On Tue, 19 Oct 2010 20:30:47 +0200, Konrado Z <konradoz@xxxxxxxxxxxxxxxx> wrote: > Thanks for your response and explanation. > > OK, and what's about situation where there are 2 parents who are > siblings at the sime time, and one proxy server under them. What I > mean is: > > > P1 ----- P2 > \ / > \ / > P3 > | > host > > and P1, and P2 are connected to the Internet. Same as scenario (3) earlier. A compound of scenarios (1) and (2). In scenario (3) DNS load-balancing determined which proxy the host connected to. In the above (scenario 5?) squid.conf cache_peer entries do the load-balancing instead to determine which P1/P2 the P3 connects to. Amos > > 2010/10/19 Amos Jeffries <squid3@xxxxxxxxxxxxx>: >> On Mon, 18 Oct 2010 20:50:02 +0200, Konrado Z <konradoz@xxxxxxxxxxxxxxxx> >> wrote: >>> Hello, >>> >>> I have to consider four scenarios: >>> >>> 1) http://img230.imageshack.us/img230/4172/52435275.jpg >>> 2) http://img2.imageshack.us/img2/6708/42467335.jpg >>> 3) http://img213.imageshack.us/img213/3600/77675899.jpg >>> 4) http://img84.imageshack.us/img84/7221/12837089.jpg >>> Host is connected only to one proxy (I think it does not matter to >>> which one). But proxies servers are connetected to each other. >>> >>> The question is, on which proxy server the website requested will be >>> cached (and why)? >>> Unfortunately I do not have RAM memory enough to install two (or three >>> in some cases) virtual machines with Squid, so I would much appreciate >>> your answers. >> >> This looks a lot like undergrad homework :P >> >> Notice that cache storage can be suppressed at any Squid on a wide range >> of criteria. So the answer to your question can change dependent on what >> you want to configure. The answers below are using the assumption that >> all >> proxies are storing everything cacheable which travels through them. >> >>> >>> My answers: >>> 1) >>> when 1&2 doesn't have website requested: >>> - 1 & 2 cached the website >> >> Yes. Everything going through #2 also goes through #1. >> >>> >>> 2) >>> when 1&2 doesn't have website requested: >>> - 1 or 2 /it depends where the host is connected/ requests the website >>> and cached it >>> >>> when 1 has the website requested, but host is connected to 2: >>> - 2 requests website from 1, and cached /two servers cached after all/ >>> >>> when 2 has the website requested, but host is connected to 1: >>> - 1 requests website from 2, and cached /two servers cached after all/ >>> >> >> Yes. >> >> With condition that special request types such as CONNECT (and for some >> older squid any dynamic web pages) are not passed to siblings when direct >> Internet access is available. These are usually non-cachable anyway so >> have >> little impact on the storage. They require admin config intervention to >> use >> the sibling. >> >>> 3) >>> when 1&2&3 doesn't have website requested: >>> - the website is cached 3 and 1 or 3 and 2 /it depends where the host >>> is connected/ >>> >>> when only 3 has the website >>> - the website is cached 3 and 1 or 3 and 2 /it is copied from 3/ >>> >>> when 1 or 2 has the website >>> - the website is cached only 1 or only 2 or 1 and 2 or 2 and 1 /it >>> depends where the host is connected/ >> >> No. Simple join of scenarios (1) and (2). >> >> - proxy #3 caches everything. Because all traffic flows through it. >> >> - proxy #1 and #2 are exactly the same as scenario (2) with same >> reasons. >> >> The fact that proxy #1 and #2 are connected to #3 instead of direct to >> internet is not relevant to what gets cached. >> >>> >>> 4) the same situation as in 2 >> >> Yes. >> >> Scenario (1) and scenario (2) are the basic units of cache hierarchy >> design. Same as with Internet route design. The complexity comes when you >> start limiting what requests can go where. Creating big loops, one-way >> requests, non-caching, and peering routes. >> >> Amos >>
