Weirds Transaction Per Second (TPS) Value

[tevfik]

Hi,

We testted two scenario :

1. Client creates request directly one application server.

2. Client created request using CSS 11503 VIP address

Results from client's side report:

------------------

Using Directly server IP Address:

TPS: 128,639

Response Time:1,113

178 Error

Using Load CSS11503:

TPS: 91,950

Response time:1,614

0 ERROR

------------------

As you see, using one server we can get bigger TPS. But, when we use Load balancer-there are six server load balanced as round robin- I think we must have much more TPS than single server TPS when we use Load Balncer.

So, I think there is a problem, can anybody give an idea about this problem?

[Gilles Dufour]

This is very well possible.

The css needs to do is own lookup before forwarding to the server, so it will for sure introduce a delay which will impact the total response time.

However, if you have 5000 clients doing the same test at the same time, I can guarantee the CSS TPS will probably still be what you see now but your server will die.

That's where the css makes the difference.

Gilles.