Dear Alex, Thank you for your advice. If I want large files to be cached when it fist seen by worker, My config should change to first worker that see large file and cache it else left it over to remaining worker for rock store worker? Below design is that, I don't want cached content to be duplicated among AUFS cache_dir and I want to use the advantage of rock store which can be shared within worker on SMP deployment. squid.conf ------------ ## If large files is seen, will be cached first by process number 1 if ${process_number}=1 cache_dir aufs /cache4/squid/${process_number} 170000 32 256 min-size=31001 max-size=200000000 cache_dir aufs /cache5/squid/${process_number} 170000 32 256 min-size=31001 max-size=200000000 cache_dir aufs /cache6/squid/${process_number} 170000 32 256 min-size=31001 max-size=200000000 cache_dir aufs /cache7/squid/${process_number} 170000 32 256 min-size=31001 max-size=200000000 cache_dir aufs /cache8/squid/${process_number} 170000 32 256 min-size=31001 max-size=200000000 endif ## Cache remaining smal object # Rock Store: SMP Awared cache_dir rock /cache1 70000 max-size=31000 max-swap-rate=300 swap-timeout=300 cache_dir rock /cache2 70000 max-size=31000 max-swap-rate=300 swap-timeout=300 cache_dir rock /cache3 70000 max-size=31000 max-swap-rate=300 swap-timeout=300 ---- Regards, Vantha On Fri, Mar 22, 2013 at 12:13 AM, Alex Rousskov <rousskov@xxxxxxxxxxxxxxxxxxxxxxx> wrote: > On 03/21/2013 02:21 AM, Sokvantha YOUK wrote: > >> ## 2. Handle large object > 32kb < 200MB. The fourth worker handles large file >> if ${process_number}=4 >> cache_dir aufs /cache7/squid/${process_number} 170000 16 256 min-size=31001 max-size=200000000 >> cache_dir aufs /cache8/squid/${process_number} 170000 16 256 min-size=31001 max-size=200000000 >> endif > > Just to make sure you are not expecting a miracle here: The above worker > may cache large files, but ALL workers will receive requests for large > files. Workers 1-3 will always miss on large files. > > On average, 75% of requests for large files will go through workers that > do not cache them and do not see them in the worker4 cache (in practice, > this could be 90% if those workers happen to be busier due to OS > scheduling bias!). > > If that is what you are after, the above config is correct. Otherwise, > you need to redesign. > > > HTH, > > Alex. > -- ---- Regards, Vantha