Re: Worst case swapping.
"Matthew" == Matthew Dillon [EMAIL PROTECTED] writes: Matthew :Believe me, I look at these things. Yes there is a lot Matthew going on and a :lot using memory. I normally have about Matthew 20% to 25% of my Gig of swap :used... meaning that I have Matthew allocated roughly double my RAM in :applications. : :And Matthew when this worst-case happens, memory is full... but the Matthew only :active application is Netscape. : :On my home Matthew machine, the same thing tends to happen. It only has Matthew 128M :and vastly fewer things going on. I see cases were Matthew I'm surfing for :20-30 minutes and I will hit this 10 to Matthew 30 second (longer, becase the :swap at home is slower) Matthew gap in netscape response. : :The only other applications Matthew running would be something like a small :UUCP transfer or Matthew a small amount of NFS traffic when the wife's :(diskless) Matthew machine changes screensavers. : :Dave. This may be a stupid suggestion, so please excuse me if it is... but... have you checked process resource limits? Especially the "memoryuse" parameter, I think, tells the system how large a working set you will allow each process. Regards, Jacob. -- Jacob Lorensen; Mosebuen 33, 1.; DK-2820 Gentofte, Denmark; +45 39560401 PGPid: 0x752EB4DE Fingerprint: F609A0BAFF393EA904F7-F344680F8EED752EB4DE To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Worst case swapping.
I'm running a 700Mhz K7 with 256M of RAM as my workstation. I have two fast SCSI drives with a Gig of swap between them. The system shouldn't normally be a bottleneck as a workstation. I find, however, that there seem to be some bad worst-case senerios popping up rather often. Netscape is a good (common) example, but other memory stresses will show if the system is busy, too. What I'm talking about is a situation where some portion of the application will be swapped out and then when the application becomes active again, the swap will grind heavily reading and writing for 10-20 seconds (pushing 5M/s out and 5M/s in). Now the application in question (Netscape) usually runs around 50 to 75 megs, so that swapping activity is effectively swapping an amount of memory equavalent to the whole application out and then in again. My fear that this is a worst case scenario comes from this fact: that some other part of the application now-just-recently-active-again is being swapped out to bring in a part that was already swapped out. Now, you could argue that this case is hard to avoid, but I find this happening during periods of constant browsing ... such that only a small amount of the application could have been out. I'm positive that its not a case of the working set being larger than physical memory; it's one of choice of page to swap. Has anyone done any thinking about this behaviour? It occurs with varying degree to many applications. Dave. -- |David Gilbert, Velocet Communications. | Two things can only be | |Mail: [EMAIL PROTECTED] | equal if and only if they | |http://www.velocet.net/~dgilbert | are precisely opposite. | =GLO To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Worst case swapping.
"Matthew" == Matthew Dillon [EMAIL PROTECTED] writes: Matthew :Now the application in question (Netscape) usually runs Matthew around 50 to :75 megs, so that swapping activity is Matthew effectively swapping an amount Matthew 50-75MB is a lot, but if you have 256MB of ram it can't Matthew be the cause unless there are other active things eating Matthew similar amounts of ram. Matthew It kinda sounds like a runaway to me. A ps axl during Matthew these heavy paging periods should shed some light on the Matthew problem. Believe me, I look at these things. Yes there is a lot going on and a lot using memory. I normally have about 20% to 25% of my Gig of swap used... meaning that I have allocated roughly double my RAM in applications. And when this worst-case happens, memory is full... but the only active application is Netscape. On my home machine, the same thing tends to happen. It only has 128M and vastly fewer things going on. I see cases were I'm surfing for 20-30 minutes and I will hit this 10 to 30 second (longer, becase the swap at home is slower) gap in netscape response. The only other applications running would be something like a small UUCP transfer or a small amount of NFS traffic when the wife's (diskless) machine changes screensavers. Dave. -- |David Gilbert, Velocet Communications. | Two things can only be | |Mail: [EMAIL PROTECTED] | equal if and only if they | |http://www.velocet.net/~dgilbert | are precisely opposite. | =GLO To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Worst case swapping.
