This section discusses the amount of data passed between the components of the ITS when it processes Web applications.

Statistics are given for typical WebTransactions delivered by SAP, for example the WebTransaction Online Store. We selected a few items from a short product list and placed an order. No pictures were associated with the products. Other ITS technologies (WebReporting and WebRFC) are not covered here because the amount of data depends entirely on the application data or the way in which a self-written WebRFC application is programmed.

The network load can be computed directly from the amount of data in each dialog step, multiplied by the number of users, multiplied by the number of dialog steps in each time interval.

The network traffic is usually dominated by images embedded into the HTML pages. There are two kinds of images: General images that are used for buttons and similar features, and application images, such as pictures of products. Our measurements can only include the general images, but no application images, because we cannot make an assumption about their size.

It is generally advisable to have network load considerations in mind when you design a Web application. Otherwise you may irritate users who have slower network connections.

In our test application, each dialog step (mouse click) resulted in an average of 4 requests to the Web server. This includes images used for buttons. Some dialog steps update more than one browser frame and so request more than one HTML page to be loaded. Each dialog step created average network traffic of about 23 KB.

The amount of data in each dialog step depends on the SAP transaction that is called by the ITS and also strongly on application data. This dependency on the application is at first surprising to anybody who is familiar with SAPgui network traffic, because the SAPgui exhibits almost no such dependency. This is because the SAPgui only receives the screen elements that are actually displayed on the screen. Any scrolling results in a new communication step. In contrast, the ITS has to fetch the entire list from the SAP System and send it to the browser for local scrolling. This results in more data for each screen, but fewer screens that have to be transmitted.

Because the WGate itself processes almost no data, the network traffic between the WGate and the AGate consists mainly of HTTP requests that are produced by the browser and the HTML pages generated by the AGate. The protocol overhead is minimal. Pictures are not sent over this connection but loaded directly from the Web server. The same applies to static pages, embedded Java applets and so on.

The size of the HTTP requests is usually small (a few hundred bytes). This means that the total load is dominated by the size of the HTML pages (usually a few KB). It also depends on the SAP transaction and on application data. The number of requests to the AGate is exactly the same as the number of requests to the WGate URL on the Web server.

In our test application, each dialog step resulted in an average of 3 requests from the WGate to the AGate and created an average network traffic of about 16 KB.

Using the built in DES encryption algorithm results in an overhead of about 10 percent.

For WebTransactions, the network traffic between AGate and SAP system is just normal SAPgui traffic. However the amount of traffic in each dialog step is larger than for normal SAPgui users, since current versions of ITS do not support compression of the SAPgui traffic. So network traffic between the AGate and SAP is about two times higher than that between the SAPgui and SAP for the same transaction.

The amount of data transmitted between the AGate and the SAP System again depends on the SAP transaction, as well as on application data. Note that multiple frames (each of which result in one Web request) do not result in multiple requests to the SAP System because they are all contained on one SAP screen. So the number of requests to the SAP system may actually be lower than the number of Web requests to the WGate.