Big Blue Launches XIV Clustered Storage Arrays
August 18, 2008 Timothy Prickett Morgan
The product might still be bearing the cryptic XIV name its founder, Moshe Yanai, gave the storage company he founded after leaving EMC, but now the XIV Storage System also has the IBM 2810 Model A14 designation, and this new type of clustered storage system is available in the Big Blue catalog.
IBM acquired XIV, which is based in Israel and which was still operating in stealth mode as it was preparing to launch its products, on January 2 of this year. Yanai, who is the engineering genius behind EMC’s wildly successful Symmetrix line of disk arrays, left EMC and joined XIV a few years after it was founded in 2002. According to press reports at the time, Yanai had kicked in some of his own dough to the $3 million in funding that enabled XIV to get started, and IBM reportedly paid between $300 million and $350 million for the upstart storage array maker. The company was founded by a bunch of graduates of the 14th class of the Talpiot technical university, which is run by the Israeli military, hence the XIV name. XIV developed its clustered disk arrays under the name Nextra, but Big Blue has killed off that name. And presumably, IBM has spent the past several months shifting the hardware technology base used in the clustered storage over to its own devices. (It is not clear what servers and disk modules XIV was using in its prototypes.) The result is a perfectly undistinguished rack of black boxes in a rack, as you can see.
As we reported back in January when IBM bought XIV, the clustered disk arrays take the idea of clustering servers and apply it to disk arrays and the file systems they support, and then add a few twists to the recipe that are key for modern storage devices. The basic node of the XIV array is an X64 server of some sort, and these controlled 7200 RPM SATA disk modules in the original Nextra designs; these nodes also ran the system software on the arrays, which implemented a data protection algorithm called RAID X, a kind of highly distributed RAID data protection that stores recovery data all around the array, not just within a RAID group, and which can recover a 500 GB disk in about 15 minutes instead of the six to 25 hours it takes in a RAID 5 array. (Part of the reason for the speed up is also that the RAD X algorithm is away of the empty areas on a disk, and doesn’t try to rebuild them.)
IBM’s formal announcement last week of the 2810 Model A14 was still a little thin, but there was a little bit more detail. The cluster supports up to 180 1 TB SATA disks, which provides up to 80 TB of usable capacity. (Wow, that RAID X algorithm, which includes a lot of redundancy for data, seems to eat a lot of capacity.) The clustered array also sports capacity on demand, utility-style activation of latent capacity inside the box, with customers able to start out with 21.2 TB activated and then turn on what they need from there. Each cluster has 24 4 Gb/sec Fibre Channel ports and six 1 Gb/sec iSCSI links for linking to host servers, and it has a total of 120 GB of cache memory. A fully configured machine has nine data modules with the disks and six interface modules (which is where the software runs and where hosts link in). The system software on the unit has synchronous remote mirroring, thin provisioning, data migration, and writeable snapshotting, all of which are managed by the XIV Storage System Software V10.0.0. (Yes, that is a silly name for a new product, but who wants to admit to something being a 1.0 product?)
For now, this is the only configuration of the XIV machine available from IBM. You buy a fully loaded box or nothing at all. The 2810-A14 supports Unix, Linux, and Windows hosts at the moment–which Unixes, IBM did not say–but will almost certainly support all of IBM’s operating systems, just like its DS family of Power-based disk arrays do. IBM added in its announcement that sometime in the second half of this year, IBM would add support for the XIV clustered disk to the SAN Volume Controller, an out-of-band storage controller that IBM has supported for years and which will increase the number of operating systems that can access the XIV unit.
IBM didn’t think anyone cared about pricing on such a machine, so it is keeping prices to itself. There ought to be a law against that.