IT Focus Area: infrastructure operations
July 26, 2012
Driving Operational Efficiency with New Network Architecture
Managing a large enterprise network is not a simple task. Network management, which includes the installation, configuration, provisioning, monitoring, testing, and debugging processes, requires detailed knowledge of many different network components, each with its own management interface. Adding to the challenge is the fact that many of these networks are based on design principles that are decades old. This means that over the years, enhancements to the network may have included component upgrades to support new services, but the basic architecture has not adapted to meet changing demands and technology innovations on the data center, making the networks highly customized, costly and difficult to manage.
That is why it is important for enterprises to take a forward-looking approach to simplifying their outdated networks in order to help decrease cost and achieve greater agility while paving the way for automation.
In this article, FOCUS sits down with Ron Jackson, practice manager of Forsythe’s network solutions, to find out how modern network architecture helps drive operational efficiency.
FOCUS: In terms of network architecture, what is a big trend you are seeing today?
Ron: We are seeing that the networks at many enterprises have become too complex and too difficult to manage. In the past, companies didn’t have the need to change network architecture, so they customized their networks to meet business objectives. They made component network upgrades, such as upgrading from 1G Ethernet to 10G Ethernet, but they really didn’t change the fundamental network architecture. So as networks continued—and still continue—to deploy new services, often in a highly customized fashion, the result is a network architecture that is purpose-built and lacks standardization. This, in turn, creates operational issues. Now, with rapidly changing technology, increased business demands and expanding network needs, more organizations understand the need to modernize and simplify network architecture as an effective solution that leads to improved operational efficiency.
FOCUS: What are some examples of the operational issues created by highly customized and purpose-built networks?
Ron: A network that has too much customization is usually a network that is very difficult to manage. In other words, it takes too long to deploy new information technology (IT) services. At some organizations, it may feel as though everything is a “one off” project. This slows down the time to market for new services and increases operational and capital costs. It’s like having to build a road every time you want to go somewhere new. That, of course, is highly inefficient, takes a tremendous amount of time and costs a lot to do. Eventually, you end up with a roadmap that resembles a sort of Rube Goldberg effect. Rube Goldberg created a series of popular cartoons that depicted complex gadgets that perform simple tasks in indirect, convoluted ways.
Today, it is important to design network architectures that are highly repeatable. Building a modular, repeatable network infrastructure means that organizations don’t have to relearn customized infrastructures. This helps companies with their ability to design, deploy and troubleshoot. By streamlining network operations, an organization can reduce costs.
FOCUS: What other factors are leading companies to revisit how their networks are designed?
Ron: Server virtualization is one of the biggest factors. It has changed the dynamics of the data center network. Not only has it changed network traffic patterns, but also the need to support the mobility of servers within the data center and between data centers. Server virtualization has also changed the classic core, distribution and access hierarchical model. This model had been the guiding archetype for most network infrastructures for the last 25 years. With server virtualization, the presence of the hypervisor presents another access layer. If you use the classic hierarchical model, you would end up with a core, distribution, access and access model. That’s two access layers. That adds complexity. New or modern data center architectures eliminate the physical access layer, thus reducing network complexity.
FOCUS: Where does convergence come into play with the new architectures?
Ron: Convergence is the consolidation of multiple infrastructure types or components. A prime example is server input/output (I/O) convergence in the data center. Fibre Channel over Ethernet (FCoE) consolidates server adapters and network ports. Server I/O virtualization replaces multiple I/O cables with a single cable for all network and storage connections. Convergence is a critical step toward maturing the data center infrastructure. Convergence simplifies the data center by consolidating infrastructure types so there are less infrastructure devices housed in the data center. It also improves interoperability with storage, server and network infrastructure platforms. And, finally, it frees up capital.
FOCUS: Can you tell us more about maturing the data center infrastructure?
Ron: In order to transform your traditional data center into a hybrid data center, there are a series of steps that an organization should take to get there. A company should start by standardizing their IT infrastructure through a series of sequential steps: virtualization, convergence, automation, orchestration and hybrid integration.
