[infinispan-dev] Infinispan Persistence Manager and JBoss Messaging Journal

Jason T. Greene jason.greene at redhat.com
Thu Jul 30 13:48:04 EDT 2009


BUT.... :) If you had a random access file (just integer offsets is all 
that is needed), combined with this journal, you could create a simple 
record based transaction db, and that could definitely be mapped. The 
key hash could be used to map to a block that would contain near entries.

Clebert Suconic wrote:
> Well .. ok.. it doesn' t fit on that at the moment.
> 
> On 07/30/2009 12:29 PM, Jason T. Greene wrote:
>> The cache store is/can be a superset of the data in memory. So, for 
>> example, you may have 1G of data, but only 200M of memory. You can 
>> configure the memory part of the cache to be an LRU (data is never 
>> lost, just evicted from memory).
>>
>> Even if the disk storage fits in memory, you likely don't want to load 
>> it right away if you dont have to (i.e. waiting on 2 gigs of disk to 
>> be read in)
>>
>>
>>
>> Clebert Suconic wrote:
>>> On restarts, you can get the whole data back on the method load.
>>>
>>> Eviction is probably a delete. so you probably don' t need to read 
>>> the Store for eviction.
>>>
>>>
>>>
>>>
>>> On 07/30/2009 12:16 PM, Manik Surtani wrote:
>>>>
>>>> On 30 Jul 2009, at 18:12, Clebert Suconic wrote:
>>>>
>>>>>
>>>>>> Yes - the cache store stores a superset of what is in memory. 
>>>>>> Overflow. So it still needs to be accessible in a random fashion, 
>>>>>> using a key.
>>>>>>
>>>>>> See the load() API on the CacheLoader interface.
>>>>>
>>>>>
>>>>> If the data is in memory, why do you need to recover it from the 
>>>>> Store?
>>>>
>>>> Eviction. Restarts. Etc. :)
>>>>
>>>> -- 
>>>> Manik Surtani
>>>> manik at jboss.org
>>>> Lead, Infinispan
>>>> Lead, JBoss Cache
>>>> http://www.infinispan.org
>>>> http://www.jbosscache.org
>>>>
>>>>
>>>>
>>>>
>>>
>>
>>
> 


-- 
Jason T. Greene
JBoss, a division of Red Hat



More information about the infinispan-dev mailing list