Browse Prior Art Database

Neural Network Program and Method to Improve Response Time for Web to Host TN3270 and TN5250 Sessions Disclosure Number: IPCOM000131173D
Original Publication Date: 2005-Nov-09
Included in the Prior Art Database: 2005-Nov-09
Document File: 2 page(s) / 22K

Publishing Venue



Disclosed is a method for detecting and programmatically learning to refine the ability to detect the end-of-data flow from an HTTP session.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Neural Network Program and Method to Improve Response Time for Web to Host TN3270 and TN5250 Sessions

Adoption of web to host solutions, is sometimes resisted because of the decrease of end user response time. In 2003, IDC published that the market space for the web to host products, is $245 Million. Heads down 3270 and 5250 emulator client users typically experience 100 ms or less response time, whereas some products that give a web browser interface to the same application, have a response time averaging 1200 ms for 3270 connections and 700 ms for 5250 sessions. This response time is about 10 times longer than a terminal emulator. The reasons for this lag is because of the heuristic algorithm use to try to determine when the host terminal has been completely refreshed. This problem stems from the mapping of a TN architecture, in which a host refresh is immediately seen at the workstation, to the HTTP request response architecture, where every update is driven from a request from and end user. In HTML, there is no mechanism for host initiated, or asynchronous message occurring from a web server, as there are from a TN3270 or TN5250 server.

A proposed solution to this problem, was a proposal draft, RFC2355, for an extension to the datastream, a direction bit, to indicate when the host screen has completed it's updates, but this draft was not adopted by the IETF and has since been deleted. For more information on this topic, see Other solutions to this problem include introducing an Applet to inform the browser client when it has missed a completed host screen, so the browser client can refresh. This solution has several limitations. It is not supported though a firewall, has scalability limitations, and some customers do not wish to run any Applets in their web browsers.

The solution is to use a back propogation Neural Network to improve the end user response time...