Search squid archive

Re: Same cache_dir for many squid process

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 1/06/2013 2:59 p.m., csn233 wrote:
     if you correctly configure your parents to not cache anything, then
you'll have no duplications. In that case, only the squid you'll choose will
manage cache on disk.

     it can be as simple as a

cache_dir null

     to the parent squids.
If the parents have cache_dir null, those parents would simply be
relaying the requests to the origin servers.

No. Critical detail: there are no siblings in a chain.

That hierarchy page has two images at the top for a reason. To clarify that there are two distinct relationship types of parent-child (chaining) and sibling-sibling (parallel scaling). The properties and behaviour of each relationsup and distinctly different. And like I wrote in that top section, they can be combined into many different layouts to perform many very different behaviours.

Leonardo has been talking about a "correctly configured chaining". Using sibling relationship in a chain is *not* correctly configured, the existence of sibling relationship converts a chain hierarchy into some other type of hierarchy. A chain is simply a sequence (chain) of at least 2 proxies like so: client<->A<->B<->C<->D<->origin with so side branches or bypasses.


  They do not add any value
- the child cache(s) can  simply send the requests direct. If fact,
you are adding overheads, the same 1 single request has to go through
the child and then the parent, ie 2 separate processes to process the
same 1 request.

CARP is the type of hierarchy you want for scalability if SMP is unavailable. It is two-layer and each layer is independently scalable. The gateway layer (childs) does not cache, but simply does access controls on clients and determines which cache-layer (parents) proxies is storing the URL being fetched to pass it through on the way to the origin.

Note that CARP typically has no sibling relationships. Yet provides cache de-duplication, failover, and scalability for high-performance beyond what a single or several parallel Squid caches can offer. It does require specific hardware and configuration arrangements however.

With SMP, if you have 2 processes, you can process 2 separate requests
simultaneously.

Amos




[Index of Archives]     [Linux Audio Users]     [Samba]     [Big List of Linux Books]     [Linux USB]     [Yosemite News]

  Powered by Linux