There may be different type of packets traversing through the WAN links of an organization. As data centres are consolidated to a central location, the WAN links almost become an extension of LAN and as a network is only as fast as its slowest link, it becomes more and more important to optimize the performance of a WAN link. One of the main parameters to optimize in a WAN link is the bandwidth. Whatever is the available bandwidth, there are always some chatty non-critical applications that choke WAN links. There are always some critical/real-time applications that suffer because of the bandwidth occupied by the chatty applications. So, there is a need to identify such chatty/non-critical applications and optimize their performance over the WAN links that connect an organization and its data centres. The following are some of the steps followed by bandwidth management/optimization vendors to achieve that.
Traffic can be categorized based on a lot of criterion. The most popular methodology is to categorize traffic based on application. But they can also be categorized based on other factors like port number, protocol, URL, MPLS labels, IP or MAC address, subnet, VLAN, database type, source, destination etc. Eg. ERP traffic to and from a specific server, traffic with an MPLS label of 8, traffic from specific port no(like 80) etc. In certain cases the categorization needs to go one step ahead and analyse the various sub-categories of a specific category as well. For example, in port 80 there may be more than one application running. And each application needs to have a separate categorization in order to observe their characteristics and optimize them later on.
The parameters required for bandwidth categorization is quite important based on the type of usage of the WAN links in a particular organization. Normally categorization based on applications is considered quite sufficient, and if not, there might be some additional categorizations based on the above parameters too.
Once the WAN traffic is categorized and the parameters for them defined, there are various metrics that can be employed to analyse the bandwidth consumed. This analysis would be used later on to optimize the bandwidth across WAN links. So, some of the parameters employed by various vendors to analyse bandwidth sent across WAN links are given below:
¤ Bandwidth usage levels per ip address, host or subnet.
¤ Top few Applications that are generating the bulk of the traffic.
¤ Bandwidth wasted due to retransmissions, dropped packets, aborted TCP connections by the server etc.
¤ Applications/ Users/ Servers that get the most traffic and send the most traffic.
¤ Identification of clients, network and server with the slowest/ fastest response times.
¤ Connection speeds and packet size histogram data.
¤ Largest and lowest number of simultaneous TCP connections at any given point of time.
Those were some of the performance monitoring parameters that can be tracked in order to optimize the WAN bandwidth. Some vendors might also give a graphical representation of various parameters for easy analysis either for the whole link or a portion of the link like a particular server/ application or URL etc.
Based on the traffic analysis, some of the following steps could be taken (depending on the bandwidth optimization vendor) to ensure that the available bandwidth is optimized across the WAN link.
¤ Virtual Links: The WAN links could be subdivided in to smaller partitions called virtual links. Each virtual link could be restricted to a certain application, user group etc. The advantage of sub-dividing the available links is to ensure that each application gets some minimum guaranteed bandwidth and that one application or user group does not occupy the entire/large portion of bandwidth available with the WAN link. Unused bandwidth is always available to others. Some vendors also allow dynamic creation of virtual links based on certain events like real time applications requiring more bandwidth suddenly, allocation of more bandwidth per user when the ERP applications are using less than a certain bandwidth threshold, etc.
¤ Policies: It is important to set priority levels to each type of application/ user group etc(In a scale from 0 to 10, for example) in order to make sure that mission critical traffic types like ERP, voice, video and others get a higher share of bandwidth than peer to peer applications, music downloads etc. which need to get a minimum share of bandwidth as long as the other critical applications are active on the WAN link. There could also be a limit to the minimum and maximum bandwidth that an application could use, for example. Certain type of applications could also be blocked by giving a very limited bandwidth or zero bandwidth (personal applications like youtube videos, movie streaming etc). Some applications (like emails etc.) are not set any priority level and go by the default average priority. Certain vendors allow automatic creation of policy tables when the most important applications/ user groups are classified and assigned a certain priority based on their classification and analysis.
¤ TCP Rate Control/ Admission Control: When TCP packets arrive at a faster rate than that can be processed, these bandwidth optimization technologies could apply TCP rate control to the WAN links and inform the sending stations to send packets at a slower rate as excessive TCP packets need to be discarded and re-requested. Certain vendors also allow to manage bandwidth by measuring the speed of such TCP transmissions and adjusting certain parameters in the receiving end. Admission controls involve creation of policies to handle the additional sessions during a bandwidth shortage like blocking them or reducing the bandwidth allocated to all other/ certain applications by a uniform rate to accommodate these additional sessions etc.
You can stay up to date on the various computer networking technologies by subscribing to this blog with your email address in the sidebar box that says ‘Get email updates when new articles are published’