I had my usual "several" (probably 3 to 5) Firefox windows/tabs open ... and Microsoft decided to send an update to my Windows 7 box. I observed Firefox memory use rise r…
I had my usual "several" (probably 3 to 5) Firefox windows/tabs open ... and Microsoft decided to send an update to my Windows 7 box.
I observed Firefox memory use rise rapidly to nearly 800MB ... while netsvcs.exe did approximately the same thing. (This machine has 2GB of RAM.)
Fortunately I had Windows Performance Monitor running, and was able to determine that Windows Update was the competing activity ... and after-the-fact, determined that it had been installing an update to MS Security Essentials (basically a malware signature update - and I have the ID of the specific update, but since it is a security update, I'm hesitant to post that detail in a quasi-public forum). Usually, such updates install quickly and transparently. I also cannot tell, with confidence, whether the memory consumption was from the process of installing the signature - or the process of scanning for malware presence after the sig was installed (although I'd *guess* the latter). I believe both of those aspects execute within netsvcs.exe.
This happened over a period of approximately an hour, during which about all I could do was "sniff around" to identify the culprits; eventually I killed the Firefox process (hoping that its History feature would allow me to return to my open tabs... which was not quite true).
I have reported the event to Microsoft as well, as it could equally be their problem.
Suggestion: Add instrumentation to Firefox to capture data on sudden increases in RAM use - paying particular attention to outside-the-norm values, in relatively short time periods (minutes, not hours) that do not self-resolve ... and consider whether Firefox can be more aggressive at releasing memory it has "claimed," when things go bad.
(Perhaps collect high-level "physical memory available" data when things begin to "go bad," and capture the "footprint" of Firefox memory use... And perhaps add intelligence to log memory allocation failures somewhere. I don't know with certainty if the issue was fragmentation, or the sheer amount requested, but experience and grey-hair lead me to suspect the former. The only Firefox window/tab that I know might have been a high-memory-consumer was a moderate-sized Google Map covering an urban area of several dozen square miles, so it was a complex image.)
Not suggesting this is Firefox's "fault" -- clearly, both processes were "to blame" in the sense that they didn't handle the situation well... but Firefox might handle it better.
I've seen (and reported) quite a series of memory-utilization issues since 28.0 ... perhaps it's time for a "tiger team" to look into these.
AH - one other detail: I keep Firefox pinned to the Windows task bar, and when it starts, it asks me which profile I want to use (of about 5 choices). When restarting Firefox after identifying all the above... I found that it no longer asked me to choose a profile, and the profile it opened was clearly NOT my most-common profile, which includes (e.g.) all my most-common, site-specific login information. I had to mess around for several minutes to get this profile "back" up and running, and I did not make note of what I tried during that effort.