> -----Oorspronkelijk bericht----- > Van: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] Namens Pierre van Rooden > Verzonden: donderdag 30 september 2004 10:47 > Aan: [EMAIL PROTECTED] > Onderwerp: Re: Optimalization Project, Step-By-Step > > > Ernst Bunders wrote: > > I can't help noticing that step 13, wich is like 'the other > half' of > > the project has no real planning yet. > > It is possible ofcourse to run this step (like most others that are > finalized at the end) concurrent with the others. > To get this done, we first need to determine the performance and > stability issues (what causes problems, in what situation?). > We can do that based on user input or on using profilers, monitoring, > or other aides. We also need to determine if the same > problems occur on > diferent platforms. > We next need to analyze the cause(s) and determine if and how we will > find a solution. > > The main problem here is that user input is in general not > very accurate > or detailed. Most of the time, when things have gone wrong, > by the time > the problem is reported, what causued it has passed and is often > untraceable. > > We need to make a list of performance issues that are known, > and these > will then get focus in this Step. The bugtracker could be a > good place > to do this - it has a 'Performance' area where these bugs can > be checked in. Wel may also desire to seek for ways to > monitor running MMBase code, or > to seek the code for bad coding samples. Suggestions are welocme. >
I have some idears: 1 I think it would be a good idear to create some kind of 'tracer' module, that keeps track of live processes and is preferably able to show relations between events. This way we can try to get data over a period in different environments to compare 2 allso we must discover the impact of the servletcontainer - database - mmbase setup. Where do problems actually occur? Who knows, maybe chucking out tomcat solves the problem... 3 allso it would be nice to have a way to replay series of request. Maybe this is common stuff and I just don't know about it, but I think about a kind of recorder for complete http requests, cookies and postbodies and so on included, so we could replay crashing situations in different environments. 4 special attention shoud go to the mmbase caching mechanism. I feel something should probably change there. Maybe we need a pluggable system of different caching engines that can be configured to cache specific parts of the cloud. For instance: a forum or other tree-like data structures do not benefit from the current caching, becouse the cached data is invalidated to soon. I'm shure specialized caching could improve performance, and is important as more and more people are building applications in stead of websites with mmbase. ernst > > -- > Pierre van Rooden > Mediapark, C 107 tel. +31 (0)35 6772815 > "Never summon anything bigger than your head." > > >
