End-to-end is a popular term in marketing circles to describe some feature that acts across an entire “something.” In the case of networking solutions this generally means the feature acts from client to server. For example, end-to-end protocol optimization means the solution optimizes the protocol from the client all the way to the server, using whatever industry standard and proprietary, if applicable, techniques are available. But end-to-end is not necessarily an optimal solution – not from a performance perspective, not from a CAPEX or OPEX perspective, and certainly not from a dynamism perspective.

The better option, the more optimal, cost-efficient, and context-aware solution, is a top-to-bottom option.


WHAT’S WRONG with END-to-END?

“End-to-end optimization” is generally focused on one or two specific facets of a connection between the client and the server. WAN optimization, for example, focuses on the network connection and the data, typically reducing it in size through the use of data de-duplication technologies so that it transfers (or at least appears to transfer) faster. Web application acceleration focuses on HTTP and web application data in much the same, optimizing the protocol and trying to imagereduce the amount of data that must be transferred as a means to speed up page load times. Web application acceleration often employs techniques that leverage the client’s browser cache, which makes it and end-to-end solution. Similarly, end-to-end security for web-based applications is almost always implemented through the use of SSL, which encrypts the data traversing the network from the client to the server and back.

Now, the problem with this is that each of these “end-to-end” implementations is a separate solution, usually deployed as either a network device or a software solution or, more recently, as a virtual network appliance. Taking our examples from above that means this “end-to-end” optimization solution comprises three separate and distinct solutions. These are deployed individually, which means each one has to process the data and optimize the relevant protocols individually, as islands of functionality. Each of these is a “hop” in the network and incurs the expected penalty of latency due to the processing required. Each one is separately managed and, what’s worse, each one has no idea the other exists. They each execute in an isolated, non-context aware environment.

Also problematic is that you must be concerned with the order of operations when implementing such an architecture. SSL encryption should not be applied until after application acceleration has been applied, and WAN optimization (data de-duplication) should occur before compression or protocol optimization is employed. The wrong order can reduce the effectiveness of the optimization techniques and can, in some cases, render them inert.


THE TOP-to-BOTTOM OPTION

The top-to-bottom approach is still about taking in raw data from an application and spitting out optimized data. The difference is that data is optimized from top (the application layer) to the bottom (network layer) via a unified platform approach. A top-to-bottom approach respects the rules for application of security and optimization techniques based on the appropriate imageorder of operations but the data never leaves the platform. Rather than stringing together multiple security and optimization solutions in an end-to-end chain of intermediaries, the “chain” is internal via a high-speed interconnect that   both eliminates the negative performance impact of chaining proxies but also maintains context across each step.

In most end-to-end architectures only solution closest to the user has the endpoint context – information about the user, the user’s connection, and the user’s environment. It does not share that information with other solutions as the data is passed along the vertical chain of solutions. Similarly only the solution closest to thecontext application has the application endpoint’s context – status, condition of the network, and capacity. A top-to-bottom, unified approach maintains the context across all three components - end-user endpoint, application endpoint, and the network – and allows each optimization, acceleration, and security solution to leverage that context to apply the right policy at the right time based on that information. This is particularly useful for “perimeter” deployed solutions, such as WAN optimization, that must be by design one of the last (or the last) solutions in the chain of intermediaries in order to perform data de-duplication. Such solutions rarely have visibility into the full context of a request and response, and are therefore limited in how optimization features can be applied to the data. A top-to-bottom approach mitigates this obstacle by ensuring the WAN optimization has complete visibility into the contextual metadata for the request and response and can therefore apply optimization policies in a dynamic way based on full transactional context.

Because the data never leaves a unified application delivery platform, the traditional performance penalties associated with chaining multiple solutions together – network time to transfer, TCP connection setup and teardown – are remediated. From an operational viewpoint a top-to-bottom approach leveraging a unified application delivery platform decreases operational costs associated with management of multiple solutions and controls the complexity associated with managing multiple configurations and policies across multiple solution deployments. A unified application delivery approach uses the same device, the same management mechanisms (GUI, CLI, and scripting) to configure and manage the solution. It reduces the physical components necessary, as well, as it eliminates the need for a one-to-one physical-solution relationship between solution and hardware, which eliminates complexity of architecture and removes multiple points of failure in the data path.


TOP-to-BOTTOM is END-to-END only BETTER

Both end-to-end and top-to-bottom ultimately perform the same task: securing, optimizing and accelerating the delivery of data between a client and an application. Top-to-bottom actually is an intelligent form of end-to-end; it simply consolidates and centralizes multiple components in the end-to-end “chain” and instead employs a top-to- bottom approach to performing the same tasks on a single, unified platform.