There’s some options I don’t understand from wiki:
max-client-connections (integer: 1..5000; Default: 600)
Maximal number of connections accepted from clients (any further connections will be rejected)
600 value is sum for all clients? If I have 1000 users behind proxy, then only 600 users will accepted (if each user only access 1 connection)?
max-server-connections (integer: 1..5000; Default: 600)
Maximal number of connections made to servers (any further connections from clients will be put on hold until some server connections will terminate)
same question as above.
In ROS v5, max client connections up to 5000, max server connections up to 5000. Total is 10000.
In ROS v6 max client+server connection cannot exceed 6600. Why?
One user could make 600 connection requests, or 1200 users could be online - 600 of them are reading the content on the pages they downloaded while the other 600 are each using one connection to load new content. This is litterally “how many streams of http data can the proxy send to clients at the same time” - Remember that one client can open more than one connection at once - suppose a web page has lots of graphics on it, many of them coming from many different websites… your browser might open one connection for each unique website where the web page’s materials are coming from…
This is exactly the same answer, except it means how many connections the proxy will use to pull the requested content from the Internet. This number can most likely be lower than the clients section. Suppose 100 users all load the google search page - the first user causes the proxy to go out and grab a copy of today’s Google Doodle - while the next 99 requests would just get the cached copy. This is how a proxy saves bandwidth.
If you’re too aggressive in oversubscribing your server connections, then this will appear to be “slow web surfing” to the users because their client will successfully connect to the proxy, but there won’t be an available “outside line” to pull down new content.