Big Blue Makes Moves To Mainstream Db2 Mirror
May 4, 2020 Timothy Prickett Morgan
When IBM brought out its active-active Db2 for i database clustering extensions, called Db2 Mirror appropriately enough, almost exactly a year ago to improve the resiliency of databases and therefore the applications that run atop them on the IBM i platform, we had a few asks. As part of the April Technology Refresh announcements for IBM i 7.3 and 7.4, some of those asks are answered and it looks like another one might be in the works.
The two most important things that we asked for are related, and it is all about making Db2 Mirror available and affordable for the relatively small system sizes that dominate the IBM i installed base. And, to be fair, both Alison Butterill, IBM i product offering manager, and Steve Will, IBM i chief architect, have said from the very beginning that Db2 Mirror was created for the largest IBM i shops that have a need for higher resiliency than the typical shop and can, to be blunt, pay the licensing fees for the Db2 Mirror extensions to the Db2 for i database that fund the development of this active-active clustering feature.
But, we like all technologies to be available to all customers, and think that there are plenty of companies that make anywhere from $100 million to $1 billion in annual revenues that have very strict availability needs for their businesses and just happen to need only modest database engines to run their businesses. And so the first ask was to have Db2 Mirror work with internal disk arrays as opposed to external disk arrays and storage area network (SAN) iron.
With the update to Db2 Mirror announced with the Technology Refreshes, the Save/Restore functions of the IBM i operating system can be used to set up the initial mirrored copy of the database on a secondary system when both machines have internal disk arrays. Prior to this, Db2 Mirror only worked with external disk arrays that had mirroring or flash copy support built into their hardware. There are some caveats if you drill down into the details, which are available here, and at least for now, the internal storage gas to flash storage either in the SAS SSD format or PCI-Express flash cards supporting the NVM-Express flash cards. You can’t use internal SAS or SATA disk drives and you can’t use a mix of disk and flash internally either. So, I was initially very happy when I read that, and then a little annoyed. But this is probably due to a need for high performance replication, and a few disk drives may not be fast enough to do the replication and not affect response times on the primary machine. If we had known this key detail when we were talking to Butterill and Will, we would have asked more about it.
For the chart above, we assumed any internal storage would be usable from this chart, because in our brains, like many of you, when we hear “Direct Attached Storage” we hear “Direct Attached Storage Device” or DASD and DASD of course has meant “disk drive” for decades. It’s cool, our bad for not reading correctly. Flash is better than disks anyway for a lot of reasons, if nothing else than to help speed up OLTP and batch performance.
IBM has also let us know that Db2 Mirror was always set to have a small tier and a medium tier price from the get-go, roughly corresponding to the P20 and P30 software tiers for the IBM i operating system, it has not yet decided on making that small tier price different from the $20,000 per core price it has set for the medium tier. We think that Db2 Mirror can and should scale with the cost of the IBM i operating system and database, period, and moreover, and somewhere around half the cost of the IBM i tier price seems about right. So, call it $1,500 for P05, $7,500 for P10, $22,000 for P20, and $29,500 for P30 and higher seems about right. It is not clear why DB2 Mirror is using AIX tiering – small, medium, and large T-shirt sizes – in an IBM i world.
In addition to these new features, IBM has added an application evaluation tool for Db2 Mirror that allows a logical partition with the base Db2 Mirror software installed to point to another logical partition and analyze the database applications on that other system (or within the original system) to see what the potential is to use Db2 Mirror to make an active-active database cluster. For instance, it can be used to examine the eligibility of applications to be replicated with Db2 Mirror, like this:
And it can be used to examine object statistics, like this:
And it can be used to dig into the journals of the system and see how they might be tweaked to boost the performance of Db2 Mirror, like this:
The updates to Db2 Mirror come to all IBM i 7.4 customers who are on Software Maintenance, of course, and that means all of you. Hopefully, IBM can get the pricing on Db2 Mirror more consistent with IBM and its integrated Db2 database soon, and we can see if smaller shops want the higher availability it affords.