IBM builds biggest-ever disk for secret customer
120 PB monster 'for simulation'. Nuke labs?
Analysis Flash may be one cutting edge of storage action, but big data is causing developments at the other side of the storage pond, with IBM developing a 120 petabyte 200,000-disk array.
The mighty drive is being developed for a secret supercomputer-using customer "for detailed simulations of real-world phenomena" according to MIT's Technology Review, and takes current large-array technology trends a step or two further.
IBM Almaden storage systems research director Bruce Hillsberg says that 200,000 SAS disk drives are involved, rather than SATA ones, because performance is a concern. A back-of-an-envelope calculation suggests 600GB drives are being used and Seagate 2.5-inch Savvios come to mind.
Won't lose any data for a million years? Really, give it a rest, this is marketing BS...
We're told that wider racks than normal are being used to accommodate the drives in a smaller amount of floorspace than standard racks would require. Also these racks are water-cooled rather than fan-cooled, which would seem logical if wide drawers crammed full of small form factor (SFF) drives were being used.
Some 2TB of capacity may be needed to hold the file data for the billions of files in the array. The GPFS parallel file system is being used with a hint that flash memory storage is used to speed its operations. This would indicate that the 120PB array would include, say, some Violin Memory arrays to hold the meta-data, and would scan 10 billion files in about 43 minutes.
RAID 6, which can protect against two drive failures, is not enough – not with 200,000 drives to look after – and so a multi-speed RAID set-up is being developed. Multiple copies of data would be written and striped so that a single drive failure could be tolerated easily. A failed drive would be rebuilt slowly in the background. The rebuild would not slow the accessing supercomputer down much if at all. A dual-drive failure would have a faster re-build. A three-drive failure would get a faster rebuild again, with, we assume, the compute side of the supercomputer slowing down somewhat due to a lower array I/O rate.
Hillsberg doesn't say how many drives could simultaneously fail. The MIT article text says the array will be "a system that should not lose any data for a million years without making any compromises on performance". Really, give it a rest, this is marketing BS. Having it work and not lose data for 15 years will be good enough.
We're interested that homogeneous disk drives are being used – presumably all the data on the array will be classed as primary data, apart from the file meta data which will need a flash speed-up. That means no tiering software is needed.
There will be lessons here for other big data drive array suppliers, such as EMC's Isilon unit, DataDirect and Panasas. It will be interesting to see if they abandon standard racks in favour of wider units, SFF drives, water-cooling and improved RAID algorithms too. ®
Storage-heavy supercomputer simulations are used in such tasks as weather forecasting, seismic surveying and complex molecular science - but there would seem no reason to keep any such customer's identity a secret. Another area in which supercomputer simulations are important is nuclear weapons: without live tests it becomes a difficult predictive task to tell whether a warhead will still work after a given period of time. As a result, the US nuclear weaponry labs are leaders in the supercomputing field.
They received my order for the porn storage farm!
I was getting worried by the silence for a while there.
MTBF and operational lifetimes
MTBF absolutely is critical when designing large storage arrays. It's the key number (along with mean time to recover - MTTR) that tells you the likelyhood of a double failure within in any one raid set. It is the size of the raid set, the number of failures it can withstand and the total number of raidsets that matter on a 200,000 array storage device (using RAID in its most general sense of storage redundancy). Note that MTTR on RAID sets including modern, very large disks can be measured in 10s of hours. One of the reasons multi-level protection is becoming more important - the other being unrecoverable read errors short of complete device failure that prevent RAID rebuilds).
There is a big question over just how trustworthy MTBF figures are. Google did a study a few years back demonstrating that failure rates are not random. They tend to be associated with particular batches, models and manufacturers (annoyingly they wouldn't identify the bad ones). Also, they found that failure predictors, including S.M.A.R.T. stats correlated very poorly with actual failures. I've had experience of that myself where very high and statistically extremely improbable failure rates were observed on a subset of disks over a month. That speaks of non-random failure modes. Those devices exhibited no warnings of any failures.
Note that some arrays make the definition of what constitutes a RAID set an extremely slippery concept. Also, the concept of RAID sets and the consequences of common-mode failures compromising redundancy can make this a very tricky analysis.
Also, I wish people would stop thinking MTBF has anything directly to do with the lifespan of a device. The MTBF is simply the average number of total operational hours that might be expected between failures for a large(ish) population of similar devices. The MTBF figure only applies to devices within given (and not well publicised) working lifespans. We have hard drives now with MTBFs approaching 100 years. However, nobody in their right minds believes these devices will actually run for 100 years before they fail - a decade would be good.
In general, very large storage arrays have to be self-healing with dynamic spares and (within the operational lifespan) will rely on a mixture of re-active and pre-emptive device swapping, but I don't know of storage suppliers who swap drives out at fixed lifespan intervals (although I have known manufacturers swap out batches of suspect drives where excessive early failures have been detected in order to pre-empt catastrophic failures).
Needless to say, manufacturers are less than forthcoming about this...
It's Microsoft I tell you
for installing a very early build of the next release of Windows