explorer wrote:I'm setting up a machine to do a lot of monitoring, that is, a lot of eValid scripts that all run regularly and in some cases in parallel. What is a good rule of thumb on the maximum number of tests per hour you would recommend?
There are several factors to consider:
(1) You want the scripts to have plenty of wiggle room in terms of playback time so that in case the application you're monitoring runs long the playback of one test won't spill over onto the next instance of the same test.
If it did you might have a kind of cascading failure where a slow test leads to interfererence and that leads to slower tests, until your monitoring machine is fully clogged.
A safety factor of at least 2:1 and maybe even 3:1 is going to be safe.
(2) If you have ANY tests that require use of the desktop you probably want to run those test serially -- not in parallel.
Yes, eValid has the Lock and Unlock commands that give any particular script exclusive access (use of) the desktop, and while these are very reliable they are a bit more difficult to deal with than a script that doesn't have them.
If you reserve one track for you desktop-unsafe scripts, that will cut down on the total number playbacks per hour.
(3) Your CPU and RAM need to large enough to handle the maximum parallelism you plan for. A full load of RAM -- 3 GB -- virtually assures a multiplier of 100:1 -- and we do see higher numbers depending on the script being run.
A good safety factor might be 3:1 here...
So, putting all three factors together suggests a maximum width of about 10-15 tracks and a nominal duty cycle of 33%. As they say, "your mileage will vary," but we have several customers who log nearly 1,000,000 tests per machine per month.
eValid Monitoring Support
by explorer » Mon Jan 25, 2010 5:08 pm