Memory Leak Detection

Jan 4, 2011 at 8:40 AM

Hi Clint, I am quite new to this interesting tool and trying to use it to check if there is any memory leak in our software client and after applying the patch whether the memory leaks have been removed a bit.

I have collected the data for two days, each for before and after applying the patch. Currently there is no alert in the report. And how can I interpret the information in the report, especially what does "Hourly Trend" mean?

Condition \Process(*)\Private Bytes Min             Avg            Max            Hourly Trend Std Deviation  10% of Outliers Removed 20% of Outliers Removed 30% of Outliers Removed
OK            Client                                111,771,648 140,197,975 158,146,560 598,955           13,952,506       138,305,391                       136,023,299                       134,124,637                     
OK            Client                                64,102,400   108,776,290 127,082,496 1,074,450        8,271,653         106,914,169                       105,895,621                       105,072,580                     

Thank you very much!

Jan 18, 2011 at 7:48 PM

Hi chanllene,

PAL time slices the counter log into equal sized time ranges. All counter values collected in each time slice is averaged. For each time slice, PAL compares the average value of the time slice to all average values from the previous time slices to produce an increasing or decreasing trend. Finally, I convert it to an hourly trend. So, each time slice has an hourly trend which is based on that time slice and all previous time slices. The Hourly Trend in the overall stats is the the hourly trend of the entire log.

The stats above show me that you have an increasing hourly trend of 584KB and 1,059KB the entire time range of the counter log. If this is over a long period of time, then it means that your second Client app is progressively leaking 1MB per hour. This is a common result because applications will typically use more memory as they are being used.