db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mike Matrigali <mikem_...@sbcglobal.net>
Subject Re: CacheManager create() semantics
Date Wed, 12 Jul 2006 19:29:56 GMT

Gokul Soundararajan wrote:
> Hi all,
> 
> I'm implementing a new replacement algorithm (clock-pro) for the CacheManager.
> In this algorithm, there is a concept of a history item. Basically, after an
> eviction, the algorithm remembers the page id for some time. I was wondering on
> how this affects the semantics of create().
> 
> Just to refresh, in Clock, create() creates a new entry and adds to the cache.
> But, it throws an exception if the cache already contains the item. In my
> implementation, the new item may not exist, exist, or exist as a history item.
> Should I throw an exception if I find it as a history item?
> 
If I am interpreting it right, the "history" is an internal 
datastructure to the cache implementation for help with replacement
decisions.  Given that I would say that create() symantics should not
change, and that the internal implementation of clock-pro somehow
manages the "history" internal without exposing it to the cache
user.

At this point do you have a writeup of the replacement algorithm you
are implementing?

> Thanks,
> 
> Gokul
> 
> 
> 


Mime
View raw message