Service Oriented Architecture (SOA) and Web Services (along with the underlying XML protocol) promise to greatly simplify the implementation of distributed computing applications, both within the enterprise and between enterprises. The widespread acceptance of Web Services across operating systems, middleware, and application vendors will lead to simplified interoperability, thus allowing for increased business agility and lower development costs.

Today enterprises can use IP (Internet Protocol) Virtual Private Network (VPN) offerings from their IP service provider for secure high-performance interconnect services both within the enterprise's own sites and with partners. Price erosion in voice services and intense competition in data services is leading many IP-based service providers to seek out new revenue opportunities by offering new services to their enterprise customers. As service providers move to offer innovative new Web Services, enterprises will benefit just as they did when service providers moved from traditional private line services to IP-based services.

The next step in the evolution of networks is the move to application-aware networking. Networks will become "smarter" by identifying, prioritizing, and moving traffic faster. In addition, distributed application-awareness in the Wide Area Network (WAN) provides a cost-effective and scalable solution for messaging applications in many sectors from financial services to supply chain services.

Today's Situation
Currently, the wide area network is unaware of the application messages that flow across it. A service provider network can offer differentiated Quality of Service (QoS) to control latency and packet loss for applications, but it's not actually aware of the applications being supported.

In the point-to-point integration across a traditional WAN, as shown in Figure 1, the number of point-to-point relationships explodes as the complexity of the Service Oriented Architecture expands across the WAN. While Web Services support location transparency through service registries, services may still need to be moved due to faults, e.g., moving services to a backup location. Techniques such as intelligent DNS can be used to redirect requests to a backup location, but they have limited flexibility.

Intelligent routing of application messages is also not easily available with the architecture shown in Figure 1. For example, consider an enterprise that wants purchase orders over a certain amount to go to one data center and other purchase orders to go to a different data center. If a Web Service interface is used to submit purchase orders, the invoking application would have to be aware of a policy to invoke the service and connect to the correct Web Service endpoint. In a truly loosely coupled environment, the invoking application shouldn't have to be aware of the business logic of its partners.

Point-to-point integration also requires the service requestor and service provider's Web Services endpoints to use the same transport protocol. In a loosely coupled architecture, the service requestor is shielded from such details.

There are also advantages to event-driven architectures in which an application can generate events that may be of interest to many other applications. Normally enterprise middleware can provide publish/subscribe services to loosely couple the applications. This means that the application that generates an event doesn't need to be aware of the applications that are interested in the event. Loose coupling increases business agility since new applications can be added quickly without disturbing existing applications. While such middleware solutions are available within the enterprise, they are typically difficult to scale across the WAN and difficult to implement between enterprises.

Application-Aware Networking: Enabling Value-Added Services
Application-aware networking addresses two current distributed technology challenges: first, the inefficiencies associated with middleware, and second, speed and scale. Speed is obvious - networks always have a need for speed - but the speed required for today's real-time financial networks, for example, is a truly demanding benchmark (see the section on financial services below). Middleware - the software that connects network applications - is an integral part of most enterprise data centers. Middleware acts as a translator between incompatible applications by exchanging information in a common format, usually XML, to distribute messages between applications, distributed geographies, or even different companies. These middleware deployments are very complex, add performance overhead, and are challenging to manage. Industry experts have recently called for this middleware function to be moved into the core of the network as a shared resource to simplify the task of running distributed applications. Today's applications become value-added services that take advantage of application-aware, silicon-based hardware to manage information distribution at wire-speed and add a layer of intelligence to the network.

Most service providers have spent the last few years upgrading their core networks to the latest IP standards at great financial expense. Although service providers are very aware of the critical need to develop new revenue-generating services, they are naturally reluctant to make any additional changes to the core of their network. A successful application-aware networking approach should function as an overlay to the existing network to avoid the risk associated with once again upgrading the entire network. Value-added services can then be delivered over the existing infrastructure and capture new revenue.

As shown in Figure 2, the service provider's application-aware network routes messages more efficiently. For example, applications can connect to the network and automatically receive messages of interest. Other applications can send messages without any knowledge of the routing policies being used by other parties. This allows intelligent application message routing across the WAN, leading to true loosely coupled applications.

The application-aware network also allows the WAN to bridge different transport protocols (e.g., between a hyper-text transport and a Java messaging service). It can route messages based on their content at wire speed, provide QoS based on the content of the message being routed, and deliver a message to many interested endpoints (i.e., publish/subscribe on a per-message basis based on message content). It can also perform wire-speed message format transformations inside the network.

The application-aware network allows endpoint applications to dynamically alter the message routing rules so that the enterprise users retain control of message routing. Similar to the underlying IP/MPLS network, the application-aware network also provides VPN so that messages are only routed to entitled endpoints.

Fostering Enterprise and Service Provider Partnerships
Service provider enterprise customers can benefit from this move to the network. As the trend towards outsourcing parts of the data center grows, more and more enterprise CIOs are seeking trusted third-party hosts. A prime example of the success of the outsourcing trend is salesforce.com, which has been providing outsourced Customer Relationship Management (CRM) solutions since 1999. Salesforce.com recognized that most CRM requirements are the same from company to company. For those companies, it makes little sense to spend years and millions of dollars building a custom system with SAP or Siebel. A standard set of services, implemented with minimal configuration, saves them time and money. This same business model is a key opportunity for service providers to help companies outsource expensive infrastructure that doesn't in itself generate revenue and focus on their core business requirements.

Beyond Connectivity - The New Service Revolution
While the promise of new revenue streams is highly attractive, service providers know there are technical challenges that must be solved to deliver new services to the supply chain, business monitoring, and financial markets.

Supply chain services
As companies begin to implement the full potential of radio frequency identification (RFID) to track goods in real-time through the supply chain, they encounter a massive increase in data traffic. Supply chain managers need to know when their pallet of cola leaves the warehouse, when it passes key milestones in transit, and when it arrives at its destination. However, enterprise networks are ill-equipped to transmit this data quickly and provide the level of monitoring required. Value-added services routing systems let service providers develop outsourced RFID monitoring solutions and track shipments at wire speed.

Business monitoring services
As networks transform from simple connections that blindly carry data to intelligent networks that understand the content flowing through them, new services that monitor and report on enterprise business activity become possible. Customers can define business rules that represent intended behavior and the network can track actual traffic against these metrics. This kind of monitoring technology lets service level agreements (SLAs) move from simple uptime and bit-flow guarantees to application-level SLAs assuring high-quality behavior of individual transactions. It can also give enterprises new opportunities to offload the burden of corporate regulatory compliance initiatives to service providers.

Financial services
A core area of focus in the financial services market (including banks, brokerages, and investment firms) is reducing information latency. Currently market data is sent from the trading floors to aggregators like Reuters, which then distribute it to the subscribers. However, this middleman system is not only expensive, it delays delivery. Although the lag is a second at most, a financial services company relying on increasingly popular algorithmic trading software needs to know that shares in Google or Exxon have changed before its competitor does. The lag can mean the difference between making or losing millions. With value-added services routing systems, service providers can provide a direct wire-speed link among trading floors around the world. They can deliver traffic, including these algorithmic trading applications, between trading offices or to remote locations. Not only does this method of distribution reduce latency times, it's more efficient. Unlike traditional direct feeds, the solution is completely interest-driven: only the data streams that traders or applications have requested flow across the network.

These services are just a snapshot of the possibilities presented by moving messaging onto service provider's networks. Since value-added service routers can identify and deliver any kind of content intelligently to interested parties at wire speed, the potential is infinite.