Web services will only become the norm if the users can be assured a level of service. They will not be deployed in earnest unless the operations department is confident that they can provide this level of service, writes Bloor Research analyst Peter Abrahams.
The operations department need tools to monitor the Web services and to react to a situation before the user becomes aware of a drop in the service levels. This is complex when the total service is provided by a bevy of web services, written by and provided by different departments and enterprises.
Woe is me
I have spent some hours today trying to find out why an email a company sent me did not arrive, this is not a Web services problem, but it should be simpler to resolve. But tracking missing emails is not a trivial problem; my machine did not receive it, the ISP did not bounce the message, nor it appears did they receive it, but the sender is certain he sent it. So who is to blame?
Several emails and phone calls between the three main parties confirmed several possible reasons that did not cause the problem. Several tests removed other possible reasons. The problem has not been resolved but there is a fair chance that it will not happen tomorrow and we will never know why it did not work today. This problem resolution is an expensive, time -consuming and frustrating process. It would not be practical to do it for a bevy of WS.
What tools does an operations manager need to automatically support web services? First, he needs to be able to ensure that the individual web service and the platform it sits on are performing correctly. This is really no different to system management of traditional applications and the same tool can still be used. The next level is monitoring the performance that one web service supplies to another and checking that it meets the service level agreed between the two services. Actional Looking Glass and, the recently announced, CA Unicenter Web Services Distributed Management (WSDM) both assist in this space.
These tools measure the performance of running processes, so will only detect a problem when a new process runs. If the web service is unavailable, the time to fix the problem will be considerably and will adversely affect the calling web service or end-user. This will mean at least one unhappy client but is likely to mean a lot more as new requests pile up.
To be really effective there is a need for a monitoring tool to discover the problem before another service or user is affected. The only way to do this is to run synthetic transaction that ping the service to check that it is still there.
Actional has just announced its Watchdog product that creates and monitors such synthetic transactions. It can be used in two ways, firstly just to ping a local service to ensure that it is still available. The second use is to initiate the ping from a remote service to check that the service is accessible and has a response time within the SLA from the remote site.
Watchdog is just the latest in a series of new functions that are needed to fully manage Web services. We expect further functions to be added as web services become mainstream. One obvious extension would be the ability to include business understanding in the monitoring, for example the SLA for a Gold Card customer would be different to a standard card.