Skip navigation

Solutions for DB2

2 Posts authored by: Andy Laredo
Share:|

In the previous blog on the Performance Testing Environment, we discussed your desire to set up a testing environment and populating the PTE.  We also spoke a little on utilization.  Since your PTE is being set up for testing code and access paths, it really only has one use in life – testing.  No production work will be done on it, and typically you won’t be doing development work either.  The utilization of the PTE will be 100% dependent on how many tests you run through there.

 

Let’s assume your organization is working 9 hours a day, 5 days a week. Furthermore, let’s say each test runs for 2 hours.  And let’s say to make each test comparable, you want to start with the same data – same amount of data, same amount of customers in the customer database, same amount of transactions in the transaction database, etc., etc.  If you could instantaneously reset the data from the end of a test to the beginning of a test, you would theoretically be able to run 20 tests a week (4 per day (9hrs/2hrs) times 5 days a week).  If resetting the data takes the entire week, then you would only be able to run one test a week.  And your utilization would be 1/20 or 5%.  If you can reset your data in 7 or 8 hours, then you can run one test a day and your utilization would be 5/20 or 25%.  How about if you can reset your data in 15 minutes, then you could run 4 tests a day and you would have a utilization rate of 100% (4 per day times 5 days a week). Now there’s a utilization rate, everyone should envy, even your boss.  No waiting around for recoveries, well you would have to wait 15 minutes for the reset to occur, but everyone deserves a coffee break, don’t they?

 

What does resetting the data entail and what are the possibilities?  Well, a rational way to look at it would be to recover the data back 2 hours or possibly migrate the production data again. I think we can safely say migrating production data again is a rather large undertaking, certainly more than 5 or 6 hours.  How about recovering the data back two hours?  Well I don’t know about your shop, but generally people think of a production recovery (remember the PTE is essentially a copy of production) as also a fairly significant undertaking, also in the 6 hour plus range.

 

Within the BMC Recovery Solution we have had an option called BACKOUT AUTO for years (others are playing catch up and just starting to sell backout, but it’s crucial to note they didn’t add the word auto, and without that there is a lot of think time and manual work involved).  Backout auto allows us to read the log backwards and undo what’s been done over a period of time.  Sticking with our example, let’s say we want to backout 2 hours worth of logs, or testing, or updates.  With backout auto, your mileage may vary, but recovering within 15 minutes is well within the realm of possibilities.  Backout auto has even run faster than 15 minutes to undo hours of log. We don’t have to touch every last production tablespace, we don’t have to lay down every production image copy and roll the log forward – all we have to do is undo two hours of logging.  And BMC, mind you, can and has done that very quickly!

 

And that my friend is how you get the utilization rate for your PTE to be 100% !!!

 

Once again, please contact myself, your local Software Consultant/Account Manager for more information on BMC Backout Auto.

Share:|

Performance Test Environment – Migrating the Data

 

Your company has decided to set up a Performance Testing Environment, referred to as PTE from now on, with production sized data volumes, allowing them to do  some real testing and get accurate SQL optimizer paths through your data.  Imagine the costs involved; the hardware costs to run the environment, the duplicated DASD to hold the data (this must have been a storage vendor’s idea), the operational costs to keep it operating smoothly, the people costs to update it and the people costs when teams of programmers are waiting for the environment to be ready for another test run.  Of course the teams are not standing around waiting; but you get the idea - when someone is waiting for something to be ready to use – it preoccupies them and they mentally wait on it.  At least I do!   And of course you will have a management layer on top of all those people costs.

 

Well, let’s get to the real gist of this blog entry, or actually the first of 3 blog entries.  This entry will discuss populating the PTE environment.  The next entry will discuss resetting the data so your application teams can run their next test or series of tests.  And the final blog entry will delve into those costs to make sure you consider them all.  

 

Populating the PTE environment initially is really a data migration exercise, and BMC has addressed this requirement over the years in several White Papers.  Like this one located here, you will see many flavors of migration; depending on outage reduction, structure modifications along the way, the ability to tolerate fuzzy source data or the need for crystal clear intact relational data.  As we are always striving for speed and outage reduction at BMC my personal favorites are:

 

      • Change Manager with Resource Maximizer – build a worklist to read production image copies with our Recover Plus product, go through ID translation and then lay those down in your PTE subsystem.
        • Benefit: Automates the build and maintenance of the data migration process, thus reducing the overhead on DBA time and runs at Recover Plus page recovery speed.
      • Online Consistent Copy on the source, followed by BMC RECOVER PLUS using OCC input with OBIDXLAT on the target
        • Benefit: No outage on the source system to create the OCC.  There is little CPU or I/O to create the OCC (it is driven by storage data set snap technology). OCC renders the OCC consistent with log apply technology. The OCC is then input to the BMC RECOVER PLUS OBIDXLAT utility on the target, where replicated data sets will be created.  Consideration: OCC requires the appropriate storage technology

 

You might have noticed in your shop, a recover product can be difficult to justify, but one that runs every day (hint, hint -- we are also going to use recovery in blog entry #2) and saves resources, should be an easy justification.  Check back next week when we go into restoring/recovering the data so you can re-run your tests.  Would turning many hours into minutes get your LOBs and managers attention? Remember those people are waiting to run the next test.

 

Change Manager allows for unattended operations when used to manage the objects in the PTE.  All of Change Manager Processes can run unattended in batch.  You can easily do the following.  Let’s say the PTE looks different than production and you want it that way, but you want to the page moving capability of Recover Plus versus doing unload/reloads.  You can take a baseline (a snapshot) of the PTE structures, drop them, copy the objects with Change Manager from production and then run a Change Manager compare, to the original baseline, to put the PTE back to the original structures, but with production data.  All of this can run automatically and unattended.

 

Please contact myself, your local Software Consultant/Account Manager for more information on any of these great techniques for populating your target environment.

Filter Blog

By date:
By tag: