Wednesday, August 6, 2008

2008 Flash Memory Summit and a vacation to boot!

Like many people that I have talked to recently, I have found myself buried under a mountain of work that seems to have no end in sight. Every time I climb my way to the top of one peak, another pops up in its place and taunts me to keep going. This week is especially busy as I am trying to finish things up before going to the 2008 Flash Memory Summit in California, all while maintaining a blooming social life full of bowling, poker and various other late night excursions.

The conference starts next week at the Santa Clara Marriott and runs Tuesday through Thursday. The preview program lists many exciting keynote speakers from great companies such as Micron, Intel, Sandisk, Samsung and Dell, but I guarantee that the best presentation is going to be on Wednesday morning during "Tutorial 1B: Designing Flash-Based Products". That is when I will be presenting an overview of the Hardware Based Flash Memory Failure Characterization Platform that just might solve our current economic crisis and end world hunger! Ok... well maybe those are a bit ambitious goals, but if you are a product engineer and want to see how long flash memory will work under specific use patters or you just have a keen interest in the failure characteristics of NAND Flash then this is definitely going to rock your world.

The presentation is based on work that I helped complete as part of a team of 5 students from the University of Utah Electrical and Computer Engineering (ECE) department for the 2007-2008 senior engineering clinic sponsored by the Micron Foundation. It was built using an FPGA and custom software and is fully open source, so if you are interested in Flash Memory check out the abstract and slides or visit the project website for more information. Needless to say I am very excited about going to my first technical conference and ecstatic about getting the opportunity to showcase the work we have done.

I am even more excited that I was able to talk our faculty advisor into extending the return trip so that I could squeeze in my first real vacation of the year. After the conference ends I will be heading to San Francisco to meet up with a group of friends that I somehow managed to talk in to flying out with me. We currently have plans to visit various tourist attractions, indulge ourselves in the great night life, and even get in a day or two of wine tasting, all while staying at one of the top 50 hotels in the country! I have been to San Fransisco before with family and enjoyed it very much, but this is shaping up to be a spectacular trip!


Herschel said...

So since you are in the flash business, do you have a magic ball estimate of when Solid State Drives will be mainstream.

Greg Bray said...

I would not consider myself "in" the flash business, but more just someone who has experience working with flash memory for a few semesters. SSDs were the talk of the town at the Flash Memory Summit, as many vendors see it as a great way to open up their product line to new markets. The problem is that there is a lot of hype around SSD drives and exactly what improvements they can and cannot provide over spinning drives. This is best illustrated by the Garner Hype Cycle, which Dean Klein from Micron included in his presentation. The graphs cover various new and emerging technologies and try to gage how long it will take for them to go mainstream. They also plot the items on a hype-curve ranging from "Technology Trigger" and "Peak of inflated expectations" through the "Trough of disillusionment" and on to the "Plateau of productivity" (aka, mainstream).

In 2008 Gartner listed SSDs as just passing the peak of hype and entering the trough of disillusionment, with 2-5 years before they hit mainstream. I think this is about right, as many people do not know the true benefits of SSDs, which really are low power and high IOPS. Also these performance benifits are ONLY if the product was designed correctly: It is just as easy to make a bad SSD drive that has no real benefits over HDs. In theory the lack of moving parts should increase the lifetime and reliability of the device, but due to the nature of flash memory it eventually wears out anyway. Also currently the sequential reads and writes are not on par with high performance spining drives, which means they are only aptly fit for workload that require lots of random access data. Eventually the big boys like Micron, Intel, and Samsung will be able to create multi-plane SSDs that beat even high end HDs on every count, but you can also expect these drives to cost a lot more than even high end HDs.

Personally I think that SSD drives will start becoming standard in high-end laptops, where low power and no moving parts are key features, as well as an option for large Storage Area Networks and servers that really need high IOPS for random data access. But for desktop and workstation applications, SSDs will always cost more then their HD counterparts, so I do not expect to see them as storage replacement anytime soon. You will however start to see new and innovative uses for flash that act as a type of cache or even integrate directly with the memory system. One of the most interesting things at the conference was a new product from Intel called Turbo Memory. It is basically a 2GB flash memory module that can be dropped into a mini-PCIe card or laptop to provide quick load times for the OS and user selectable applications. It is like Vista ReadyBoost and ReadyDrive but using an addon card and more user controllable settings. This type of product could be benifitial to both the retail and business markets as it is a low cost addon (around $25-$40) you can buy to squeeze extra performance from existing hardware. Google "Intel Turbo Memory" or check out these links for more information:

Intel presentation on Turbo Memory at FSM 08
Intel page about Turbo Memory (see PDF overview or video at the top of the page)

Post a Comment