I warn readers to regard SSD reviews with a high degree of caution". He says he's going to stop using certain test suites, and the discovery that he can't trust all the results which he has worked hard to collect and publish made him feel sick. Coles is a very experienced benchmark tester - and his article makes interesting reading. ...read the article
.
Violin Advances flash SSD Architecture for Enterprise Apps
Iselin, New Jersey - November 10, 2008 - Violin Memory, Inc. today reached out to markets beyond those accessible to its RAM based storage appliance line by announcing the availability of a fast 2U 4TB SLC flash SSD.

Designed for enterprise server acceleration - Violin's flash technology enables over 100K sustained random Write IOPS (4K block) and write throughput is upto 400MB/s. Latency is 70-300 microseconds and the internal architecture eliminates the multi-millisecond variability seen in most current SSD arrays. ...Violin Memory profile

Editor's comments:-
although I had been expecting Violin to announce a flash SSD product since last year - I got some real surprises in the detail of the new model announced today.

You'd expect a product which contains over 500 internal flash interfaces and a lot of hardware controller technology to be fast compared to a RAID array built from COTS flash SSDs, but due to some clever tricks and design choices (which Violin pointed out are patent pending IP) the overall product achieves an edge over potential competitors in a number of key areas. You'll have to ask the company for more details - but are some key points.

It starts fast - and stays fast. Violin says the sustained performance does not drop off in the same dramatic way as in most other flash SSDs.

Most publications mistakenly attribute the performance droop in flash SSDs to "fragmentation" - but that's not correct. The effect is due to what Violin calls "garbage collection".

That includes a bunch of stuff like... Can the internal controller maintain its house-keeping, supply of erased blocks, wear leveling, and virtual mapping at the fastest speed? Or is there an inbuilt halo effect - which means the product looks good in benchmarks - but degrades over time? I already discussed these in my article Can you trust your flash SSD's specs? Violin's got a good story on this - which goes into a lot more detail.

Violin's specific implementation of "RAID like protection" solves 2 problems at the same time.

1st - it uses less flash chips than would be needed in a traditional RAID built from COTS SSDs to achieve the same level of fault tolerance. The lower chip count also reduces electrical power, increases MTBF and helps make the box more competitive.

I asked "Are you seriously hoping to compete with boxes stuffed with commodity SSDs?" - The answer - it's going to a big market - and being competitive due to intrinsic design is a much better option than (in my words) simply lowering margins (which is where the consumer SSD market is today - and where the enterprise SSD market could be tomorrow.)

2nd - it enables the internal controller to perform a write erase simultaneously to a virtual block which is being read. This is what Violin calls "Non-blocking erase". It's a feature (which when added to everything else) enables the product to offer an industry leading envelope of R/W IOPS and reliability for an enterprise acceleration flash SSD.
.
Can You Trust Your Flash SSD's Specs?
Editor:- I've noticed is that the published specs of flash SSDs change a lot -from the time a product they are first announced, then when they're being sampled, and later again when they are in volume production.

Sometimes the headline numbers get better, sometimes they get worse. There are many good reasons for this.

The product which you carefully qualified may not be identical to the one that's going into your production line for a variety of reasons...
ssd specs article And here's another thing to worry about...

The enterprise flash SSDs which you benchmarked yourself - may surprise you by running much slower when deployed in your own applications due to common "halo" errors which are implicit in the set ups of many performance test suites which were originally designed for HDDs. ...read the article
.
Cypress Integrates Non Volatile Static RAM in Controller

SAN JOSE, Calif. - September 22, 2008 - Cypress Semiconductor Corp. today introduced the industry's first device to integrate a non-volatile static random access memory (nvSRAM) and a programmable system on chip.

The new PSoC NV family combines the flexible design capabilities of Cypress's flagship PSoC architecture with an infinite endurance nvSRAM in a single package. PSoC NV devices integrate configurable analog and digital circuits, controlled by an on-chip microcontroller, providing both enhanced design revision capability and component count savings. The secure-store data logging devices target the computing, networking, telecomm, automotive and industrial markets. ...Cypress Semiconductor profile

Editor's comments:- I thought you might want to think about the potential of using nvSRAM in flash SSDs.

In the future I think that very high performance flash SSDs will incorporate small amounts of non volatile RAM technology - to get better internal controller performance.

Current designs use large flash arrays and relatively small volatile RAMs. As there is always a risk of power failure - the way that the internal state of the SSD controller is managed is a defensive compromise which protects against data loss - but doesn't permit the same level of random access time that you would get if the designers were confident that the RAM was non volatile. This design trade-off is compatible with high throughput - but (even in the best designs) results in an order of magnitude worse latency than RAM SSDs.

For those interested in seeing how nvSRAM technology could be integrated in an SSD controller take a look at the article called nvSRAMs eclipse battery-backed memory. The example given in the article is for industrial products. But remember that most innovations in SSDs originally came from either hard industrial or military applications.
.
600MB/s SATA-3 SSD Controller in the Pipeline

Editor:- in August 2008 - Indilinx unveiled its 230MB/s flash SSD controller, and said it is working with MOSAID Technologies on a 600MB/s SATA-3 design.
.

storage search banner

STORAGEsearch Solid state disks flash SSDs RAM SSDs the Solid State Disks Buyers Guide Storage news
STORAGEsearch is published by ACSL