Browse Prior Art Database

A Method to Analyze Mass of Trace Log by Mapping Customized Text based Expression Pattern to Graphic based Expression Pattern

IP.com Disclosure Number: IPCOM000212476D
Publication Date: 2011-Nov-15
Document File: 5 page(s) / 153K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method to transfer massive text messages into graphic representations for easy understanding and efficient debugging. It maps each line of trace message from a text based pattern to a graphic based pattern. The method uses a two-dimensional representation to describe the parallel concurrency in a dynamic way. The text pattern to graphic pattern mapping can be customized.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 43% of the total text.

Page 01 of 5

A Method to Analyze Mass of Trace Log by Mapping Customized Text based Expression Pattern to Graphic based Expression Pattern

Currently, many software products provide trace mechanisms for diagnosing problems that happen inside the software product at the customer's site or deployment environment. Most of the trace codes are part of the product, compiled together with the product. Normally the trace can be controlled (i.e., enabled/disabled) by setting some switch environment variables. With the trace information, support engineers could locate the internal data and problematic code quickly.

The quantity of the trace messages and the size of the trace log depend on the detail levels of the trace system designed. Often, there is a set of levels (e.g., OFF, INFO, WARN, FINE, FINEST) recognized by the system that one can designate to get fine-grained information for a software application. In a rich information output level, even though there is more accurate runtime information about the application, there are some obvious drawbacks. The quantity will become so massive that it is hard for a human being to read and parse. Analyzing the trace log and identifying the areas in which the support engineers are interested often is extremely time-consuming. The situation becomes worse when facing a complex multithreaded software product.

Figure 1: Example of some trace outputs from an application.

Here, the support engineer was trying to diagnose a performance bottleneck. The engineer enabled application trace and had a trace log to examine. Then, the following problems occurred:


• The trace log is the only useful information to understand the issue


• The call to "obj_get_lock" from line 5292 took 1 second, but not all the calls to

1


Page 02 of 5

"obj_get_lock" took such a long time. The engineer was trying to figure out the caller of this function. It might be "acc_check_locks_on_obj_genl" from line 5291, but it might not be. Once the call stack became large in size with too many layers/frames, the performance bottleneck became quite hard to be located from the trace log.


• The support engineer realized that the application in question is a multithreaded application, so the current trace became meaningless.


• The real problem might have involved two or more threads.

A method to transfer these massive text messages into graphic representations for easy understanding and efficient debugging is needed.

The disclosed solution provides a method to transfer massive text messages into graphic representations for easy understanding and efficient debugging. It proposes a method to map each line of trace message from a text based pattern to a graphic based pattern. The method uses a two-dimensional representation to describe the parallel concurrency in a dynamic way. Text pattern to graphic pattern mapping can be customized.

The invention involves:


• A programming time Application Programming Interface (API) to generate trace messages,


•...