Re:A Tour Inside CloudFlare's Latest Generation Servers
by eastdakota on Monday July 22, 2013 @02:03PM
(#6085362)
That hasn't been our experience. While we've optimized our file system to minimize wear, we do an extremely high volume of reads and writes on our SSDs. We have many SSDs (previous generations) that have been running full steam for 3 years in production. We've been pleasantly surprised with the number of write cycles they can endure without failure.
Re:A Tour Inside CloudFlare's Latest Generation Servers
by peterwwillis on Monday July 22, 2013 @02:46PM
(#6085690)
How do you measure writes in this case? Do you use SSD write caching? Is your filesystem caching the writes? Would love to see some stats/graphs to show real-world metrics of disk resilience.
Re:A Tour Inside CloudFlare's Latest Generation Servers
by eastdakota on Monday July 22, 2013 @02:58PM
(#6085771)
We'll do a post at some point on file system benchmarks and what we do to get the most performance and life out of our SSDs.
Re:A Tour Inside CloudFlare's Latest Generation Servers
by bifrost on Monday July 22, 2013 @11:30PM
(#6088080)
Care to share the brand/variety? So far I've killed Intel, Kingston, Micron, OCZ and a couple offbrands. Wasn't terribly surprised by the offbrands, a little miffed at the Intel though...
Re:A Tour Inside CloudFlare's Latest Generation Servers
by alexchamberlain on Tuesday July 23, 2013 @05:23AM
(#6089002)
You haven't considered the load balancing. You don't write everything to a single disk, you stripe it across multiple disks.
Uh, I've never had one fail gradually...
> "Intel reports that the 520-series drives have a mean time between failure (MTBF) of 1,200,000 hours (about 137 years)."
Yes, but they have a maximum write cycle, you can blow through the average consumer drive in a month and a half of concerted writing.