Is your network VoIP ready?

Enterprises and equipment vendors are learning the value of a complete readiness assessment before deploying voice over internet protocol (VoIP) across an organisation. The assessments are a critical step to a successful VoIP deployment, but many enterprises do not perform an assessment because of cost or because, after performing a baseline on network utilisation, they assume that their network has enough bandwidth to accommodate the voice traffic. Fluke Networks’ Wayne Allen* reports on the importance of assessment and the consequences of not assessing the network correctly.

It is essential to perform a readiness assessment both for initial VoIP deployment and expansion projects to avoid unplanned additional costs and deployment delays. Over the past several years, enterprises have been deploying VoIP with more frequency and many have learned the hard way that the first step in the VoIP life cycle process should be a complete assessment of network readiness. Too often, IT organisations fall into the stereotype of “Let’s put it on the network and see what happens”.

As convergence grows, the ‘let’s see what happens’ approach with VoIP will almost certainly not be successful. While it is a data application at the core, VoIP does not act like traditional data applications. For most enterprises, VoIP will be the highest profile application on the infrastructure, meaning if the phones do not work, everyone will know.

With early VoIP deployments many enterprises had major issues with initial rollouts. One key reason was a lack of pre-deployment assessment. When there were initial problems, there was ‘finger-pointing’ between the enterprise, VoIP equipment vendors and service providers.

Today, most VoIP equipment vendors require network assessments to be done before they will sell or service the internet protocol (IP) private branch interchange (PBX) and phone systems. While this is a good first step, many enterprises still do not perform comprehensive assessments.

Pre-deployment assessments are the first step in a successful VoIP deployment, which must include monitoring, management and optimisation.

Approaches to pre-deployment assessments

Polling-based analysis

One approach uses quantifiable usage of VoIP to estimate readiness. For example, an enterprise plans to use the G.711 codec and sets up the quality requirements. Each VoIP call might require 64 Kbps of bandwidth. Therefore, if a location is expecting five concurrent calls maximum, there should be at least 320 Kbps of bandwidth available just for VoIP.

With this approach, a polling-based solution will focus on using the existing infrastructure and polling standard management information bases to trend utilisation on each device. With this information, the enterprise can attempt to calculate how much bandwidth is available and/or needed for each location.

Problems with this approach are the limited granularity of the information, the polling interval can skew and flatten the bursty nature of data applications and the traffic is not broken down by application so there is no correlation of performance.

Granular converged performance

Granular converged performance uses the quantifiable aspect of VoIP with a different tactic. Instead of polling and averaging usage every one to 15 minutes, this approach focuses on detailed monitoring of the network and provides up to a one-second level of granularity.

For example, instead of one data point every five minutes with a polling-based solution, one-second granularity provides 300 data points in the same five-minute period so peaks and valleys with bursty applications can be seen.

The benefit is the deep granularity it provides by actually monitoring and measuring every packet going across the network. A disadvantage with this approach is the requirement of a device or appliance at each location. It is impossible to collect this level of data and granularity without a dedicated appliance.

Simulated VoIP calls

With simulated VoIP call testing, traffic agents are used to generate simulated real-time transport protocol traffic and measure the performance across the network. This approach provides detailed results of mean opinion score (MOS), jitter, packet loss and latency or delay for each call. From these results, the maximum number of calls, above a user-defined MOS threshold, will be calculated.

The network performance is further validated to ensure deployment can support the VoIP requirements under normal and peak situations by generating the maximum number of calls at specified intervals to ensure the voice traffic is not affected by data applications.

The simulated calls need to be generated simultaneously from selected areas throughout the network and should also ensure that additional voice traffic does not affect the performance of existing applications. This process is typically run for two or more weeks and if there are no problems, the network is deemed ready and the assessment is over.

It is also necessary to consider how long it will take from assessment to full deployment and what changes could affect the organisation in the meantime such as new users, new applications, increased usage and new locations.

These variables highlight why a single assessment is not sufficient. If the calls are below quality metrics, then changes in infrastructure should be made and another assessment performed prior to deployment. Before making changes to the infrastructure, it is necessary to ascertain exactly what is contributing to the degradation in voice quality.

Without knowing whether the degradation is a result of jitter, delay or lost packets, it is impossible to determine where the changes need to be made. Consequently, enterprises require additional troubleshooting products to determine the root cause of the problem.

Other considerations for VoIP pre-deployment

In addition to performing an assessment to estimate the number of simultaneous calls and the expected call quality on the network, it is essential to assess the network topology to determine how devices are interconnected at both layer two and layer three and to verify virtual local area network (VLAN) configurations and device membership within those VLANs.

Infrastructure devices also need to be evaluated to ensure they are capable of handling voice traffic and are correctly configured, especially with regards to quality of service (QoS) policies or class of service (CoS) settings.

Other considerations include Power over Ethernet (PoE) support, adequate cooling for the devices and sufficient central processing unit and memory capacity in infrastructure devices.

Another major issue in today’s converged world is that many enterprises are combining applications and networks, but are not bringing the voice, data, application, IT and infrastructure groups together.

All too often the voice group needs to replace the legacy phone system and begins the migration to VoIP. The data or networking group may not be aware of impending change or the potential impact.

The consequences of ignoring pre-deployment assessments

Organisations that want a successful first-time deployment must find a way to not only do an initial assessment for VoIP, but find a cost-effective and user-friendly way to continuously monitor readiness for VoIP in the bursty and high-demand converged networking area.

Since VoIP deployments typically occur over a period of months or quarters, enterprises must consider how to carry out dynamic and on-demand assessments without raising the cost several times.

A solution to meet VoIP assessment needs

Whether planning a new VoIP deployment or an expansion to an existing deployment, an organisation must determine:

• whether its LAN and wireless area network (WAN) circuits can handle the increased traffic from the VoIP deployment
• that QoS policies are correctly implemented
• the MOS call quality expected
• the major factor affecting voice call quality such as delay, packet loss, jitter or Codec
• the number of concurrent VoIP calls the network can handle without impacting existing application performance

The success of VoIP deployment relies on many individual components and when one is not performing efficiently call quality will be degraded. The key to having a successful voice deployment is to build a strong foundation across the distributed enterprise.

VoIP will raise the level of attention of poor performance because most users will never notice or complain if an email takes 10 extra seconds to send or web applications are a little slow.

With VoIP, users will immediately notice degradation and most likely complain, which will increase the exposure and risk to your organisation. This exposure adds to the criticality of ongoing performance management for both voice and data applications.

* Wayne Allen is the program manager - Australia and New Zealand, Fluke Networks. He is responsible for sales and technical support for Fluke Networks products within the region.

Allen has been associated with Fluke and Fluke Networks products for the past 20 years - initially with Philips Scientific and Industrial Electronics, the then agents for Fluke in Australia and New Zealand, and more recently with Fluke Australia.

Prior to joining Philips Scientific in 1989, Allen spent 12 years with the Department of Defence in various technical roles as a specialist in military communications.