ECDN 101: Advanced Topics in ECDNs

Building an ECDN

The adoption of those applications and services affects everyone in the enterprise, particularly the WAN architects assessing and re-assessing wide-area network bandwidth requirements, building out redundant data centers to support the applications and services, and eliminating slow response times that affect a user's ability to perform his or her job.

Building Out an ECDN Begins at the Core
The core of the network is where the servers and applications reside. Think of the core as the supply-side of the network. The first thing businesses must do is guarantee that the supply of content is always available. This requires the creation of a server farm with high availability and load balancing across those resources--ensuring that users can consistently gain access to the content and applications. Secondly, and equally important, is the quality of that supply. Content and applications that are out of sync or corrupted shouldn't be accessible by anyone, regardless of whether they are internal or external customers.

The third concern is that the supply must be secure. Most, if not all, the data on those servers is the intellectual property for that organization, and must be protected.

The Local Area Traffic Manager
Therefore, at the core of the network, a smart local area traffic manager is required. It can address intra-data center availability, and guarantee that Web services and applications are always available by providing high availability and load balancing across the server farm. The right traffic manager is smart enough to intercept all IP traffic, inspect the traffic to determine requestor information, transform the traffic to support QoS, security or business criteria, and then direct the traffic to the appropriate resource based on business policies. The right product here can also verify the quality of the supply and the health of the applications or content through scripting and other mechanisms.

Disaster Recovery
Furthermore, to build a highly available ECDN, businesses should have a disaster recovery plan to ensure that the supply is always available. The architecture of the first data center should more or less match that of the second data center. This second data center can either be an exact replica of the first, or a slightly scaled-back mirror image. ECDN solutions allows businesses to build an architecture that load balances and provides high availability across both data centers, thus giving them an active-active or an active-standby deployment. High availability across both data centers is achieved by a global load balancing product.

This product ensures that users are directed to the most available data center. Its global redirection capability guarantees both availability and performance across data centers at the network level by aggregating data, such as round trip times, router hops, and completion rates. It then combines this information gathered from each data center, such as the health of a network connection, and metrics collected from the local area traffic manager and the caches to intelligently direct the user to the most available supply of content and applications.

The Need For Caching
So we've talked about the supply of content. But what about the demand? Demand for the content and applications is scattered throughout the enterprise network. Making sure that this demand gets fulfilled is critical. Think of the cache component of the ECDN as the mechanism that fulfills the demand. Whereas the majority of the supply of content and applications exists at the enterprise network core, caches are deployed in both data center and regional and branch offices to assist in fulfilling the demand by allowing far reaching distribution of the supply. With the cache, the performance and accessibility of the content and applications are greatly increased.

But the fulfilling of content isn't the only benefit cache deployment offers. By deploying caches, businesses can delay any costly bandwidth upgrades. Caches also act as a policy enforcement point in several ways. First, only users with the proper permissions can access wide-area resources. This authorization can take place against an LDAP server or by using NTLM. Second, users can only access appropriate content based on corporate policies. Finally, bandwidth usage can be metered and controlled through the cache itself. These policies can be configured at the IP address level, or at the protocol level (for example, 75 percent of mission-critical applications can be allocated to flow through the cache, with 20 percent for streaming media, and 5 percent for non-mission critical applications).

Those are all important points in maximizing network resources within an enterprise CDN. The problem that arises with this deployment of local traffic mangers, global redirection technologies, and caches is this: how can a business make sure that the servers at the data center in Tokyo have the same content as those in the San Francisco data center? How can a business in Seattle guarantee that the latest stream file that describes some new manufacturing process is up-to-date in their plant in London?

Content Distribution: Guaranteeing That Content Is Up To Date
This brings us to the fourth critical component of the ECDN: the content distribution solution. Its most important function is to guarantee the quality and availability of the content and applications on the ECDN. Imagine having the wrong server object or .DLL on a server, and allowing user to access that application. The application may still respond, but who knows what that response might be?

By pairing the content distributor with the local traffic manager, businesses can now guarantee that the server will have the right object. If it doesn't, the content distributor tells the traffic manager not to direct users or cache requests to that resource. Or take another example. Suppose a business has over a hundred GB of content in PDF and Word formats. Suppose they average approximately one hundred MB of new or changed content every month--content that must be distributed to over one hundred and fifty branch offices. Some of these offices have 64K bit frame relay connections. Before building out an ECDN, each time a user accessed any of those documents, their requests had to go all the way back to the network core--slowing response times and clogging up the network.

By combining a content distributor with a cache, the business can now schedule distribution of this content during, say, weekends or evenings (off-hours). Users will have access to the content that is waiting for them when they arrive at work Monday morning while eliminating the risk that the document they are accessing is an older, inaccurate version.

ECDN Management
The fifth component of the ECDN is its ability to be efficiently managed. As the ECDN gets rolled out, management of the many devices becomes difficult at best. What if a policy change needs to occur on the cache in Munich? A centralized device manager solves this type of problem. For example, assume that the amount of streaming traffic allowed through the caches is set at 10%, and the CEO of the company just announced that he wants to send out a company-wide stream for an important speech. With the right device manager, businesses can quickly change the policy on the cache to 90 percent, and distribute that policy to its three hundred caches located throughout the network. After the event is complete, policies on the caches can be restored to their previous levels.

The Open Architecture
This last component of the ECDN is the open architecture. I talked more extensively about this in my previous class, but basically the open architecture allows open communication between the devices. By having these elements in the network "talk" to one another, and also getting the network infrastructure to talk to the applications that are driving the content, businesses can automate many of their operations. This interaction allows the environment to scale up much more elegantly, without the operational costs of having to adjust every single element whenever a change occurs.

An open architecture with well-defined APIs (application programming interfaces) enables both customers and their partners to create interactions with each component of the network, as well as creating intercommunication between 3rd party applications and the network.

So that's it. Six components are required for building the ECDN, including local area load balancing (traffic management), global load balancing and redirection, caching and streaming, content distribution, network management, and an open architecture. Building an ECDN infrastructure with these key components lets companies deliver rich digital media, web-based applications and web services to employees, vendors, partners and customers worldwide, in the fastest amount of time at the lowest possible cost. And for ECDNs, that's really the bottom line.