Throughput

Throughput in general means the rate at which the output is generated.
In performance terms, the most literally used meaning is rate at which the data is sent from the servers back to the users.
However, throughput may also be used to describe transaction throughput ie, number of transactions per sec.

For running tests to compare the earlier baseline performance results to current level of code, it is important to maintain the same transaction throughput and not the data throughput (as data throughput can vary due to changes on the page such as intoduction of new images, different result set in different tests, etc).

To maintain same throughput for different tests, loadrunner and other load test tools provide a the feature called "Pacing". This allows to control the number of transactions executed over a period of time.

Analysis of data throughput graphs between test runs with similar TPS can easily point to issues such as addition of new image/javascript/other non html files, increase in response page result set, issues related to compression, and thus leading to high response times without affecting any other server resource, issues with caching of non-html resources at both web server and the browser, etc.

Dynatrace AJAX - A Must tool to check actual end user performance

Just started to have a look at the Dyantrace AJAX tool which can give a very clear picture of the actual end user performance when he is working on a browser.
It breaks down the request response time on the browser into network time, server response time, embedded image/javascript download times, javascript execution time to render the page, etc.

Even before starting a performance test with any of the load test tools, a simple study of requested pages of the application with Dynatrace tool can provide probably lots of pointers to start tuning the application.

http://ajax.dynatrace.com/pages/