Automating scalability and high availability services

There are a lot of SOA governance solutions out there that fall into two distinct categories of purpose: one is to catalog services and associated security policies and the other is to provide run-time management for services, including enforcement of security and performance-focused policies.

Vendors providing a full "SOA Stack" of functionality across the service lifecycle (design, development, testing, production) often integrate their disparate product sets for a more automated (and thus manageable) SOA infrastructure. But very few integrate those same products and functionality with the underlying network and application delivery infrastructure required to provide high-availability and scalability for those services.

The question should (and must) be asked: why is that?

Today's application delivery infrastructure, a.k.a. application delivery controllers and load-balancers, are generally capable of integration via standards-based APIs. These APIs provide complete control over the configuration and management of these platforms, making the integration of application delivery platforms with the rest of the SOA eco-system a definite reality.

Most registry/repository solutions today offer the ability of external applications to subscribe to events. The events vary from platform to platform, but generally include some commonalities such as "artifact published" or "item changed". This means a listening application can subscribe to these events and take appropriate action when an event occurs.

1. A new WSDL describing a service interface (hosted in the service application infrastructure) is published.

2. The listening application is notified of the event and retrieves the new or modified WSDL.

3. The application parses the WSDL and determines the appropriate endpoint information, then automatically configures the application delivery controller to (a) virtualize the service and (b) load balance requests across applicable servers.

4. The application delivery controller begins automatically load-balancing service requests and providing high-availability and scalability services.

There's some information missing that has to be supplied either via discovery, policy, or manual configuration. That's beyond the scope of this post, but would certainly be a part of the controlling application.

Conceptually, as long as you have (a) a service-enabled application delivery controller and (b) an application capable of listening for events in the SOA registry/repository, you can automate the process of provisioning high-availability and scalability services for those SOA services.

If you combine this with the ability to integrate application delivery control into the application itself, you can provide an even more agile, dynamic application delivery infrastructure than if you just used one concept or the other. And when you get right down to it, this doesn't just work for SOA, it could easily work just as well for any application framework, given the right integration.

There already exist some integration of application delivery infrastructure with SOA governance solutions, like AmberPoint, but there could be more. There could be custom solutions for your unique architecture as well, given that the technology exists to build it.

The question is, why aren't folks leveraging this integration capability to support initiatives like SOA and cloud computing that require a high level of agility and extensibility and upon which the ROI depends at least partially on the ability to reduce management costs and length of deployment cycles through automation?

It's true that there seems to be an increasing awareness of the importance of application delivery infrastructure to architecting a scalable, highly available cloud computing environment. But we never really managed to focus on the importance of an agile, reusable, intelligent application delivery infrastructure to the success of SOA.

Maybe it's time we backtrack a bit and do so, because many of the same architectural and performance issues that will arise in the cloud due to poor choices in application delivery infrastructure are the same as those that adversely impact SOA implementations.

Governance in the Cloud

Reliability does not come from SOA Governance

Building a Cloudbursting Capable Infrastructure

The Next Tech Boom: Infrastructure 2.0