discuss Unified Communications

Q&A with Sean Armstrong, Director of Product Management, Apparent Networks www.apparenttools.com

Download the latest whitepaper from Apprent Networks, Best Practices for voip Network Assurance

Why is it important to conduct a network assessment prior to VoIP deployment?

We actually just conducted a survey of network engineers about the applications that stress their network the most, and 67% of respondents said that VoIP stresses their network more than any other application. Much of this has to do with VoIP's sensitivity to jitter and packet loss, as well as the performance requirements it places on networks.  A small change in network performance can have a big impact on the usability of VoIP. Therefore it is critical that network managers assess their network's performance prior to a VoIP rollout.  In 2008, Gartner analysts cautioned that 75 percent of enterprises that do not perform a pre-implementation analysis of their IP network infrastructure will not achieve a successful VoIP implementation.

So why don't most network administrators conduct a pre-deployment assessment?

Based on conversations we've had with network administrators, most do not do a comprehensive VoIP pre-deployment assessment. While VoIP vendors recommend this as part of their deployment process, most vendors do not supply or recommend tools; and network administrators often do not have the tools to adequately perform one.  While their existing tools might provide great insight into the operations of devices in their network they lack the end-to-end performance view that is critical to understanding whether a VoIP deployment will be successful. And often VoIP traffic flows through third-party networks like carriers, further limiting their ability to measure end-to-end performance.

Some VoIP vendors and resellers will do a one-time assessment for a fee which can provide data and insight prior to deployment. But, this is a one time view of the network's performance and after installation users are often on their own. Considering limited budgets, network administrators need a more cost-effective option that provides ongoing value.

Isn't a network bandwidth test sufficient?  What if network administrators know they have plenty of bandwidth to support VoIP?

Network administrators should remember that bandwidth is only one of many factors in VoIP performance. This is why just throwing additional bandwidth to solve performance problems doesn't work. They should pro-actively measure all of the factors that can negatively impact voice quality - like network utilization, latency, jitter and packet loss.

They should also take measurements over an extended period of time. Traffic from voice and data will vary depending on the day and week so running the assessment regularly is as important as the upfront assessment.

Also, test the capabilities of the network to support end-to-end QoS. The inability to guarantee end-to-end QoS was actually listed as one of the top problems network administrators face when deploying applications like VoIP. QoS mismatches between WAN and MPLS network segments can severely degrade call quality and is extremely difficult to diagnose.

What types of things should network administrators look for when they assess their network for VoIP?

The objective should be to look at the network from the perspective of the VoIP application.  This means viewing the whole network in a comprehensive way to assure low latency, jitter, and packet loss while providing the bandwidth required by the VoIP calls. And, it's reasonable to achieve this by testing and assessing your network for the following:

• Is this LAN & WAN capable of supporting VoIP? What portions are? Which aren't?
• Are the service level agreements being met by the service providers? Is the service level sufficient to meet needs?
• What configuration faults need to be fixed?
• What components need upgrading?
• What is the call load capability?
• What is the best and worst call quality given a variety of conditions?

OK, so how do network administrators know if the assessment was effective?

There are a few questions that network administrators should be able to answer after their assessment to assure they are all set for their VoIP rollout: 1) What is the call load capability of the network?

2)  What is the overall network's ability to handle data and voice traffic?

3) What are the key faults and shortcomings of the network to delivering VoIP?

4) Are there equipment or configuration changes necessary to properly deliver VoIP?

5) Are my service providers SLAs being met?

Once a VoIP installation is complete, are there other considerations?

There are three areas to focus on once a VoIP or UC application has been deployed:  post-deployment impact assessment, ongoing monitoring, and troubleshooting.

A post-deployment assessment is important to understand the actual impact of the VoIP application on the network.  Is VoIP performing properly?  Are other applications being impacted? Are periodic application events like network-based backups affecting VoIP performance? Ideally, the pre-deployment assessment and planning increased the likelihood of a successful VoIP rollout.

Next is ongoing monitoring of the network and VoIP performance. This is especially important when there is any kind of network change. We know that changes in a network environment - either internal or external - disrupt the regular traffic and flow which can lead to faults and problems. By monitoring this activity and doing regular change assessments, network managers can be aware of network "health" and also foresee and mitigate issues before their users are impacted.

The last area to focus on is troubleshooting. We all know that issues will arise. So what happens when calls are disrupted? With most applications, a network performance problem just means slower application response times. With VoIP, a network problem may mean that VoIP is unusable. Because of this, network managers must be able to find and fix problems in a much more timely fashion.  Having the proper diagnostic tools in place is even more critical with latency-sensitive applications like VoIP.