Hi Narayana
Pls. see below and I assume your questions are wrt GF 2.1.1
On 6/2/10 7:24 AM, Narayana Rallabandi wrote:
> Hi,
>
> I have few questions on GlassFish Clustering. I tried search over net
> and docs but in vain. I need some help in this reg. optimum
> configuration for a cluster with in-memory replication works efficiently?
A lot depends on the Hardware being used (multi-core vs single core,
single threaded versus multithreaded cpu), OS tuning, RAM available,
JVM tuning and heap configuration, JVM bit support - 32 vs 64 bits, the
application characteristics and tuning, how sessions are being used in
the application, the number of concurrent sessions, number of users,
size of sessions, and so on and so forth.
There is no one-size-fits-all optimum configuration for a cluster to
make in-memory replication work efficiently. In general you should take
the above factors into consideration for each instance, and calculate
the heap space needed, then double it for replica sessions since each
instance acts as a replica for another, and add some extra headroom in
the heap configuration for accommodating failover sessions. This is
just the starting point and depending on how complex the application is,
you may have to iteratively run sizing tests to determine optimal
configuration for a particular deployment.
Are you or a customer seeing problems? If so please send us details
including domain.xml, details as indicated above and server logs.
>
> * The number of nodes/box or nodes/cluster ?
>
To some extent, a limitation can come from how many instances can be
handled by a DAS in a domain. That is again dependent partially on the
admin infrastructure code and partially on the deployment environment.
Hard to tell though in terms of number of instances per domain without
details.
> * No. of Instances/node
>
>
> Regards,
> Narayanaa
>
>