• Post Reply
  • Bookmark Topic Watch Topic
  • New Topic

Techniques to guarantee response times?

 
andy armstrong
Ranch Hand
Posts: 154
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
What techniques can an architect use to guarentee an N second respone time?
Can you configure web logic, jboss, websphere to
return a response page to the user (telling them not to worry .. we are still processing) every N seconds if for some reason the business request is exceeding the maxium request period?
 
Dragan Milic
Ranch Hand
Posts: 31
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
> What techniques can an architect use to guarentee
> an N second respone time?
> Can you configure web logic, jboss, websphere to
> return a response page to the user (telling them
> not to worry .. we are still processing) every N
> seconds if for some reason the business request is
> exceeding the maxium request period?
The architect cannot (at least - should not) tweak the application server. His responsibility is to design a system that is vertically and horisontally scalable - which should help maintaining response time by adding more hardware (clustering of the app server for example). The other posibillity is asynchronous (JMS) data processing which could be used if the user does not need immediate outcome of the data processing and must be used if the expected processing time is not acceptable for the user .
 
Thomas Taeger
Ranch Hand
Posts: 311
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Originally posted by dmilhdef:
The architect cannot (at least - should not) tweak the application server
[/QB]

Hi Dmildhef,
there are situations or business rules when you might want to serve the "most paying" clients within a certain time frame and letting "guests" wait, or in technical data processing etc. .
These situation might be very seldom and unpredictable, like a heavy increase of load after an advertizing campain, or in technical pdp any environmental events are flowding the server.
As soon as the application server SUDDENLY is no longer able to serve all requests in the specified time frame, it is too late to add more servers or other hardware. Scalability can not _guarantee_ availability within economical/effordable limits without a more sophisticated architecture:
The only way I know to handle this case is to design several actor groups having different priorities:
- JMS offers 10 levels of priority (9=highest, 4=default).
- With an array[0..maxPrio] of queue of Command you can implement the decoupling needed for letting the low prio actors wait in business logic if no JMS is used.
In my oppinion to be prepared for such situations is a good architecture if needed to match the business rules.
Thomas.
 
Thomas Taeger
Ranch Hand
Posts: 311
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
P.S.:
Originally posted by Thomas Taeger:
- With an array[0..maxPrio] of queue of Command you can implement ...

... and you could even run over all Commands waiting in low priority queues all 10 minutes and send those clients a "<timestamp> Please wait..." as originally intended.
Thomas.
 
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic