On Wed, 2009-05-20 at 19:28 -0600, singh@xxxxxxxxxxxxxxxxxxx wrote: > This is pretty cool explanation .. thanks so much and absolutely my bad > this particular thing of port 80 should be asked in tomcat user group ;) > > But any thoughts on handling of load for tommcat and apache in this case : > 2500 hits / min. There is no straight forward way of answering this as the number of hits/min is meaningless without an understanding of the work required to achieve a single hit. For example, if each hit is served from memory cache, by fast clients, where each request only takes 0.001s to serve from start to finish, then you would only need a few worker processes, as each worker would be relatively uncontended. If on the other hand, each request takes 3 seconds to complete, maybe due to having to contact a remote WS, go to the DB, whatever, then you would need many more worker processes. In an ideal situation, you would manage the number of workers to allow enough spare workers to handle your load. You can monitor this with most web servers/application servers, in apache you would use mod_server_info. However, it is not that simple (nothing is). By increasing the number of workers, you increase the potential load on your server. You also need to monitor your load, and not allow your server to spawn too many workers. You can monitor this in many ways, depending on your server OS. Finally, you have to take into account how scalable the resources you are using are. For instance, if you use a database, it needs to be able to cope with handling simultaneous queries from the maximum number of workers your app/server is configured to use. Even if you are doing something relatively 'minor' (say, load a template from disk and transform it with some info from the request), you need to check that you can load that many templates from disk simultaneously. Obviously, in that last example you could obviate the need to load from disk by caching the template. If your app feels 'slow' after increasing the number of workers available to it, or has a measurable increase in processing time, then it is a fair indication that your resources are not scaling effectively. This advice is all generic web application tuning, and applies to any web or app server. Specifically for apache though, in the prefork MPM, you control the initial number of workers with ``StartServers'', the maximum number with ``MaxClients'' and you control how it ramps between the two with ``MinSpareServers'' and ``MaxSpareServers''. For other MPMs, those directives have slightly different meanings, so always check the manual for both the directives and the MPM you are using. Cheers Tom --------------------------------------------------------------------- The official User-To-User support forum of the Apache HTTP Server Project. See <URL:http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx " from the digest: users-digest-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|