(Working with Yuval Meir and Limor Gavish on this project)

IBM has a policy that Open Source products may use it's patent so we
didn't see ARC patent as a barrier.
(PostgreSQL problem was with ARC interfering with closing the source
of PostgreSQL in the future)

We will also try to setup a parameter so that caching algorithm can be
changed at will.

There are other algorithm that provides comparable results.

Nezer
On May 24, 10:12 pm, Yuval M <[email protected]> wrote:
> Hello,
>
> My name is Yuval Meir and I am a M.Sc student in Tel Aviv University.
> As part of my academic research I have implemented the ARC (Adaptive
> Replacement Cache) caching policy in Memcached
> I ran benchmarks using a modified version of memaslap and the results were
> very good: cache misses were reduced by up to to 30% on some loads, and in
> terms of run time, memory usage and CPU usage, my implementation performed
> similar to the original Memcached.
> I attached the diff (based on Memcached 1.4.13) and the results I have so
> far in the excel file (note that it has several tabs).
>
> ARC was introduced in the following publication by Nimrod Megiddo,
> Dharmendra Modha:http://www.almaden.ibm.com/cs/people/dmodha/arcfast.pdf
> which was published in USENIX 2003.
>
> I will be very glad to get you reviewes and comments.
>
> Regards,
>
> Yuval Meir
> Computer Science department
> Tel-Aviv University
>
>  memcached_with_arc.patch
> 61KViewDownload
>
>  ARC_Memcached_Results_2.xls
> 106KViewDownload

Reply via email to