Bookmark and Share
RSS

Recent Posts

PowerHA and Zero-Downtime Backups on IBM i

February 09, 2016

I recently decided to solicit some ideas for blog topics from the community. I tweeted my invitation for comments and received several replies–some via Twitter and even more via Facebook (which posts my tweets as status updates).

You-and-i-figure-1.jpg

 One of the replies was the following, and it got multiple reinforcing comments and “Likes.”

You-and-i-figure-2.jpg


Well, the people have spoken, and it’s a good idea too (as anyone who uses Facebook knows, the good ideas are only a proper subset of the popular ideas) so here we go – a blog post about that topic.

Straight to the PowerHA Source

To make sure I get the information right, I decided to go straight to one of the sources for PowerHA knowledge, Steve Finnes, the product offering manager. Here’s a portion of our discussion.

Steve Will (SW): So Steve, you saw the tweet. Larry’s asking for a post about PowerHA, SAN and FlashCopy. He mentions two scenarios–backups with no downtime and quick copying of production data for testing. Let’s get one thing straight first: are these things customers actually do with PowerHA? Is one more common than the other?
Steve Finnes (SF): Oh, absolutely, customers are doing these things. One of the keys is the move of many of our customers to SAN. SAN storage has come a long way over the past decade. I would have to say that if you’re serious about HA/DR you’re going to be deploying SAN storage and PowerHA. I should point out that for those customers with smaller shops and internal disk, PowerHA with non-SAN storage works wonderfully and is very inexpensive. We can talk about that in another discussion. But, to your question, FlashCopy is a part of copy services and it enables the client to quickly create a point in time copy, which as Larry points out can be used for tape backup operations and also for testing purposes. PowerHA for IBM i has integrated the SAN storage copy services functions including FlashCopy as an integrated extension of the PowerHA solution. The backup operation can be completely automated from the flash operation to the save to tape. It eliminates one of the oldest and most vexing problems; that of getting the backup done in the available time slot. But its gets better. The save window is essentially eliminated with this technology. The testing capability is also widely used. You have this point in time copy; you can run tests against it and throw it away. I like the DR compliance test use case; they make a flash copy at the DR site, run a test bucket and discard it. Of course how one does DR compliance testing depends on the customer requirements, some shops require a complete roll over to the DR site, the great thing about a PowerHA cluster is that that is actually a simple operation.

SW: Maybe before we go on, we should first let people know what FlashCopy really is. How do you describe it to customers?
SF: Well, I like to use pictures because it’s easier to explain that way but here we go. FlashCopy is part of the Copy Services function provided by IBM SAN storage servers. There is a production volume (IASP) and a target volume in a different partition (all in the same cluster). The storage server pauses production momentarily and creates an identical point in time copy (the flash copy) into a flash copy partition. Production resumes and that FlashCopy is now ready for whatever purpose you choose. There are several different ways that a FlashCopy is used. For example, some in the banking industry create a FlashCopy before the nightly batch run starts. In case of an issue occurring during the batch run, the FlashCopy data can be used  to restart the job.

SW: How is PowerHA important for solving those business problems? In other words, if you just had a SAN capable of FlashCopy, and not PowerHA, what would you be losing?
SF: Without PowerHA the SAN Copy Services can still be used; however, it’s what people often refer to as the “full system” replication technology. With PowerHA, the database and IFS are in the IASP, which is the entity getting flashed and this operation is fast. Same goes for the replication services whether Metro Mirror or Global Mirror. Full system operations on the other hand, are replicating everything as well as flashing everything. To activate a copy a full IPL is required and this takes longer. With the PowerHA cluster the IASP is being varied on which is relatively fast..Since there are active operating systems on all of the partitions in the cluster, you are able to do real time software maintenance on say a target node (partition) while running production on the source node. So the business value of PowerHA is about planned and managed outages as well as operational simplicity.

SW: What SAN devices can PowerHA use to accomplish these backup and test data scenarios?
SF: For example, a DS8000 storage server is hosting various partitions so internally to the SAN server, it flashes from one partition to another and then that FlashCopy is connected to BRMS or Robot to upload the data to the tape library. To use Copy Services including FlashCopy, you do need an IBM storage server. You can’t do this type of operation with internal disk. PowerHA supports the DS8000, Storwize and the SVC copy services technology. Storwize is popular in the IBM i midmarket and there are several choices.

SW: How close do the two systems have to be? Or perhaps I should ask how far apart CAN they be?
SF: The source and FlashCopy partition share the  same storage device. The production systems and storage servers in a PowerHA cluster can be in the same room, the same campus or half way around the world. Normally you want to have a close proximity for the HA operations and a distant site for DR operations. The close proximity part of the cluster enables either a shared storage configuration or Metro Mirror configuration and in both of these configurations the solution provides an RPO of zero. That is, the data being replicated via Metro Mirror is synchronous to the application state. All data that gets paged to the storage subsystem is mirrored real time and therefore the distance is finite before the write operations slow down production too much. When we go unlimited distance we use Global Mirroring, which is asynchronous to the application state while at the same time maintaining consistency boundaries (or sync points if you will). When you’re using asynchronous replication technology you are trading performance with unlimited distance for an RPO greater than zero. If there is an unplanned outage event, some amount of data didn’t make to the consistency boundary and is therefore lost. Having said that, we do have customers with two systems in two sites separated by a great distance and that is their entire PowerHA cluster configuration. We also have customers who are using Metro Global Mirror for a three-site PowerHA cluster configuration. This is becoming more frequent in the banking industry and it enables three copies of the data in three separate locations all in one centrally managed PowerHA cluster. The FlashCopy operation can be performed on any of the systems in the cluster.

SW: What else should we tell potential users about this PowerHA and SAN capability?
SF: PowerHA is ultimately about automation and certainty of outcome. If you’re not doing it now you will likely be doing it in the future. There is a steady movement to the IBM SAN storage technology and PowerHA. As I mentioned when we started this discussion, the smaller customer shops with internal disk are well served with PowerHA and geomirroring replication. I would estimate that there are now over 1,200 of these smaller shops so deployed around the world. Customers love the simplicity of use and the price tag. Geormirroring comes with the PowerHA license. We should talk about this in another discussion.

SW: Where can people find more information if they want to learn more?
SF: Check out our PowerHA website, get the new set of Redbooks that are just now being published externally (Preparing for IBM PowerHA SystemMirror for i, SG24-8400; IBM PowerHA SystemMirror for i: Using DS8000, SG24-8403; IBM PowerHA SystemMirror for i: Using IBM Storwize, SG24-8402 and IBM PowerHA SystemMirror for i: Using Geographic Mirroring, SG24-8401); I can tell you that they are excellent and will replace the current ones that are out there. You can talk to your business partner or the IBM Lab Services team. For more information, you can also contact Mark Even.

What Do You Want to Hear About?
So, there you go–a post on a topic which comes straight from the community. I’ll be looking for other ideas from you, so please keep sending them to me on Twitter, Facebook or via e-mail. I have a few topics which I need to do, as well, with some announcements approaching, but I always like to know what our users are interested in hearing about.

So, until next time, think about the business challenges you have, and how IBM i and its technology can help solve them.

Posted February 09, 2016 | Permalink

Post a Comment

Note: Comments are moderated and will not appear until approved

comments powered by Disqus