Mike Bailey of Ipanema tells DCM about the options for wide area network (WAN) operations

As business reliance on data centres increases, so too does the reliance on wide area networks (WANs) to connect the operational hub with offices or outlets around the country, or even the world. With a considerable percentage of enterprise employees based in branches (40% - 60% depending on the industry and size of the company), and companies reducing the numbers of local servers for security and cost reasons, the WAN now plays a crucial role in enabling access to business-critical applications. However, the bottlenecks and constraints of the WAN will often impact application performance, and this has the potential to seriously damage the productivity of the business as a whole. As such, there is a great deal of pressure on network managers to guarantee the performance of the network and the delivery of key applications.

Examining the options

In approaching the challenge of WAN optimisation, network managers need to examine the main root causes of poor application performance over the WAN: insufficient bandwidth, distance, and fast growing chatty recreational applications, dynamic IT environment, etc. These issues result in packet loss, high delay and jitter, all of which deteriorate application performance. Insufficient bandwidth can be addressed either by increasing the overall bandwidth available or by better allocating existing bandwidth. Network managers can increase available bandwidth either directly by buying fatter pipes, or by applying compression technology.

You can better allocate existing bandwidth through Quality of Service (QoS) techniques and bandwidth allocation management technologies. Unfortunately most of these techniques are static, with the exception of dynamic bandwidth allocation. Dynamic bandwidth allocation starts with the performance expected for each application, then analyses the traffic mix in real time and automatically matches the required performance with the available bandwidth. It is the most efficient way to handle micro-congestions and sudden surges in the number of users without massively over-provisioning.

Applications that need to travel long distances over the WAN are the second cause of poor performance. Network managers can address the latency issue caused by long distances and ‘over-chattiness’ of some protocols or applications in two ways: either by reducing the network delay itself, or by reducing the dependencies of applications on delay. Nothing can be done to the ‘speed of light’ limitation; however, router-induced delays can be reduced through intelligent packet forwarding techniques. One solution consists of forwarding packets looking simultaneously at the nature of the flow – i.e. data transfer, transactional or real time – and its business criticality. In this manner, business-critical applications that are more sensitive to network delay are forwarded first.

Network managers can reduce application-dependency delays either by reducing the time to execute a turn or by reducing the number of turns. TCP acceleration technology and local acknowledgment techniques are useful for reducing a turn’s time. The number of turns for overly chatty and widely used protocols such as Common Internet File System (CIFS) and Exchange can be decreased through application-dependent protocol optimisation techniques. Finally caching also contributes to reducing the number of turns.

None of these solutions are without drawbacks or solve all issues. In fact, network managers might use a different technology to resolve each issue that impacts application performance. For example, a network manager may deploy a compression solution, but this will only have a significant impact on data-transfer applications. Similarly, application-dependent optimisations will have an impact solely on the application they are built for; and QoS techniques work best for protecting VoIP.

Finding the right WAN

The problem with using different optimisation technologies to accelerate applications is that it can lead to end-users deploying ever-more recreational applications, which in turn tend to grow their resource demands faster than critical application traffic. This then harms the performance of critical applications and leads to a need for ever more bandwidth – or to an ever-increasing array of single-purpose optimisation tools. It is the start of a vicious cycle.

The only way out of this trap is to discriminate among applications and ensure the performance of critical applications under all circumstances. After all, no network manager will lose sleep over the fact that downloading the latest Madonna clip is slow; however, he or she might worry if that clip's torrent harms the performance of the end-of-the-month SAP data consolidation and it doesn’t perform to the CFO’s liking. In this way, WAN optimisation should be a function that is aligned with larger business goals to protect business-critical applications, rather than all traffic. Attempting to accelerate all applications in an indiscriminate fashion is the pursuit of an impossible, costly and misguided goal.

If network managers wish to satisfy end-users, it is vital that they not only prioritise traffic based on importance to the company, but that the performance of business-critical applications is guaranteed under all circumstances. As such, it is vital that network managers take an approach that recognises and discriminates between different applications and traffic types, and can respond to changing demands on the network in real-time.

The Ipanema System

The Ipanema system offers data centre managers the opportunity to guarantee the performance of business critical applications through three main functionalities: Visibility, Optimisation and Right-sizing.

Visibility enables full monitoring, reporting and trouble-shooting on the WAN. As well as measuring Round-Trip Time (RTT), the solution can deliver one-way metrics that indicate the exact contribution of the network to application performance. All packets, including both performance and volume information for each flow, are measured, and the metrics are collected and aggregated by the central management software. There are dedicated reports that allow users to accurately monitor the network service delivered by the telecommunications provider. This feature makes it possible to understand, with no ambiguity, the exact makeup of the network to end-user application performance.

The second function of the system, WAN optimisation driven by application-performance objectives, enables network managers to guarantee the performance of business-critical applications under all circumstances. Optimisation features are based on global performance objectives that define what resources the network should deliver to provide a good quality of experience to every user of each application. The system dynamically calculates optimal local policies, enabling the meeting of the objectives, in each device, based upon a real-time and continuous analysis of the available network resources and user demand.

The system’s optimisation capabilities are based on co-operation between devices, and as such it handles meshed flows, providing total control over quality, without having to statically reserve bandwidth. The solution can utilise tele-optimisation to control remote TCP meshed flows with devices only in the data centres and none in the branch offices. When equipment is placed in both data centres and branches the system is able to control meshed flows using any protocol, including UDP. This means that it is possible to provide optimisation of non-hub-and-spoke networks running any client-server applications, using a lower number of devices compared with other products.

The final function of Ipanema’s system is an ability to right-size the WAN according to application performance objectives, controlling expenditure on bandwidth. Essentially, the solution calculates the relationship between available bandwidth and application service levels, and automatically responds, preventing unnecessary over-provisioning.