IBM’s latest announcements in its Storage Tank strategy are impressive but somewhat limited in scope compared to the intentions of the Storage Tank project.
Instead of providing open storage across heterogeneous platforms, both at the front end (server clients) and back end (different vendors disk arrays and backup systems) Storage Tank provides a SAN File System (SFS) merely for Windows 2000 and AIX clients and support for IBM disk arrays. The software support is entirely IBM-based, and very limited, with just DB2 integrated and Tivoli Storage Manager providing backups and Tivoli Storage Resource Management providing management information.
IBM says it is planning to talk to other vendors and mentions Oracle, Veritas, McData and others. This is small beer and, were it not IBM talking, one might be inclined to dismiss it as stretching the marketing-speak a tad too far. But as it is IBM talking, perhaps we should give it the benefit of this doubt.
However, SQL Server support should not be expected anytime soon. Interestingly, the main support is for Windows 2000 and AIX clients. The obvious standouts are Linux, about which IBM is so energetically positive, and OS/400. Other Unixes not included are HP-UX and Solaris. The Linux omission is very odd as the actual virtualisation engine runs on a pair of Linux boxes. One might imagine that the use of Linux client servers in the market space IBM is aiming for is very limited. That will most probably change.
Mainframe clients are not included either. They have their own virtual disk facilities but we might expect them to slowly begin to support SFS/SVC technology over the next five years. There us probably a revenue hit to be taken by IBM here as an SFS/SVC system for Windows and Unix is probably lower priced than equivalent mainframe facilities.
Disk array support will extend out of IBM arrays to Hitachi Thunder series arrays, shortly. This is no surprise as Hitachi has bought IBM’s drive manufacturing business and is IBM’s drive partner. True heterogeneity on this front will come with support for EMC and StorageTek arrays. Don’t hold your breath.
Ditto on the SVC supporting other in-fabric platforms, such as the EMC one. McData’s is a distinct possibility now that McData has an intelligent switch capability, courtesy of its recent Arohi acquisition.
The virtualisation engine has been developed from one used inside IBM's Shark disk arrays. In effect, it is a drive array engine applied SAN-wide. Previously IBM sold the DataCore SANsymphony product but this partnership was terminated earlier this year. Virtualisation is the necessary technology needed to provide in-SAN fabric provisioning and management facilities and IBM, HP, EMC and Cisco are all progressing efforts to provide versions of a utility storage service based on it.
The target markets are the highly information-intensive financial services, retail, and life sciences, where billions of files, and potentially petabytes of online storage, are going to be needed. The system is being piloted at CERN where 5-8 petabytes of data a year may be generated by 2007. Most global businesses won’t be at this level for many years though. What customers will see is, hopefully, the same swiftness, convenience and familiarity of current file systems, with the scalability and sharing of a SAN, all provided transparently across Windows and AIX servers, potentially thousands of them.
It is clearly a large scale computing technology and one that will be used in-house by large information-intensive organisations, or by vast server farms offering facilities in a hosting, ASPS, or outsourcing manner. It won’t trickle down to the ordinary enterprise or SME space anytime soon.
IBM has set up a special sales force to deal with its SFS/SVC products which includes both direct and reseller elements. This suggests that the average system sale value could be quite high, well in the six figures area.
The company bangs both its on-demand and autonomic drums, as we might expect. On-demand storage is provided via pre-set policies which tell the SVC what to do when file space is requested by an application in a server. There are in-SAN data protection facilities like FlashCopy and we can expect the integration of tape libraries and some kind of storage hierarchy facilities to be added as well.
Storage Tank is not a single product; it is a road forward and IBM has delivered a significant but, in the scheme of things, small, step on that road. The architecture and products look promising but limited and its development should be keenly watched, not least for responses by EMC, HP and others.