26th October 2009
Day 2 of the Gold Briefing, and I’m not the only one to see the irony of suffering from information overload at a data management conference!
Yesterday I talked about IBM’s grand vision for Information-Led Transformation and the way it builds upon the innovations and acquisitions made over the past few years. I also mentioned that, important though these higher-level initiatives are for making today’s enterprises more agile and efficient, there is an ongoing requirement to continually enhance the underlying data management technology in order to address more fundamental issues such as availability, scalability and total cost of ownership.
So today I want to spend some time on a technology that was originally supposed to be announced at IOD, but for various reasons actually broke cover a few weeks ago at the IDUG Europe conference in Rome. It’s called DB2 pureScale and in my opinion it is the single most significant advance in the DB2 to LUW product in the last decade.
For over 15 years, DB2 for z/OS has provided an option to dramatically improve scalability and availability by utilising a shared disk clustering solution known as data sharing. Large enterprises all around the world have been taking advantage of this architecture ever since, allowing them to grow their workloads and support true 24×7 availability requirements. pureScale implements the same fundamental architecture within DB2 for Linux, Unix and Windows, allowing IBM to use its extensive experience in this area to deliver a mature solution that is significantly different (and in my opinion superior) to competing architectures such as Oracle RAC.
Under lab conditions and running a read-intensive workload, pureScale can scale to well over 100 members with more than 80% scalability. Given the power of individual servers and the ongoing trend to cramming more and more cores into a single processor, this represents a massive amount of potential capacity and makes DB2 for LUW capable of handling the kind of workloads that would have been firmly in mainframe territory before. And with that massive capacity comes great data availability, with the loss of individual members/servers only causing a temporary “blip” in processing under most conditions.
In its initial form, pureScale is limited to an AIX environment running on IBM’s Power 550 Express and Power 595 machines, with availability expected in December 2009. This obviously limits the appeal of the technology for shops that have other DB2 running on other platforms, so it wouldn’t surprise me to see pureScale being made available on other platforms in the near future. The technology is also a natural fit for mission-critical packaged applications such as SAP and Siebel, so I’m expecting those vendors to move swiftly to exploit this technology and take advantage of the significant scalability and availability improvements it can provide for their systems.
That’s it for today. Tomorrow I’ll talk about some of the announcements made during the first day of the IOD conference, and also cover a few interesting items for the mainframe fans out there.