Explanation of “Off-Box Inclusive testing” and how it impacts troubleshooting


An “Off-Box” request is any request that is made by a client to a machine that is not part of the core environment. As an example, if a user opens Internet Explorer and goes to the main Hanes.com website, Internet Explorer will go to the Hanes machines (the core environment). However the browser will also go to a few 3rd party sites on the Internet to retrieve images and possibly data. These “Off-Box” calls are made by Internet Explorer because the core environment told IE to go to this place to get the image or data.


Once the core environment tells IE where it should get the image or data, the core environment is no longer involved in any way with completing the final page in IE. The fact that these calls are made to external sources, which are shared by applications from other companies means we have no way of directly influencing, or even knowing what load or issues they might have that can change our numbers.


As an example, here is a simple test to the <Contoso> default web site. The request was made five times in a row and timing was measured from each one.

The results show:

·         The performance improvement between runs 3 and 4 was:

o   Total with OFF-BOX: 2.9 times longer (289%)

o   Total of ON-BOX only: 6.4 times longer (642%)

·         The performance improvement between runs 4 and 5 was:

o   Total with OFF-BOX: 1.6 times longer (157%)

o   Total ON-BOX only: 4 times faster (26%)

So test run 4 would give us a *false sense* that degradation was not nearly as bad as it really was.  Likewise, test run 5 makes us believe that our machine was slower than in test run 4, when in fact the machine itself was 4 times faster.


To see the rest of the post and to get some sample code, please visit this entry on Geoff's blog site.

Comments (0)

Skip to main content