I am working for a client who is moving from a legacy system to J2EE web based system. But they have ridiculous requirements for response times. Where the old system had 16 seconds, the client made it 2 in reqs.
Is there a document/standard anywhere which gives average response times based on server, no of users and records involved which i can use to support my team?
btw, i am using weblogic 8, oracle 8.1.7, 100 max users with about 10,000 records involved for any query.
Please give me details reagrding J2EE Architecture u r using ...
is us using XML/XSL or directly using JSP's Plz give me details...
Is client is requesting.. response time in 2 Sec.?
What type of application is it?
As per standards maximum 8 seconds is response time expected for any web page ... except that if page is having any bathc processing at back end or middle layer ...then u have to clearly tell customer about the same ...
Quality is never an accident, it is always result of intelligent efforts.
The j2ee application is usng struts, and EJBs.
app server is weblogic 8.1. No xml/xsl. presentation layer is JSPs. its an intranet application.
You said "As per standards.." thats exaactly what i want.. where do i get those standards? i need to present them to the client..
2 seconds is ridiculous. This requirement will not pass whatever you do.
About standards: There is really no standard because it all depends. I can have users extremely pleased with a response time of 1 minutes and I can have them extremely displeased with a response time of 0.1 second.
Andy King wrote a book on Web Optimization and in the first two chapters he addresses response times quite well. The book is worth the price for just the first 2 chapters.
Check it out at http://www.websiteoptimization.com/
Response times are a completely subjective measurements. It is really up to your client and what they feel is acceptable. Your only concern is conveying to them the realistic situation and what is technically feasible at a given cost. Assuming you have a properly designed and tuned system, achieving very low response from a system under load (linear scalability and a robust system) can come down to cost and what they are willing to spend to achieve that performance.
One factor in making any of these decisions is the nature of the business and what users can be excpected to deal with.
For example, lets look at 2 hypothertical system that could be implemented with J2EE..
System 1) free discussion forum for basket weaving enthusiasts.
The typical user is probably not very concerned with performance. They can be expected to endure longer page load times than a user of a mission critical business system. Designing for page loads of under 8 may not be much of a concern.
System 2) Real-time trading system for institutional equity investors.
This may be a large revenue producing system for a high profile business that engages in trading of stocks worth billions of dollars daily. It could be a cutthroat market where a few seconds added on to a transaction time could mean large swings on pricing. The user might connect to your system over a high speed private network. Even mentioning 8 second response time could cause them to cancel a contract. System outages usually mean people lose jobs. They want mega scalabilty and super fast performance, and will pay top dollar for it.
....... so, those are 2 hypothetical system. My point being that you can not apply a standard of system performance to anything. The key is to know your client, what they need, how important it is to them, and that they understand costs involved in achieving these needs. And with enough money and time, there really are no technical containts.