Dave Hewetson, Focus Group Chairman, opened the proceedings and welcomed members, thanking ICL for hosting the event. Being a small group he took the opportunity to invite everyone to introduce themselves.
Ian Manton is Candle's European Business Development Manager for Electronic Business Assurance. We have long talked about and used SLAs but increasingly important, particularly with the growth of business via the Internet, is delivery of service. With increasing use of Internet and web-based services customers' traditional expectations of service remain and indeed are often higher. Companies may be tempted to provide extensive and complex services through Internet technologies, They must not lose sight of the over-riding need for ease of use. It is essential to monitor customer needs. They will be regularly seeking enhanced services. Keep an eye on the competition. Customers want good performance; good response times. Some recent research has shown that 65% of Internet-based transactions are not completed. 40% of customers said that poor performance caused them to leave the service.
Ian reviewed appropriate metrics and then took a look at the key issues by reference to Vodafone as a case study. It was an opportunity to identify the Candle product ETEwatch. Using this to monitor transactions it was possible to establish where gains could be made and provide immediate access to information for Call Centre Managers.
Roger used a presentation that he had recently delivered to UKCMG to explain how ARM was used for Inland Revenue services by EDS. IBM NUMAQ (ex Sequent) and HP V-Class application servers support a network of some 60,000 client NT workstations, some with local NT servers. Whilst ARM can be used on both client and server systems EDS currently deploy it only for servers.
A Best/1 ARM agent on HPUX currently shows high overheads. Action is with BMC to sort this out and hopefully deliver a similar agent for Dynix. In the meantime, Roger has developed code to produce similar information, he calls it BARM, which has tested successfully on HPUX and is about to be validated on Dynix. BARM has been based on Richard Spurr's MINX ARM agent. BARM provides:
- transaction count
- count over 1 second
- count over service level objective
- average service time
- standard deviation service time
- maximum service time
The ultimate aim is to use Best/1 ARM on Dynix.
End to End response time monitoring does not provide statistics about queues for access to servers. Currently EDS are evaluating some products, all of which are based on windows event trapping. They now call ARM and all use their own processing architectures. Use of ARM is possible at the workstation but would not give information for business applications without some modification, and, significantly, testing would be difficult.
In this presentation Simon opened the debate about the use of agents versus robots. His claim is that robots are more effective and do not carry the potential overheads one might expect from software agents. He spoke about the use of robots in End to End Response Time Monitoring drawing some experience from companies such as British Airways, Co-operative Bank (Smile), Marks & Spencer, and Merill Lynch.
The use of robotic, continuous measurement, is proactive. Applied to end-user monitoring through the use of appropriate scripts it offers end-user information which IT managers can share with users to provide a foundation for informed discussion between the IT team and end users. Such information helps to overcome user distrust, provide support for investment dialogue and helps to detect failure in unmeasured components. Simon used a case study - Bell Atlantic - to show the use of robotic measurement. The fundamental tool is EnView running as a dedicated slim-line NT workstation installed at user locations. It simulates user activity, accessing the same user applications. Logs of monitoring sessions are passed to a "Collector" which holds such details as response time objectives, and so enables comparisons to be made an triggers chosen alerts as required. A benefit is to be able to identify performance issues before impacting the end user - a critical benefit for Bell Atlantic.
The scope for monitoring end user activity is such that if used by management to identify exactly what a particular user may be doing it may raise some sensitive questions particularly in the minds of trade union representatives. Food for further thought!
James described some of the performance aspects of a major project which will touch all of us who support their local post office. The Pathway project to bring every Post Office online in support of both Post Office Counters Limited and Benefits Agency business is very high profile. As at 12th May, the end of the week prior to this meeting, 6363 Post offices were up and running (this included 14,463 counters). The roll-out rate is 300 per week.
Two Data Centres, the main one at Bootle and the Standby at Wigan, support the service with individual Post Offices connecting via ISDN over a triple routed network. The size of the network has made it impossible to do full-load tests, but from the graphs displayed traffic load predictions have been very well assessed.
At each Post Office counter an NT workstation provides application support. Only the user application is visible through touch screen capability. A gateway server at the Post Office synchronises with the Data Centre periodically, typically at 30 minute intervals. Each Data Centre synchronises with the other almost simultaneously. At the Data Centre the EMC configuration provides a third mirror disk which is described as the Business Continuity Volume. During the night this is 'split away' from the message servers to enable back-up to be taken.
Metron's Athene is in use for automatic collection of performance data.
Peter and Baz provided a double-headed presentation looking at their product VitalSuite in the context of supporting a large Financial Services Provider with mission critical information needs. The perception of the problem sounds familiar, desktops running slowly, poor response between locations, and a network problem. VitalSuite follows the idea of using software performance monitoring agents, which Peter and Baz argue, can hold their own in the agents versus robots debate. The Suite has three key componenets:
- VitalAgent : centrally managed, configured and deployed to the desktop
- VitalAnalysis : historical application performance reporting tool, form an end-user perspective, SLA comparison
- VitalHelp : real time alerts and desktop troubleshooting tools.
Dave Hewetson thanked ICL for hosting the meeting and providing an excellent buffet lunch. He thanked all speakers for their presentations.
8. Next meeting
The next meeting is expected to be held late November/early December 2000. Would any member volunteer to host the meeting? Provision of lunch is very welcome but not a prerequisite. Please email Dave Hewetson if you wish to host the meeting and also if you have a topic(s) which you would like to air or to be aired.
to AXiS Home Page.Copyright © AXiS 1999 Created: 23rd May 2000