January 17, 2013

159 Views

WAN (Wide Area Network) traffic Optimization

Blog Context

Here are some thoughts on new avenues and directions for the WAN optimization technology domain

Problem Statement

The fundamental problems we are trying to solve using WAN optimization are:

  • Get higher traffic throughput on provisioned/available bandwidth over the ever-expensive WAN links

  • Optimize the usage of expensive WAN links

In simple terms, squeeze “more out of less,” and in business terms, a higher Return on Investment (RoI) and lower Total Cost of Ownership (TCO).

Standard approaches to solving the above problems:

  • Deploy better transport mechanisms on the network (independent of optimizations performed by host operating systems)

  • Content compression techniques with additional contextual information about the network flow

Some directions where the WAN optimization domain can evolve are

The Cloud

Cloud (Storage, Compute and Network Virtualization) technology provides a unique proposition where optimizations done on the network can be augmented by optimizations done at respective application/host layers. More precisely, cloud technology facilitates all the elements such as storage, compute and network elements to contribute respectively to the overall optimization achieved, to a certain common orchestration.

Dynamic vs. static content “strategic” positioning

Conscious positioning of redundant information for an enhanced user experience:

As an analogy, well-known XML technology is intended to achieve a separation of the data model from the presentation (GUI – Graphical User Interface) aspects.

Along similar lines, the presentation aspects, or in general an application skeleton (GUI), can be moved to the network periphery and the active data set that changes frequently can be allowed to flow across the network.

Today, there is no clear distinction made by protocols on the presentation and data model. To a certain extent, having a client application solves this issue. But more often, we may need to separate them by inspecting the flow which is a major throughput bottleneck for WAN optimization technology. More specifically, HTTP-based data transfers are not very efficient at this.

Given such a possible mechanism, for the WAN service provider counterpart playing their part in the midst, how an application skeleton can be maintained across such administrative domains (ISP realms) in a “secure” manner will still be a challenge.

Weaving data within secure conversations

Weaving data within secure conversations in an acceptable and secure manner, or in other words, creating a legitimate man-in-the-middle scenario.

With enterprises opting for secure hosting options like HTTPS (Hyper Text Transmission Protocol – Secured), one possibility is to offload the SSL Layer (Secure Socket Layer) onto WAN optimizer equipment on the content server side to enable the elimination of information redundancy.

Addressable data chunks for parallel data acquisition

Enterprise deployments with mirror servers distributed across locations, the WAN Optimizer equipment can be configured to have a point to multi-point deployment (connecting a single client with multiple servers) and parallelize data acquisition (in a funneled way). Such configurations can be made possible by maintaining addressable data chunks.

New generation of applications and requirements for newer optimization/acceleration mechanisms:

Keeping an eye on emerging traffic patterns (evolved packet core at the telco SP, Android/iOS, Google and Apple), would help identify new protocol optimization opportunities (study on information redundancy, protocol behaviors causing transfer delays), customization based the environment (tuning the timers, window size, etc). This is a continuous R&D process to be performed.

Offloading/acceleration opportunities for WAN optimizers from the application developer context:

As a convenience measure, application developers just use a well known underlying protocol to avoid creating a new protocol (and related hassles) that can be efficient in terms of encoding. For custom applications, determining the optimum encoding option can be a daunting task for the developer community. A “codec factory” can absorb this risk and can suggest binary or textual encoding, length, redundant semantic checks, etc., based on the long-term inspection of exchanged data. This information can be used by application developers or can be used by WAN optimizer equipment for translating inefficient streams into optimized traffic. To make a long story short, we can learn from the application information exchanges, then build a suitable translation codec on the fly.

Conclusion

The above thoughts post a good platform to develop proof-of-concepts around the WAN optimization/acceleration technology domain. The entire domain, as such, keeps evolving with an ever-growing demand for applications across verticals and ever-growing open initiatives/communities throwing challenges toward the network.