:Believe me, I look at these things. Yes there is a lot going on and a :lot using memory. I normally have about 20% to 25% of my Gig of swap :used... meaning that I have allocated roughly double my RAM in :applications. : :And when this worst-case happens, memory is full... but the only :active application is Netscape. : :On my home machine, the same thing tends to happen. It only has 128M :and vastly fewer things going on. I see cases were I'm surfing for :20-30 minutes and I will hit this 10 to 30 second (longer, becase the :swap at home is slower) gap in netscape response. : :The only other applications running would be something like a small :UUCP transfer or a small amount of NFS traffic when the wife's :(diskless) machine changes screensavers. : :Dave. Hmm. How large a memory-cache do you have configured for netscape? Disk cache? What is the RSS and VSZ of the netscape binary while the paging is going on? Please post a ps axl of the state of the system while the paging is going on. -Matt Matthew Dillon [EMAIL PROTECTED] To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Worst case swapping.
Matthew Dillon wrote: :Believe me, I look at these things. Yes there is a lot going on and a :lot using memory. I normally have about 20% to 25% of my Gig of swap :used... meaning that I have allocated roughly double my RAM in :applications. : :And when this worst-case happens, memory is full... but the only :active application is Netscape. : :On my home machine, the same thing tends to happen. It only has 128M :and vastly fewer things going on. I see cases were I'm surfing for :20-30 minutes and I will hit this 10 to 30 second (longer, becase the :swap at home is slower) gap in netscape response. : :The only other applications running would be something like a small :UUCP transfer or a small amount of NFS traffic when the wife's :(diskless) machine changes screensavers. : :Dave. Hmm. How large a memory-cache do you have configured for netscape? Disk cache? What is the RSS and VSZ of the netscape binary while the paging is going on? Please post a ps axl of the state of the system while the paging is going on. Netscape reallys goes to pot in a hurry if you allow it to use more than 1-2MB of memory cache. A friend was seeing a terrible response and tracked it back to Netscape's memory cache. He had a lot of memory and started out with something on the order of 16MB. By the time he was satisfied he was allowing less than ~2MB of memory cache, which is all I had ever allowed it to use. I seem to remember an affect on how much disk cache but that part of the memory has evaporated. Kent -- Kent Stewart Richland, WA mailto:[EMAIL PROTECTED] http://www.3-cities.com/~kstewart/index.html FreeBSD News http://daily.daemonnews.org/ SETI(Search for Extraterrestrial Intelligence) @ HOME http://setiathome.ssl.berkeley.edu/ Hunting Archibald Stewart, b 1802 in Ballymena, Antrim Co., NIR http://www.3-cities.com/~kstewart/genealogy/archibald_stewart.html To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Worst case swapping.
Kent Stewart drunkenly mumbled... Netscape reallys goes to pot in a hurry if you allow it to use more than 1-2MB of memory cache. A friend was seeing a terrible response and tracked it back to Netscape's memory cache. He had a lot of memory and started out with something on the order of 16MB. By the time he was satisfied he was allowing less than ~2MB of memory cache, which is all I had ever allowed it to use. i never screwed with the memory cache, but i've seen some pretty heavy memory leakage with navigator. how long have you had netscape running? an hour, a day, a week? my experience (it seems to have gotten better lately since the upgrade to 4.0 however) has seen starting size 16M, ending size 172M, running time of 13 days. of course, the fact that netscape actually ran for 13 days without crashing is a bit of a miracle itself. :) cheers, -brian To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
Re: Worst case swapping.
On Sun, 11 Jun 2000, Brian Hechinger wrote: :Kent Stewart drunkenly mumbled... : : Netscape reallys goes to pot in a hurry if you allow it to use more : than 1-2MB of memory cache. A friend was seeing a terrible response : and tracked it back to Netscape's memory cache. He had a lot of memory : and started out with something on the order of 16MB. By the time he : was satisfied he was allowing less than ~2MB of memory cache, which is : all I had ever allowed it to use. : :i never screwed with the memory cache, but i've seen some pretty heavy memory :leakage with navigator. how long have you had netscape running? an hour, a Netscape has lots of memory leaks. the worst seem to be in Javascript, with java being a close second. I find I get the best performance and stability out of it if I leave those off, except when you need them. I also keep the memory cache size small, 2 or 3 megs; I leave the disk cache at a large size, since I am behind a slow link. The FreeBSD buffer caching does a good job of not throwing stuff away that I am actually using, so that works out quite well. To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
