Comments on: Design Tradeoffs for SSD Performance

By: Aboard the SuperSpeedDrive train | Fun To Think About

Aboard the SuperSpeedDrive train | Fun To Think About — Wed, 17 Feb 2010 05:18:03 +0000

[...] Intel, Samsung or the smallest newcomer, Indilinx. These three companies make the SSD controllers, geekily explained here, though it simply be thought of as the [...]

By: wei Gao

wei Gao — Tue, 02 Dec 2008 11:50:46 +0000

By the way, I am using CMU’s Disksim and SSD Extension for DiskSim Simulation Environment. I hope your paper can give me some help.

By: wei Gao

wei Gao — Tue, 02 Dec 2008 11:49:09 +0000

I am learning SSDs, and I think this paper is excellent! absolutely!!

By: Ed

Ed — Wed, 23 Jul 2008 05:13:57 +0000

Slightly off topic and sorry for asking stupid questions.
The old / current controller chip all uses 4 channel. The only different between faster and slower SSD is the use of Slow MLC, Fast MLC or SLC SSD.

The newer generation uses 8 Channel and achieve 120MB/s + using MLC only. The speed that previous 4 Channel generation would require SLC Chips.

Do we see a future of 16 or even 32 Channel? I dont know if 32 Channel is even possible because a normal 2.5″ would only fit 16 chips inside. Unless the 2 die per chip and uses a channel individually.

Therefore controller chip is obviously the bottleneck in current SSD. As Intel has mentioned. So would be see controller not fast enough to cope with the SSD speed?
The current controller is getting fairly big. I dont know how process it is built on and if die shrink helps.

And How will OS help with performance on SSD. What about File System?
The funny things is that NO ONE has ever mention OS / File system problem when we setup Virtual Drive using DRAM. Which is much faster.

By: rakesh

rakesh — Thu, 17 Jul 2008 11:28:09 +0000

Nice article.You summarized the points very well.
Now as the SSD’s has to support wear leveling and load balancing, Microsoft have to redisgn their file system design.