HP Integrity NonStop H-Series - Manual (Page 20)

◦ Loading the session data from Infinispan Cache server prevents old session data from
being used.
◦ Failure of one or more NSASJ serverclass instances or Infinispan Cache server does not
impact the web session replication as NSASJ Server instances are configured in the
TS/MP serverclass and the Infinispan Cache servers work in a distributed mode.
•
Stateful session replication handling: State of a Stateful session bean is maintained by Infinispan
Cache server across multiple client requests. Hence, failure of any number of NSASJ instances
is tolerated across multiple client requests.
NOTE: Failure of Infinispan Cache server during client request processing is tolerated as
Infinispan Cache server is backed up by another node and both these nodes are configured
to work in distributed mode. In case of an NSASJ instance failure during processing a client
request “Remoting connection can no longer process messages” is returned to the client. In
such cases, the client needs a restart.
Scalability
Scalability is the capability of NSASJ to increase the total throughput in an environment of increased
load when new hardware or software components are added.
In other words, scalability is the ability of NSASJ to achieve the following:
•
Take complete advantage of all the available resources to provide optimal performance.
•
Take complete advantage of any newly available resource (hardware or software) and provide
a performance that is proportionally higher, prior to adding the new resource.
The subsequent sections discuss the various NSASJ components that can scale when new resources
(hardware or software) become available.
Post Master for Scalability
The main purpose of the Post Master is to listen on the remoting port and transfer all data to the
NSASJ serverclass. Post Master does not participate in data processing. It just acts as an agent
that transfers data between the remoting client and NSASJ serverclass.
The number of instances of Post Master that can be started is dependent on the TCP/IP provider
that is used. Post Master is a state less component, and the scalability of Post Master is determined
by the number of instances of Post Master that can be made functional at runtime.
The two different types of TCP/IP providers facilitate Post Master to scale. When using TCPIPv6,
each CPU can have only one instance of Post Master. As a result, the ability of NSASJ to handle
connections is limited by the number of instances that can be running on a CPU which is typically
one. A new instance of Post Master can be started for every new CPU that is added. This new
instance becomes functional immediately and is used to serve requests. Since, only one instance
of Post Master can be started per CPU, it puts a restriction on Post Master to fully utilize the available
resources in a CPU.
Whereas, when using IP-CIP, each CPU can have any number of Post Master instances. New
instances of Post Master can be started for every new CPU that is added. Since, many instances
of Post Master can be started per CPU, Post Master can fully utilize the available resources in a
CPU.
20 Features of NSASJ