• Post Reply Bookmark Topic Watch Topic
  • New Topic

How can huge hits be handled  RSS feed

 
Anonymous
Ranch Hand
Posts: 18944
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi All,

In one of the interviews, I was asked about a design approach for an enterprise application. The question is like this

Assuming huge hits like 30,000 or 40,000 are trying to access your webapplication, which in turn has access to Database through a Business tier . How do you manage these many hits, without impacting the performance and all these requests should have a more or less same response time. Also How do you handle EJBs in your Business tier. How many Entitybeans do you wish to manage in your container? What performance bottlenecks would you be considering for such kind of scenario? What design patterns would you recommend, in such cases?


I tried to answer this question, saying that ideally 30,000 hits may not be able access the application simultaneously and Container should be able to manage sufficient hits. I know, this answer is a lame one and would not satisfy my interviewer, but I didn't see any other option for this scenario.

Can any one help me in solving this design issue? Thanks in advance.
 
rajesh bala
Ranch Hand
Posts: 66
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Ok..

1. What does the intereviewer really mean by 30,000 or 40,000 hits.
2. Is it for an hour or for a day or for a second. Definitely not for a second for sure..
3. If its for a minute, it translates to 666 hits per second (40000/60).

The system might be able to support 666 hits/persecond provided, the response time is low comparatively. We dont need 666 EJBS for this. System might be able to manage with 10 ejbs also with less response time.

So basically without response time requirement, you can't solve it. Also, there is no way you can gurantee same response time as you increase load. That is basic performance theory. Only for certain load, the response time would tend to the same. After that, response time is inversely proportional to the load on the system.

~Rajesh.B
 
  • Post Reply Bookmark Topic Watch Topic
  • New Topic
Boost this thread!