|Preferably it should be a working set, i.e. use the bean a lot then it
|gets more resources. But there is a need for a max, since what would
|happen if 500 EntityBeans were used at the same time without max!?
|Kabooom.
500 EB x 1 kB (big bean!) = 500 kB = .5 M in cache
<snore/>
alright let's even go higher,
let's say that we reach the win2000 soft limit of 2000 clients in // and
that each bean in memory is 10kB (!)
we go up to a wonderful 20 M in cache... we can't generate more load from
the apache loads and the cache size won't go higher.
That being said Sebastien and I have seen the server run into "out of memory
error" because we started the VM with too little memory.
The implementation today is with an LRU that can cycle, only on very extreme
conditions will the bean LRU'ed come back immediately... maybe it is ok, I
don't know
marc
|
|> |Marc, when release 2.1 is planned ?
|>
|> I am shooting for the end of this week.
|>
|> Rickard stuff with EJX needs to make it in, some andy stuff, the roberto
|> work, some deployer stuff and this would be great...
|
|I will get the server conf stuff in, but not EJX. Getting the EJX stuff
|working is doable, but then there's the whole porting of JBoss metadata
|usage... wheee.... I guess we could ship the new EJX *GUI* though, and
|wait with the metadata usage in the server a bit.
|
|/Rickard
|
|--
|Rickard �berg
|
|Email: [EMAIL PROTECTED]
|
|
