Today`s networks increasingly incorporate a voice or telephony component - commonly known as voice over Internet protocol (VOIP) - in a bid to save on high telephony costs through least-cost-routing and other options now available to companies in SA.
To implement VOIP successfully, all components, including the network wiring or cabling infrastructure, have to be optimised.
To be intelligible, voice transmissions over the network require a constant stream of packets. This is unlike data transmissions that can accumulate packets and send them in bursts - because only the content (and not voice quality) counts.
In other words, to maintain a reasonable quality of conversation, the Internet protocol (IP) voice packets must arrive at their destination in the correct order and more importantly, they cannot take too long to do so or conversation clashes will occur.
Delays
It`s important, therefore, to minimise the small delays that are inherent in every switch, router, traffic shaper, firewall and hub.
Major causes of transmission delays which can impact voice quality on the network are "packetisation" delays associated with the time taken to convert an analogue signal into a digital signal and vice versa.
Jitter buffer
Another big cause of hold-ups is "jitter buffer" delay, which is linked to the time taken to queue data packets inside a jitter buffer. This is the component tasked with collecting and collating packets arriving at irregular times, ensuring they are in the right order, and then sending a smooth stream to the listener, thus ensuring a good quality conversation.
The cabling infrastructure is not easily replaced, so it`s vitally important that cabling is installed correctly and tested.
Graham Vorster, chief technology officer, Westcon AME
Delays often propagate "bit errors" which result in IP voice (and data) packets being lost or discarded which in turn leads to quality of service (QOS) issues and listening quality problems.
In fact, bit errors are one of the major causes of QOS problems on any application-aware network infrastructure. Because of the real-time nature of many time-sensitive applications, such as video, lost data is never recovered.
For these applications, the luxury of several re-transmissions via TCP applications is not available as it is for computer data transfers.
Role of cabling
Bit errors are most often introduced into the system through faulty equipment, incorrectly installed structured cabling systems, mismatched cabling components and patch cords, and by external "noise" sources.
After the network system is installed, the faulty equipment causing bit errors can be easily replaced or repaired and external noise sources can usually be traced and eliminated.
But the cabling infrastructure is not easily replaced, so it is vitally important cabling is installed correctly and tested to ensure there are no situations where the physical cabling is likely to present problems.
Tests
When installing structured cabling systems, one of the first considerations should be the quality of the cable infrastructure. Poor grade cable, cheap connectors and patch cords as well as inadequately designed installation pathways and less-than-professional installation practices used can each become a hidden bandwidth killer.
Fundamentally, for every cable related problem there are significantly more TCP/IP problems created - which will impact all applications running on the network.
Another consideration should be the amount of testing and measurement to which the entire infrastructure is subjected.
Tests for bit error rates, bandwidth reservation, application ID (differentiation) and Web application interference (eg Web radio streaming) need to be completed.
The amount of overhead demanded by the various applications, the volume of management traffic and the amount of broadcast traffic received (including unauthorised radio services!), need to be documented and evaluated.
Successfully installed?
When the cabling contractor requests sign off of a "successfully installed" network, what really is being asked of the purchaser?
A "successful" installation should warrant more than mere operational functionality. It should certify that the network is operating up to its maximum potential.
Cabling contractors usually certify their installations independently of the network. Due to this, performance criteria are difficult to evaluate. It`s almost impossible to determine when or where errors will occur - or whether data packets will be lost as a result.
To realise the true impact of the quality and integrity of a structured wiring installation, it is necessary to use testing procedures that will address a live environment.
Hand-held test units are not up to the task. What is required is a device with the ability to locate online problems, identify weak network interface cards, marginal connectors and borderline patch cables. More importantly, the solutions must be able to evaluate network throughput, taking into account collisions and data errors.
Quality is key
Finally, from the voice perspective, it is key to evaluate quality against a number of criteria. There are tests available, although, a number have only been proven in laboratory settings.
As time progresses, more credibility will be given to these tests, and the infrastructure specialists will be held accountable, based on their findings.
They include:
* PSQM - Perceptual Speech Quality Measure (ITU P.861)
* MNB - Measuring Normalised Blocks (ITU P.861)
* PESQ - Perceptual Evaluation of Speed Quality (ITU P.862)
* PAMS - Perceptual Analyses Measurement System (British Telecom)
* E-Model - A computational model for use in transmission planning (ITU-T G.107)
To maximise network throughput, the impact of structured cabling must be considered. Failure to do so can have serious consequences.
Most likely, hidden bandwidth killers lurk on many operational networks today - without their owners being aware of how much potential is being lost and how much it is costing them in largely unnecessary bandwidth optimisation and data compression solutions.
Share