The new or modern data center architectures specifically integrate the standardization, virtualization and convergence steps within the network platform. However, the architecture also supports the other steps of automation, orchestration and ultimately the hybrid cloud data center. To achieve these other steps, it is beneficial to consider the new data center network architecture.
FOCUS: What are some of the characteristics of new data center network architecture?
Ron: There are four main characteristics of newer architectures that bring about standard uniform expansion: pods, multi-tenancy, service levels, and automated management and provisioning.
FOCUS: What are the first two characteristics?
Ron: To remove complexity in the infrastructure, organizations can build a data center network through the use of pods. Pods provide the uniform homogenous modular unit of expansion that is helpful with scalability. Within the pod architecture, the network components are specified, the amount of compute power that can be attained is determined and interconnectivity with storage systems is defined. This pod or fixed unit of infrastructure can then be cloned to accommodate expansion with pre-determined cost structures and management burdens. With old network architecture, any customizations that need to be made rely on various physical components. Using newer network technologies, customizations are made through virtualization technologies, thereby keeping the physical infrastructure very pod-like.
Another key aspect to the new network architecture is multi-tenancy. Multi-tenancy allows you to securely separate compute environments so that one compute environment, under any circumstance, cannot affect the other. In a classic case, multi-tenancy supports multiple customers, much like a service provider. Many enterprises are increasingly finding the need for multi-tenant-based network architectures. For this application, tenants are the various business units, organizational units or even application sets. In each of these cases, there is a clear need to securely separate services.
For example, let’s examine how a retail organization may have a need for multi-tenancy. That retailer may have multiple brands that it may need to quickly introduce or remove. The retailer may also sell products online and may want the ability to completely secure and separate various brand content and shopping carts on the same physical infrastructure. To make this happen, it would require large capital and operating expenses to ensure the data center had the ability to virtualize network infrastructure resources in order to support this multi-tenant environment. Another example may be that an organization wants to create separate tenants for top-tier applications and other applications within another tenancy. In other words, multi-tenancy requirements depend on your business objectives.
FOCUS: What are the other two characteristics?
Ron: Another important characteristic of the new network architecture is the need to quickly adapt to increased demands with guaranteed service levels. A central part of cloud architectures is the ability to provision resources in an on-demand fashion. This means that if a tenant requires additional resources for a brief period of time, the infrastructure can support the increased demands but maintain the required service levels.
The last characteristic is automated management and provisioning. To efficiently scale and reduce time for provisioning, service orchestration is central to the picture. Automation and orchestration are key elements to cloud-based data centers, and are required to facilitate self-service portals and obtain operational efficiency. With the new data center architecture, all provisioning is done logically, or virtually, so there is no need for physical moves, adds or changes to accommodate any service. This means that services can be deployed within any tenancy, with varying service levels and varying security controls, in an automated fashion. The only thing that an organization needs to add is the orchestration element.
FOCUS: Do the new network architectures support cloud computing?
Ron: Yes, absolutely.
FOCUS: Do you have any advice for those looking to modernize their network architecture?
Ron: Changing the data center architecture is a big deal. It is important to involve IT architecture or design groups, as well as the operations, security, server, storage and IT service management teams. Since the goal is to move toward a service-oriented IT infrastructure, all infrastructure groups should be involved in the planning of the new network architecture. Often, new network architecture is part of a larger IT service optimization project. In this case, starting with the network makes sense since it usually has the longest refresh cycle out of all the different components of an IT infrastructure.
FOCUS: So basically you “rip and replace” to get this new network architecture? Or is there a transition plan?
Ron: No organization can afford a rip and replace. The new network architectures can easily be integrated into the existing data center networks. Generally, a company would start by building out a single pod and then integrate that pod into the existing data center environment. Any net new services deployed or infrastructure purchased is directed toward the new pod. Functions like orchestration can be deferred within the pod until a later date.
FOCUS: What kind of benefits will organizations expect to see with new network architecture?
Ron: With new network architecture in place, organizations will see many benefits. They include improved agility, better support of their data center environment, better efficiency and scalability through standardization, and lower capital and operating costs.