InfoCenter Home >
7: Multimachine management >
7.1: Using WebSphere Application Server in a multimachine environment >
7.1.2: Availability management
7.1.2: Availability management
One of the benefits of scaling up to a multimachine configuration is that it improves
the availability of the system. Applications hosted on multiple machines generally have
less down time and are able to service client requests more consistently.
The following commonly used scaling techniques can be combined to take advantage of
the best features of each topology and create a highly available system. (Note that this
is not an exhaustive list of ways to improve availability.)
- Eliminate single points of failure in the system by providing hardware and process
redundancy:
- Use horizontal scaling to distribute application servers over multiple physical
machines. If a hardware or process failure occurs, clones are still available to handle
client requests. Web servers and IP sprayers can also benefit from horizontal
scaling.
- Use backup servers for databases, Web servers, IP sprayers, and other important
resources. This ensures that they remain available if hardware or process failure occurs.
- Deploy an application in multiple administrative domains. If an entire domain goes
offline, the others are still available to handle client requests.
- Run administrative servers with workload management enabled. The failover support that
workload management provides eliminates a single administration server as a point of
failure.
- Provide process isolation so that failing servers do not negatively impact the remaining
healthy servers in the configuration. The following configurations provide some degree of
process isolation:
- Deploy the Web server onto a different machine from the application servers. This
ensures that problems with the application servers do not affect the Web server, and vice
versa.
- Use horizontal scaling, which physically segregates application server processes onto
different machines.
- Deploy an application in multiple administrative domains. Problems are confined to one
domain while the other remains available.
- Use load-balancing techniques to make sure that individual servers are not overwhelmed
with client requests. These techniques include the following:
- Use workload management. It is automatically implemented for cloned application servers,
but must be explicitly enabled for administrative servers.
- Use an IP sprayer to distribute requests to the Web servers in the configuration.
- When using remote OSE, direct requests from high traffic URLs to more powerful servers.
You can also make use of OSE's simple load-balancing facility to distribute requests among
servers.
- Provide failover support. The application must continue to process client requests when
servers are stopped or restarted. Ways to provide failover support include the following:
- Use horizontal scaling with workload management to take advantage of its failover
support.
- Use remote OSE to distribute client requests among application servers.
- Use the servlet redirector to distribute client requests among servlet engine clones.
- Enable the Session Manager to store session information in a persistent database. This
preserves session state in case of server failure.
|
|