QoS - Availability (part 2)?

[Trond Hjelmaas]

Hi,

I should be done with the architecture soon, but I am still not sure how to address some of the Qulity of Service (QoS) requirements...:

Availability - FBN has bought a Sun E10000 as the app server. Availability has a few keywords like fault tolerance and redudance, but we only have one machine... I guess it is possible to run multiple JVM on the E10000. One machine is still a "single point of failure" if for example the powersupply brakes down...

Reliability - ability to assure consistency and integrity. I would think transactions is a keyword here.

Any suggestions for the thoughts above?

Thanks is advance, Trond

[Akar Rafidj]

Trond,

For your first question: http://www.sun.com/servers/highend/e10000/specs.xml
For reliability, you should take in mind also:
-robustness:strong test plan and exception handling mechanism...
-resources management: finding, acquiring, using and releasing resources.I'll recommand studying Pattern-Oriented Software Architecture: Patterns for Resource Management, Volume 3 by Michael Kircher and Prashant Jain ISBN:0470845252
John Wiley & Sons � 2004
Hope this helps

Akar
[ October 17, 2005: Message edited by: Akar Rafidj ]

[Trond Hjelmaas]

I looked at the E10000 description in more detail and it looks "unbreakable"

BUT, I am still not sure about the multiple JVMs and/or active/passiv replication. Any more comments ?

Thanx!

Originally posted by Akar Rafidj:
Trond,

For your first question: http://www.sun.com/servers/highend/e10000/specs.xml

[ October 17, 2005: Message edited by: Akar Rafidj ]

[ October 17, 2005: Message edited by: Trond Hjelmaas ]

[Akar Rafidj]

I don't think that you have to go at this level of details.There're many clustering strategies for scalability and availability:
Clustered pairs
Two-node clusters
Ring
Scalable (N-to-N)
N+1 (Star)
Usually, you select your clustering strategy after load tests. This strategy depends also of your architecture(local vs distributed) and your hardware. (one or more servers).
For a local architecture and one server,one scenario may be: 3 J2EE server instances at 60% max capacity (weighted load balancing strategy), each running in its server domain, le first and the second for scalability, the third (active or in stand by) for availability (star strategy)

Akar
[ October 17, 2005: Message edited by: Akar Rafidj ]

[Trond Hjelmaas]

spot on. Thank you!