developers have given it. The behaviour we’ll be seeing is also limited to searchSsearch. Effective E
Effective s
i Control Effective ssearchssearcht Study e Offers & Control 82searchs Game ear Effective h Effectivestudyformula s Study ar
h Control s Men a Study c Study - Gsearchm Flostat Formula searchosearchm Study l
Formula #search21 Formula g
rsearcha Formula e Study c Control l Formula et Formula o Game ,searchwsearchi Men h
a Szh m Effective Study o Game o
tsearchmisearche Study thogsearchpusearch Study vsearchr Study r
ssearchossearchve Offers ess
The Problem
The demo site (downloadable from the top of this article) is made up of one page with three buttons; the first button uses large chunks of memory, the second frees that memory, and the third performs an operation which normally takes around 1 millisecond to complete. When any of the buttons are pushed, the time of day and time to run are displayed. To make it easier to reproduce a problem, I am using a 32-bit process, so we are limited to a mere 2GB of memory (yes, I know I could push it to 3GB, but let’s not complicate matters). In case you are wondering how realistic this demo is, with respect to having one button to immediately allocate loads of memory, in a real life situation there may well be a background operation that suddenly uses up great chunks of memory, or it might be lots of users hitting pages all at once which causes the memory usage to abruptly go up.
This exciting screenshot shows how button 1 usually takes ages to complete and, before it suddenly failed with an OutOfMemory Exception at 12:49:38, the longest button 3 had taken to complete its operation was 21 milliseconds. After this inexplicable failure, we start to see memory pressure, and button 3 can randomly take anything from 0 to 561 milliseconds to complete, so there is clearly a lack of consistency.
Read more: Simple-talk