In-Depth

A management framework

• By Tony Baer
• March 27, 2003

Web services may be complex or simple. They may resemble conventional RPCs or provide a dramatically different form of application deployment. Obviously, if Web services are deployed as simple RPCs, heroic systems management measures would not be necessary. But if Web services assume more complex behaviors such as aggregating multiple synchronous or asynchronous interactions, or using discovery or involving third parties, then new approaches that integrate security and higher-level, semantic-oriented functions may become necessary.

Under that scenario, what would a management framework look like?

One proposal would be a dual-layer approach, where the external layer acts, in effect, as the "ambassador" or gatekeeper that performs gatekeeping and compliance roles, including security, life-cycle management and compliance. It would be complemented by an "inner" layer that would manage the service provider infrastructure and service content.

Here is how it would work: For each incoming request, the management framework would apply policies regarding which requestors gain access and, where appropriate, what levels of access to grant and which versions of services to provide. If the request has previously passed through intermediaries, the framework might have to verify the actual integrity of the message and ensure that any relevant public keys are still intact.

Complementing its role as gatekeeper, the management framework could meter usage if appropriate, track compliance with service-level agreements, and provide audit and fault tracking for problem conditions when appropriate. In turn, this data could be submitted to enterprise applications such as ERP, CRM or supplier relationship management systems that handle customer relationship and contract issues.

Internally, a Web services management framework would oversee the functions involved in delivering the service. It would start by provisioning access through brokers or other measures that grant access to underlying infrastructure resources and the services themselves. If any form of service request/discovery process is involved, the framework may also inspect the structure and content of incoming messages to ensure that the requests are acceptable and, in effect, speak the right language when it comes to methods or data structures.

Additionally, although the management framework may not necessarily map or translate data or choreograph multiple workflows or processes, it may have to ensure that all the necessary coordination and mapping processes occur to deliver the service. Underneath the hood, it would act as the traffic cop -- prioritizing, queuing and routing messages; checking service availability and acknowledging that services are delivered; and maintaining the necessary transaction logs and resource utilization statistics necessary for optimizing internal performance.

Please read the associated story "Managing Web services" by Tony Baer.