I have noted KeepAlive coming up in a few of the Serf threads. I have been looking at setting up a load-balanced Subversion cluster, somewhat like what Blair describes in this wiki:
http://www.orcaware.com/svn/wiki/Server_performance_tuning_for_Linux_and_Unix Anyway, in looking at load balancers, I noted this in the documentation for HAProxy: http://haproxy.1wt.eu/ Keep-alive was invented to reduce CPU usage on servers when CPUs were 100 times slower. But what is not said is that persistent connections consume a lot of memory while not being usable by anybody except the client who openned them. Today in 2009, CPUs are very cheap and memory is still limited to a few gigabytes by the architecture or the price. If a site needs keep-alive, there is a real problem. Highly loaded sites often disable keep-alive to support the maximum number of simultaneous clients. The real downside of not having keep-alive is a slightly increased latency to fetch objects. Browsers double the number of concurrent connections on non-keepalive sites to compensate for this. With version 1.4, keep-alive with the client was introduced. It resulted in lower access times to load pages composed of many objects, without the cost of maintaining an idle connection to the server. It is a good trade-off. 1.5 will bring keep-alive to the server, but it will probably make sense only with static servers. I do not really have any comments about this, just thought I would share it. -- Thanks Mark Phippard http://markphip.blogspot.com/
