Evolution of Web Services Infrastructures

The Web service infrastructure is middleware, which is embraced by both early adopters and mainstream enterprises interested in reducing costs and integrating legacy applications. The Web service paradigm is platform neutral as well as hardware and software neutral. Thus enterprises can easily communicate with employees, customers, business partners, and vendors while maintaining the specific security requirements, access, and privileges needs of all. The examples used in this book will focus on Web-based systems, which simplifies our discussion from a networking perspective.

FIGURE 1-1 shows a conceptual model of how this new paradigm impacts the data center network architecture, which must efficiently support this infrastructure. The Web services-based infrastructure allows different applications to integrate through the exposed interface, which is the advertised service. The internal details of the service and subservices required to provide the exposed service are hidden. This approach has a profound impact on the various networks, including the service provider network and the data center network that support the bulk of the intelligence required to deliver these Web-based services.

Figure 1-1. Web Services Infrastructure Impact on Data Center Network Architectures

Data center network architectures are driven by computing paradigms. One can argue that the computing paradigm has now come full circle. From the 1960s to the 1980s, the industry was dominated by a centralized data center architecture that revolved around a mainframe with remote terminal clients. Systems Network Architecture (SNA) and Binary Synchronous Communication (BSC) were dominant protocols. In the early to mid 1990s, client-server computing influenced a distributed network architecture. Departments had their local workgroup server with local clients and an occasional link to the corporate database or mainframe. Now, computing has returned to a centralized architecture (where the enterprise data center is more consolidated) for improved manageability and security. This centralized data center architecture is required to provide access to intranet and Internet clients, with different devices, link speeds, protocols, and security levels. Clients include internal corporate employees, external customers, partners, and vendors each with different security requirements. A single flexible and scalable architecture is required to provide all these different services. Now the network architect requires a wider and deeper range of knowledge, including Layer 2 and Layer 3 networking equipment vendors, emerging startup appliance makers, and server-side networking features. Creating optimal data center edge architectures is not only about routing packets from the client to the target server or set of servers that collectively expose a service, but also about processing, steering, and providing cascading services at various layers.

For the purposes of this book, we distinguish network design from architecture as follows:

• Architecture is a high-level description of how the major components of the system interconnect from a logical and physical perspective.

• Design is a process that specifies, in sufficient detail for implementation, how to construct a network of interconnected nodes that meets or exceeds functional and non-functional requirements (performance, availability, scalability, and such).

Advances in networking technologies, combined with the rapid deployment of Web-based, mission-critical applications, brought growth and significant changes in enterprise IP network architectures. The ubiquitous deployment of Web-based applications that has streamlined business processes has further accelerated Web services deployments. These deployments have a profound impact on the supporting infrastructures, often requiring a complete paradigm shift in the way we think about building the network architectures. Early client-server deployments had network traffic pattern characteristics that were predominantly localized traffic over large Layer 2 networks. As the migration towards Web-based applications accelerated, client-server deployments evolved to multi-tier architectures, resulting in different network traffic patterns, often outgrowing the old architectures. Traditional network architectures were designed on the assumption that the bulk of the traffic would be local or Layer 2, with proportionately less inter-company or Internet traffic. Now, traffic is very different due to the changing landscape of corporate business policies towards virtual private networks (VPNs), consumer-to-business, and business-to-business e-commerce. These innovations have also given rise to new challenges and opportunities in the design and deployment of emerging enterprise data center IP network architectures.

This book describes why these network traffic patterns have changed, defining multi-tier data centers that support these emerging applications and then describing how to design and build suitable network architectures that will optimally support multi-tier data centers. The focus of this book spans the edge of the data center network to the servers. The scope of this book is limited to the data center edge, so it does not cover the core of the enterprise network.