I've been looking through the posts and documentation on Delay Pools. I suspect I've missed a few important examples and notes. I know very similar questions have been asked before. The following scenarios seem to be the most common uses - but I'm still not sure how I would implement them. All the questions are either how do I distribute my bandwidth among groups of machines, or how do I limit the amount of bandwidth for any destination, web site, protocol? What I really need are examples with the information to calculate their effect. [I started running webalizer against my current squid over the last few months - and while I doubt I have webalizer configured correctly my monthly throughput is exceeding what I can handle)
Feel free to RTFM me, but include links or at least search terms so I can gather up the answers I missed from the archives. (I tend to use the one at http://marc.theaimsgroup.com/?l=squid-users&r=1&w=2 so if you could check a search or link on that archive it would be most helpful, though I'm sure there are other places as well.) Lets say there are three groups on three subnets: Research on 192.168.2.0/24 Marketing on 192.168.3.0/24 Administration on 192.168.1.0/24 Scenario One: insuring an even distribution of available bandwidth. Let's say I have a 1Mbps interent connection and I know I need 100kbps for external connection, email, web server,etc on my network. That leaves 300kbps for each of the tree groups.Can I enforce that with delay pools. Scenario Two: insuring a hierarchical preference so the following types of documents are treated as more important: 1. Text, html, pdf, 2. graphics. 3. audio files 4. video files 5. anything being streamed. Nothing should be prevented per se, but the simpler and older basic web services should always have a good response, while streaming video and any bandwidth hogs I know about should be limited. Scenario three: Keeping available bandwidth proportionate. Say Marketing sub net has 10 machines. First, how can I make sure that the same percent of the Marketing groups bandwidth is available to each machine. Second, is there a way to keep it proportionate but not count machines when they aren't being used. Scenario four: Limit bandwidth based on the online service. For example say you know a lot of people will be using Aol Instant Messenger, which is fine from time to time. But as it is a text based protocol you know that should never really take any bandwidth and you know you can restrain it quite a bit before you render it unusable. How would you do that. -- Josh Kuperman [EMAIL PROTECTED]
