Friday, January 14, 2011

WAN performance management, mission impossible? [part 2]

Application visibility

A sound approach is to start with an assessment of your IT infrastructure by building a baseline of your application flows crossing the corporate wide area network. This results in application visibility. In many cases this baseline discloses quickly where on the (highway) connection traffic is building up, when flows of traffic are congested, how often this occurs and ho

w sever the perceived performance pain is and if it’s an isolated event or a trend.

A baseline in essence is a “picture” of an hourly, daily, weekly and monthly busi

ness period showing the utilization of IT infrastructure. The performance data must be actionable data, i.e. data that enables the company to make decisions upon.

An example of actionable data in layman’s terms is shown in the picture here. Detailed insi

ght into the application data that flows across the available data connections between corporate offices.

Shown here is that the data connection is used for various types of corporate user

s, applications and services. It also shows that at

several instances the

connection was claimed for 100% by a file transfer application an

d later on by a disk backup process (storage layer). This task also delayed other business applications in delivering data to and from the offices.

When measured with appropriate technology, the graphical output could be something like the graph shown here.

A so-called ‘brown-out’ on your infrastructure, a traffic jam, without any certainty of when it will resolve, resulting in end-user performance degradation, loss of good-will as IT department towards the business and loss of business opportunity and loss of revenues.

The baseline resulting in the application insight means also access to actionable performance information. Events like these, whereby uncontrolled application usage claims corporate resources at any given time, cannot be tolerated in any corporate infrastructure.

Results of such a baseline could already pin point to potential root causes of the problem, such as:

  • · Simply not enough bandwidth between the clients and application server



  • · So called “chattiness” of applications whereby many packets between client and server create a completed transaction.


  • · The distance between the client and server is so long that the data packets between the clientPC and the server simply take very long. This is called high latency.



[In part 3 I will review a number of senario's to resolve these issues using WAN Optimization and Acceleration techniques]

Friday, December 31, 2010

WAN performance management, mission impossible? [part 1]




Imagine any corporation with a public Internet connectivity or connections to offices across Canada connected by a private data communication (network) infrastructure, using WAN links, MPLS or Internet with or without VPN tunnels.

This could be the application performance as perceived by your end-user: “Slow, unpredictable, no one knows what’s going on.”

Whereas the IT department believes that the quality of service delivered is spot-on: “all lights are on green, we don’t have any problems, all systems are up”.

Not too difficult to spot the differences between the two pictures. Please allow me to depict some scenarios that could have led to this situation.

· The IT department responsible for the “highway” connections between the two corporate locations assumed that a connection with 4 traffic lanes was sufficient.

· The old infrastructure was, according to the usage reports of the service provider, only used for 50%. When the contract renewal was due, the service contract manager ordered half of the bandwidth.

· The application support team was full of confidence that the server system, serving the local users, could also service the remote office users on “the other side” of the network connection, without having tested the network-friendliness of the application.

· The business unit Marketing launched a new customer contact campaign requiring the Marketing team in the remote office to interact on-line with web-based customers through the corporate Internet gateway.

· The EDP auditor, as part of the annual risk auditing, advised to have all data residing on every corporate server, to be backed-up on a central, secure, storage system.

· The corporate ERP application had to be implemented before year-end, end-user testing was limited to a functional test.


The end result of one or more of these decisions: chaos on the highway between the decentralized offices and the corporate office annex data center.

Monitoring and management of these individual technology silo’s such as databases, web servers, networks, storage, applications, etc. does not guarantee that the end-user gets deliver what is agreed upon. Nor what the perception or expectation of the end-user is. Each “silo manager” may conclude that he (or she) delivers in accordance to the agree terms, however the end-user gets confronted with all the individual events and hick-ups in the application chain.

Each discipline oversees in its own technology or functional silo. It is at the end-user where the performance pain is felt. Not surprisingly is that studies show that more than 70% of all performance problems are reported by the end-user.

Time to analyze this phenomenon. The key lies in actionable performance information. Only with actionable performance information organizations are able to tackle application and end-user performance problems and challenges.

Creating this performance information starts with gathering performance data. This sounds a trivial conclusion, but it is far from that. The challenge starts with choosing a technology (or two) to obtain the required data. So, where to start?

[more in part 2, to be published soon]