Hi Folks,

I have been searching on the internet on how to set up an active/active cluster in db2 similar to Oracle RAC and haven't been able to find a straight answer. Can this be done? If yes, do I need additional software to take care of the database load balancing. I was reading about xkoto's GridSCALE software, is anybody using that?

Will appreciate your help.

Thanks

Has anybody worked on setting up an active/active cluster in DB2?????

Thanks

Look through the forum ... within the last week or two, there was a discussion about HA and Marcus_A had mentioned a few lines about xkoto's products.

Thanks Sathyaram_s.

If anybody else is also looking for that thread:

Seeking DB2/AIX64 9.1.5 desaster recovery solution

This means that there is really no way in DB2 to achieve a HA solution with NO downtime at all....

Hallo rangupt,

I think xkoto is not the same as Oracle RAC. RAC has one database and has SGA distributed between the cluster nodes. If data buffer is already in SGA the other node need not to read this buffer from the database file. It reads it from other cluster cache with interconnect. In xkoto each node has its own database. It is mirroring.

DB2 does not have the equivalent of RAC clustering for scalability (of course RAC doesn't scale very well for OLTP because of the distributed lock manager, but that is another issue). In DB2 clustering for reliability is done with HADR, which is superior to RAC, since RAC has a single point of failure on the shared disk sub-system (and usually the network).

One thing to keep in mind is that you only need to license DB2 for 100 PVU's on the HADR standby server (which is the equivalent of 1 dual core CPU), regardless of how many CPU's are on the machine.

Xkoto is a 3rd party tool that provides clustering for reliability and scalability, but it only provides scalability on Select statements. Inserts, Updates, and Deletes are single threaded, whereas Selects can be distributed out to any available server in the gridscale cluster (you usually need a minimum of 3 servers for high availability/scalability and a minimum of 2 if scalability is the only concern and you have maintenance window). Xkoto is somewhat complicated, and reports are that it is not completely bug-free, but it does work well if you have OLTP and need scalability of a mostly read-only database, and you don't have a data warehouse (in which case DB2 DPF would be a better choice).

Can DPF combined with HACMP achieve a similar solution?

I would forget about using HACMP with DPF. You can get scalability with DPF in certain situations on an OLTP database, but those situations are limited, and I would not recommend it the vast majority of cases. There is a reason why DPF is now called InfoSphere Data Warehouse Edition (no longer called DPF) in V9.5.

SMP machines can be scaled very large in the same frame, and IBM has a machine (ours is Intel based running Linux) where you can have 4 separate machines effectively linked together as one SMP machine, which pretty much covers all the same failures that a RAC system would protect against. So you can start out with one machine and grow it to 4 machines running regular DB2.

Thanks Marcus.