eValid™ -- Automated Web Quality Solution Browser-Based, Client-Side, Functional Testing & Validation, Load & Performance Tuning, Page Timing, Website Analysis, and Rich Internet Application Monitoring.

Overview
Good engineering practice is needed to assure the maximum effective value from eValid's LoadTest feature on PCs running Windows NT/2000. Long-running, deep ("coherent") tests often download large amounts of data and consume relatively large resources from the PC. Machine features such as CPU speed, available RAM, I/O speed, motherboard speed, video-card speed and on-board RAM, and a host of other factors affect how much load generation performance you can achieve.

Recommendations
Here are several recommendations on how to maximize eValid performance.

1. Use eVlite
   eVlite generates "background activity", albeit not the full transactions which eValid can handle.

   The idea here is to keep the WebSite servers busy so you can experiment with response times from multiple eValid playbacks being served by a [very] busy server.

   Each eVlite takes about 8 MBytes of RAM, and can play back up to 1000 threads, each representing a separate user session. eVlite is especially helpful to generate background activity when a long, deep transaction script is being used in the foreground with full-eValid playbacks.

2. Carefully Budget RAM Use
   Each eValid takes about 9 MBytes of RAM and each eVlite takes about 8 MBytes of RAM. If you use the System Summary information on your machine you can read out the "Available Physical Memory" value. This is the value you will want to budget your eValid/eVlite usages against.

   Adding additional memory is relatively inexpensive and generally getting cheaper. Recently we saw 512 MByte chips at about $80.

3. Adjust Virtual Memory Settings
   You probably will need to have about 900-950 MBytes of Virtual Memory (on a machine with 512 MBytes of RAM) if you plan to run a large number of eValid's in parallel. Virtual Memory size is resettable by the user.

4. Adjust HeapSpace Settings
   You probably will have to set your HeapSpace setting to the maximum of either 16,384 or 32,768. You may need to experiment to find the optimum setting for best results from your machine.

5. Increase The Available RAM in Video Card
   As you know, when you have the minimal amount of available RAM in the video card, i.e. 4 MBytes, strange results in the display happen with high eValid parallelism counts. Interestingly, it appears from what we observe that the technical results themselves do turn out to be correct. It is only the display you see that is goofy.

   We get most of our data from a 512 MByte machine that has a hi-performance video card which has 64 MBytes of RAM. Attached are two printouts from the web that describe Hi-performance video cards with 64 MBytes of RAM are widely available at about $250-$350.

6. Include Delays To Decrease Effective Playback Duty Cycle
   The duty cycle on eValid goes down if you have a higher percentage of wait times and accordingly a larger number of eValid's can run simultaneously if they are not all running "as fast as possible."

   We have observed very wide variations between scripts with a 100% duty cycle and those with a 10% duty cycle, for example. You should expect to see the same kind of variance.

7. Calculate and Budget Machine Bandwidth Carefully
   A test that downloads 100 KBytes in 10 seconds implies a bandwidth of 10 KBytes/sec or 80 Kbps. Ten such tests would require 800 Kbps to run without slowdown, other factors being equal. eValid's log files tell you precisely the total download bytecount for the test just completed.

   Typical PCs have effective and actual I/O bandwidth that varies between 2000 Kbps and 4000 Kbps (2 Mbps and 4 Mbps). This number varies depending on the speed of the mother-board, the LAN that it is attached to, etc.

8. Use Multiple Playback Machines
   The best solution to achieving large capacity tests is to use multiple machines. Special short-term eValid license keys for multiple machines are often made available to deal with such crunch times.