SOS for SOAs

The deployment of service-oriented architectures (SOAs) has sparked a Web services revolution. Recent surveys indicate that nearly 90 percent of enterprises will employ Web services within the next year, and two-thirds will use them in a significant way.

Web services create reusable, cross-platform, software modules, which expedite the cost-effective development and rollout of new applications and the sharing of information between applications. However, that sharing comes at a cost to the wide area network (WAN).

To ensure continuous operation of mission-critical applications, more bandwidth and higher network performance must be provided to support Web services and their related remote data replication traffic.

In fact, the proliferation of Web services within enterprises will have three major impacts on WANs:

1. Inter-site traffic volume will grow exponentially.
Today, applications are normally executed within a single location. When Web services are used, code execution is distributed at the location of the particular Web service program or applet typically among multiple sites, adding more WAN traffic. And it’s not just the number of packets traversing sites, but the size of those packets. A key Web services standard is eXtensible Markup Language (XML), which is a particularly inefficient protocol, with some estimating that it requires two to 100 times more bits than non-XML binary transfers.2. Low latency will be increasingly critical.
When applications execute on a single server, inter-process communications are handled by the server’s system bus. When processing is distributed with Web services, inter-process communications are carried over the network. The performance of the network can significantly affect the end user’s applications experience. Having a network that behaves like a system bus – that is, predictable, deterministic, and low latency – will be critical to dramatically improve total applications performance and, ultimately, the end-user experience.

3. High priority packet identification will become increasingly difficult.

Web services can be combined in often unpredictable and complex ways. Going forward, it will be difficult to ascertain when a particular Web service is being used as part of a mission-critical business process and how to prioritize the associated traffic.

To address this impact on WANs, two primary methodologies have emerged – one that focuses on application layer networking and one on the optical and data layers.

The application layer approach addresses the Web services onslaught with packet content processing and switching at Layers 5 to 7. While this brings some XML processing and application load balancing benefits, there are significant security, cost and performance issues as it depends on expensive and time-consuming “tuning” of applications.

The other approach avoids those issues by focusing on high-performance bandwidth at Layers 1 and 2, rather than optimizing application prioritization over networks of limited bandwidth. And, by handling applications transparently, no applications tuning is required. The end result is a flexible, adaptable, WAN that is always available, never drops packets, and delivers predictable, deterministic, network response so enterprises are able to cost-effectively support Web services and their real-time, transaction-oriented, application traffic over multiple sites.In the end, enterprises need to react to the impact of Web services with an optimal blend of application-oriented networking and Layers 1 to 2 network technologies. Addressing these challenges with only application-layer networking tools to prioritize high-priority traffic during peak traffic periods, enterprises will force their applications to struggle and compete for clogged WAN resources.

More importantly, their investments in Web services hardware and software will not deliver the expected benefits.

— Michael Mullaley, Director, Enterprise Services Ciena Corp.