More than often, IT and network sizing remain based on an intuitive mix of facts (measures), expectations (what should be the traffic in a few months) and a few drops of magic sauce (rule of thumb). It is actually a difficult task, as IT managers don't want to be short on resources, can't justify the budget for unused infrastructure and lack of a clear relation between sizing, quality of experience (QoE) and business impact for their company.
Looking to the network, the usual way to decide about the size of links is typically to ensure that congestions will not occur: the network manager looks to the traffic at peak time of a business day (say 11 am or 3 pm), add a security factor of its own (20% to 50%) and order for the next available bandwidth her/his Telco can provide. The outcome is acceptable in term of performance (there is generally no undersizing) but is not cost efficient, as the lines will be mostly oversized. Moreover this decision is always hard to justify to the management and the Business Units that pay for it.
Let's note that this decision process is more and more dangerous to apply to a traffic the critical business portion (~15%) of which is much lower that the recreational one (~50%) in most of today organizations. It finally consists to open doors without giving a sense of responsibility to users, encourage bad habits and set unreasonable expectations.
WAN governance proposes to shift from this empirical method in favor of a business driven logic. Beforehand, let's examine briefly two ideas generally accepted (even by experts):
- Conventional Wisdom #1: good network sizing will prevent congestion to happen à FALSE! And this is BAD news. TCP flows are elastic; their 'mission' is to fill the pipe up to the point where packets drop. Servers (and their Windows or Linux TCP stacks) are more powerful than ever and can burst very quickly. Applications and GUI get richer and embed heavy content. Cannibals (bandwidth greedy applications like YouTube, Facebook and all sorts of peer-to-peers applications) are on the prowl.
- Conventional Wisdom #2: network congestion means performance issues à FALSE! And this is GOOD news. Modern traffic management solutions, able to understand applications and to manage individually each flow according to its nature and behavior, are able to guarantee good performance even in the most congested situations.
The consequences are that congestion does not mean restriction: it is a normal situation where the enterprise can use the totality of the resources it pays for, while getting the right performance its business applications deserve.
How to conciliate Oliver's strength with Stan's format? What are the right criteria to rightsize network accesses? How to get sure that rightsized networks provide users with the required level of performance?
The essay writing just about this topic, I would find at the writing services. Buy the pre written essays and custom essay just about this topic.
Posted by: Konnie19cY | 01/08/2010 at 05:18 PM