Andy Bechtolsheim on the Solid State Storage Revolution

At the MySQL conference Andy gives a talk that is up on YouTube.



He gives a good overview of the hardware changes, but doesn't talk about the software challenges, where many of the assumptions and algorithms that are built into all the common operating systems, filesystems and databases are simply wrong. There is a lot of work to be done here. My own work at Sun involved several attempts to point this problem out over the years, and if I had done a better job of getting traction for my ideas, there would be a lot more preparation and research in this area by now. The specific examples include trying to explain the benefits of embedding NVRAM products into servers during the 1990's, and doing extensive testing with solid state disks around 2002. The current size and cost benefits that we see were very predictable, and so are the problems. The SSD's at the time were very expensive, but they were perfectly capable of supporting research and development of new algorithms, and an opportunity was missed.

The next thing that is coming is non-volatile RAM (all the RAM in the system), in case anyone is listening. There need to be fundamental changes in operating systems (user level memory protection models etc.) and we then have the ability to build truly diskless database servers, with safely persistent in-memory databases, in user space, no device drivers.

Intel high speed SSD

Here is Tech Report's review of Intel's high speed SSD, which confirms the trend I've been talking about for a while. SSD's were always faster for random read and write, now they are faster for sequential read (250MB/s), and the "extreme" Intel model is faster for sequential write as well (170MB/s). They use less power and have comparable MTBF to the best enterprise disk drives.

The remaining disadvantages of total size and cost are being eaten away over time...

Samsung 256GB SSD coming later this year

The current generation of SSD's are smaller and slower than regular disks in the same form factor, but this is starting to change as products like this Samsung 256GB SSD reach the market. It has a 200 Mbyte/s read speed and 160Mbyte/s sequential write speed, in a standard 2.5" disk form factor.

So the sequence goes like this:

1) SSD's will be faster for random access (already happened)

2) SSD's will be faster for sequential access (coming later this year)

3) SSD's will be higher capacity (maybe next year?)

4) SSD's will be lower cost per GB (when production volumes ramp up)

The advantage in random access and reliability (due to no moving parts) means that relatively fewer SSDs are needed than spinning rust disks to provide the same availability and performance, so the end user cost per configured GB should switch in favor of SSDs earlier.

Flash based SSD from Samsung

A nice review of a Flash based SSD in Engadget. This 64GB drive from Samsung is a drop-in replacement for a 2.5" hard drive. Its fast, but still too expensive for common use.

Flash Solid State Disk (SSD) for Millicomputers

The days of keeping bits on spinning rust are coming to an end....

Both Samsung and SANdisk have announced 32GB SSDs. SANdisk's comsumes 0.9W max (competing disks take 1.9W) and fits in a 1.8 or 2.5" drive form factor with ATA interface. The sequential performance of these SSDs is similar to normal disks for reads, a bit slower for pure writes, but as soon as you start doing random reads or writes they are an order of magnitude faster than disks. The smaller the random accesses the bigger the relative speedup. The latest announcement from Samsung is a 1.8" 64GB version, and there is some discussion about the growth of this market in the press release.

This makes perfect sense for millcomputers. Small millicomputers can be directly connected to gigabytes of NAND flash via the SDIO interface, and larger millicomputers can use ATA interfaces to connect to flash-SSDs. The extra random performance of the SSD offsets the lack of disk spindles in a compact design and will make IO intensive workloads extremely competitive for millicomputing.

The MTBF (reliability) of SSDs is also far higher than disks. A mirrored pair of disks may be replaced with a single SSD since it has much higher reliability. This helps offset the current price premium paid for the SSD.

In the past SSDs have been built using technologies that were far more expensive than disks. Flash based SSDs have now reduced the gap, and the trend is that SSDs will eventually become bigger and cheaper than disks, the only question is when, and my answer is sooner than you think!

Update: here is a detailed benchmark review from Tomshardware.com.