Ajax Logger. Client Monitoring Technique with High Installability & Scalability. November 2007

Transcription

1 Ajax Logger Client Monitoring Technique with High Installability & Scalability November 2007 Hitachi Ltd., Central Research Laboratory Tomohiro Nakamura 1

2 1 Introduction The importance of Web system administration is growing - web applications everywhere - mission critical web applications are taking off Web system administrators having problems because - real user-side performance is uncertain - which server logs have a key to a complaint from users Defects are found everywhere from servers to clients - client monitoring is required (not only server monitoring) - link client-side performance monitoring logs to server-side logs Normal performance? Performance Monitoring Logs Clients Admin Servers User Complaint? Logs 2

3 2 Existing Client Monitoring Technique Stopwatch Method & Monitoring Agent Stopwatch method - observer measures user-side performance with stopwatch Operators (Users) Monitoring agent - client application to monitor user-side performance Monitoring Agents Observer Operators (Users) High cost No scalability No sustainability Stressful No server-side analysis Messy installation (low scalability) Expensive Unwilling users No server-side analysis 3

4 3 Proposed New Client Monitoring Technique Ajax Logger -Features & Functions- Client Monitoring Technique using Ajax Technology - no need to install tools on client PCs, only Web browser - no need to modify web applications, only to change server settings HITACHI Net Shopping - No change in Web page s appearance - Script code runs background Web Response (HTML) Functions of the script code (1) Event handling (makes event logs) (2) Log store on local disk (using userdata persistent) *The userdata behavior persists information across sessions by writing to a UserData store. Web server inserts script code into the response Clients Servers Store logs on local disk userdata persistent Local disk Send logs to server if necessary 4

5 4 Elemental Technology Ajax Logger - ID Handling - ID Associates Logs - can retrieve server log from client log & Web page access history Client-side (Web browser) Client log log (Web (Web page page A) A) Previous page ID Current page ID N/A 100 set ID of Web page A to cookie link the previous Web page Client log log (Web (Web page page B) B) Previous page ID Current page ID Request for page A Previous ID Response of page A Request for page B Previous ID Response of page B Link the server-side log Server-side (Web server) Server log (Web page A s trace) Current ID Trace ID 100 insert ID of Web page A get ID of Web page A from cookie Server log (Web page B s trace) Current ID Trace ID 200 insert IDs of Web pages A & B * Trace ID is unique value 5

6 5 Implementation Example Ajax Logger - Perfomance Monitor - Client-Side Performance Monitoring - break performance problem down into communication, rendering, etc. record onactivate event record onload event Web Browser Java Script Engine HTML Interpreter Communication time Rendering time Request get prev. ID from cookie insert Ajax Logger code & prev./current ID into resp. Response Web Server Servlet Filter Application Trace Logger Ajax Logger Filter Trace collaboration get trace ID ( = current ID) set current ID to cookie record beforeunload event store records into local disk Operation time Next request 6

7 6 Application Example Ajax Logger - Log Monitoring (1) - Prototype Log Monitoring - user-side performance monitoring Access history Each page access has unique ID linked to server-side log easy inquiry using ID User-side performance break down 7

8 7 Application Example Ajax Logger - Log Monitoring (2) - Prototype Log Monitoring - statistics analysis Sort by page access count analyze page access trends Best, worst, average performance analyze each page s features (detect heavy pages, tricky pages, etc ) 8

9 8 Conclusion Ajax Logger - Client Monitoring Technique using Ajax Technology - High Installability: no need to install tools on client PCs no need to modify web applications - High Scalability: store logs on local disk, no server-side additional loading - Application Example Client-side Performance Monitoring break performance problems down, easy inquiry using log ID Future Work - More types of applications (ex: collaboration with anomaly detection) - Apply for usability evaluation 9