i think the main point to solve this problem is just seprating 'addressing'
and 'hotness'.

'addressing' i used is quite simple, that is 'hash'.
in this case, to balance 'hotness', i'll cut resources into small pieces,
so that resources could be well-distributed and so as 'hotness'.

further more, in some other cases with meta data for addressing the
resources, i'll also use 'replication' to handle different 'hotness'
requirement.


2014-12-14 16:34 GMT+08:00 atul anand <[email protected]>:

> It is a system design problem .
>
> Suppose a http request  is sent to server . Now Server maintains cache for
> fast retrieval . if link is present int the cache then it just takes a data
> from cache and return it to user but if not , then user will fetch that
> http address and then store it in its cache and return same to the user .
>
> Problem is that there are many server and many global cache as expected in
> distributed system. Now when request is received by a server then how can
> we maintain global cache such that server can know which cache to query
> instead of querying each global cache as it will be inefficient.
>
> one approach can be...... maintain 26 global cache . Now when request is
> received by server it check the web link say , www.*a*bc.com ... here
> server will query cache-1 . Similarly cache-2 will take care of links with
> starts from "b"...www.*b*bc.com ....and so on....
>
> above method will avoid duplicity in caches but will not be very efficient
> as a cache may have higher query rate than others...
>
>
> any other approach ??
>
> --
> You received this message because you are subscribed to the Google Groups
> "Algorithm Geeks" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
>

-- 
You received this message because you are subscribed to the Google Groups 
"Algorithm Geeks" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].

Reply via email to