EMC on XtremIO SSD brickup ballsup: Its lifetime downtime is under 3 minutes
Replace a failed X-Brick SSD once every 5 years
No DOAs here: EMC’s XTremIO arrays are expected to have less than three minutes downtime in their rated life, with X-Brick component SSDs failing once every five years or so.
We have reported that the DOA rate was too high and the co-founder and general manager of EMC's acquired XtremIO business, Ehud Rokach, has written blog which is partly a counter to that, writing that our “article presented speculations that may lead to false conclusions.”
He says that, using data obtained by monitoring “hundreds of XtremIO X-Bricks at customer sites globally and across all major verticals”:
- XtremIO delivers world class 99.9999 per cent (six nines) field-proven availability (less than 32 Seconds of unavailability in a year, and less than 3 minutes of unavailability over the lifetime of the product.)
- Our SSD Mean Time Between Part Replacement (MTBPR) was field-measured to be 922,240 hours, or 105 years.
- Our Annual Replacement Rate (ARR) for SSDs was field-measured to be 0.009. For an entire X-Brick (holding 25 SSDs), the probability of encountering SSD failure at any time during a 1-year period equals (1-0.991^25), or 0.2.
- A 0.2 ARR means that on average, based on our actual field data, you’ll need to replace a failed SSD (Due to a non-endurance related failure. …) in an X-Brick roughly once every 5 years.
Pretty darn convincing. He hammers away: “Our actual measured field performance demonstrates exceptional SSD and array-level reliability. Since initiating XtremIO’s Directed Availability program we have seen a grand total of single-digit SSD failures out of thousands of deployed SSDs.”
To refresh your memory, in the comment we reproduced from Xtremio chief techie Robin Ren, he said: “I am not too happy about our field hardware failure rate for many reasons. However, the vast majority of failures – we have seen over 150 X-Bricks so far – [pauses] in real customer environments … [and] another 200 systems internally. I think we have seen a lot of DOAs in terms of drives.”
Ren spoke in October by the way, several months after Directed Availability started.
Rokach says our story, based on Ren’s comments, “referenced (unknowingly) a couple of early DOA SSD failure events during Beta, prior to product being released for Directed Availability. The two failures during pre-release Beta were analysed, and corrective action applied (firmware update). Not surprisingly, ever since we started Directed Availability and to this very day, we have seen no excess SSD failures of any kind (in the field or DOA). This is indeed a non-issue.”
Happy to hear it. ®