On 02/13/2014 11:54 AM, Dr.x wrote: > hi all , > assume have > 4 workers ( i used worker 4 in squid.conf) > 2 disker ( i used two rock dir ) > 1 coordinator > > assume that google website has been cached on hardisks (disk1 and disk2 > which are represent by diskers process) > > > assume that that userx is pumped to squid and requesting google.com > now , > who will answer the user ?? > the worker ? or the disker ?
The worker. Diskers do not speak HTTP. > well i will answer this question and need make sure that my answer is > correct . > > as i understand , the responsible of answer the request are the workrers not > the disker , the worker will try to solve it by memory , if it failed it > will go to diskers ( responcilbe of disk caching and rock ) and try to > server it from hardisks. Yes. > it will go to one of disker and if not found in one of them it will go to > the 2nd disker and so on. The worker does not exactly iterate diskers in search for a disk cache hit. Instead, the worker uses disk indexes to check whether a disk cache has the requested object. Those indexes are stored in shared memory so the check is relatively fast. If the worker detects a hit, it sends the corresponding read request(s) to the disker responsible for the disk/index where the hit was found. > if not found on memory or disker , squid will try to cache it if its size > less or equal 32 kbit ! Approximately 32 KBytes (not 32 Kbits), provided you are using small rock and the response is cachable based on HTTP rules and other factors. HTH, Alex.
