Integrate the concept of stress in its software testing operations – JDN

If you do not put enough stress to the right places, then your load scenario does not meet its objectives.

Let
 therefore a practical example. Let’s say you test a commerce site
 line. Your load scenario probably will simulate the activities
 the entire site: product pages, search, shop page,
 control procedure, cart and payment. But do you know how much
 distribute the load between these different areas? What proportion of
 scenario must be dedicated product pages with respect to the function
 search?

Say
 you dedicate 10% of the scenario in product research, as it represents
 the average number of users who perform that action on the site. But,
 and if this assumption was wrong and that there was more research
 at the site, and that when the proportion of load research
 15% going to the problems began to appear? Although a
 5% increase seems almost negligible, it corresponds to a
 50% increase of the load on the search the database
 of data. If the back-end database is actually a sensitive link, then
 it can have an impact on the stability of the entire application. But she
 must be subject to sufficient stress to update its weaknesses. For
 say clearly, if you do not put enough stress to the good
 places, then your load scenario does not meet its objectives.

So what
 can you do ? Of course, you can always charge your
 system with more users / stress than under your data
 “Brutes”. But still, you should also try to modify and
 increase / decrease the proportions between the different transactions. Your
 role as a performance engineer is to test even the unexpected and
 identify locations or scenarios where increased stress puts
 your “failed” system.