Sun preps Flash-based servers, storage for '09
Sun Microsystems plans to introduce Flash-based servers and storage systems in early 2009, according to company sources.
This would build upon Sun's earlier embrace of flash in its ZFS storage platform. Sun officials have argued that Flash pushes down the total operational cost of storage, particularly when it comes to data or performance intensive applications. Earlier in the year, there had been speculation about a faster move to Flash before the end of 2008.
Sources say they expect performance gains of between 10 percent and 20 percent in the upcoming systems. An announcement is expected in the first quarter of next year.
"Certainly, there are lots of Web 2.0 customers out there who are trying to get performance gains so anything that pushes out their scale is going to be valuable," said one executive familiar with the plans.
Sun Chief Technology Officer Greg Papadopoulos told a gathering of analysts and reporters in brief welcoming remarks Tuesday evening that "Flash memory will be the catalyst...that leads to a rethinking of storage." He did not get more specific.
Earlier in the day, Sun disclosed that it had stopped accepting new customers for its Network.com pay-per-use utility service. Dave Douglas, senior vice president of cloud computing and developer platforms group, acknowledged that it had been slow going for the 2-year-old project.
"I think the model makes sense...it definitely (appealed to) a set of customers but not as big a set as we would have hoped," Douglas said.
Lew Tucker, who was hired in August to take over as vice president and CTO of Sun's Cloud Computing initiative, left the door open to a "retooled" Network.com sometime in the future.
Charles Cooper has covered technology and business for more than 25 years. Before joining CNET News, he worked at the Associated Press, Computer & Software News, Computer Shopper, PC Week, and ZDNet. E-mail Charlie. 
I'm not sure flash memory is going to improve the performance of Web 2.0 apps. As I understand it, flash memory will not be used to store databases, only data that does not change a lot. This basically means operating system and applications, static HTML and dynamic pages. Once the system goes up it will have the operating system and application code in RAM, and the static internet content (and dynamic bytecode) will get cached into RAM after its first access.
Flash memory could be useful for servers that hold large downloads, and useful for increasing the speed of the first access of a file, but I'm not sure it's especially a good answer for servers. I would be especially worried about databases as they can have frequent writes and large storage demands - neither of those things really suits flash memory very well.
Hard to see flash competing with mechanical drives that, at least for consumers, are almost down to $50 per terabyte.
- by peeweehoi December 17, 2008 6:16 PM PST
- I run a web presence with around 10m weekly visitors. A huge majority of them want access to the same content. If flash memory lets me avoid short stroking 15k rpm drives, I save money, power, get better performance, and lower my cost. No one in the industry is doing this, but from the update I just got on what Sun's up to, they may have actually solved the problem - if they have, they're going to see a lot of interest from folks running high scale web 2.0 companies. Latency kills us, ZFS/flash kills latency - simple as that.
- Like this Reply to this comment
-
(8 Comments)