Hi Saikat,
Comments inline
On 08/01/15 12:27, Saikat Maitra wrote:
Hi Andy
Thank you for your email and sharing the external cache implementation. It
certainly helped me to understand some of the usecases better.
I am thinking to implement the caching layer with a provision to use any of
the options below.
Good - I think the key is to get the right abstractions in the code with
different implementations.
1. We can use in-memory operation using Guava library and have regular
schedule for execution of task to persist the data in disk.
2. The second option is to provide capabilities to use couchbase buckets to
store the cache results and have a separate couchbase cluster instance.
Realisically, for most deployments, in-memory is more important.
Especially as the cost-effective size of servers RAM these days seems to
be going up so quickly.
I don't know enough about couchbase buckets -- which features do you see
as most relevant here? As a pratical matter - can they be used
separately from couchbase?
I am still reading the Fuseki2 server code to understand the server work
area.
Andy
Regards
Saikat
On Thu, Jan 8, 2015 at 2:09 AM, Andy Seaborne <[email protected]> wrote:
On 07/01/15 15:43, Saikat Maitra wrote:
Hi
I am Saikat, I am new to Apache Jena and I would like to contribute to
SPARQL Query Caching (https://issues.apache.org/jira/browse/JENA-626).
I would like to know if someone has already taken up the issue and
currently working on it. I have reviewed the old messages in developers
mail list and I suppose this issue is currently still Unassigned.
Regards
Saikat
Hi Saikat,
Good to check - no one seems have mentioned it any time recently.
There is some code:
https://svn.apache.org/repos/asf/jena/Experimental/sparql-
cache/README.markdown
which is an external cache and has some issues:
1/ Updates don't invalidate the cache
2/ It's a rough prototype in scala I knocked together, not production
tested
3/ Claims of "rate control" are overstated (there isn't any).
(seems strange to have it in the cache code at all)
In Fuseki2, there is a server work area on disk so putting results
somewhere on disk now makes more sense. Or memcached.
Andy
