As enterprises continue to adopt SaaS applications and migrate to cloud-based VoIP services, the dependency on the Internet, a public “best effort” network, has increased. In today’s modern hyper dependent business environment, IT teams need to be able to analyse application performance and service delivery across all infrastructures, including ones they don’t own or directly manage.

In the era of the cloud, application monitoring and network visibility should not be an afterthought – especially when it comes to maintaining a reliable Unified Communications as a Service (UCaaS) solution. Enterprises migrating to UCaaS should follow these best practices to maintain a successful voice strategy before, during and after deployment.

Baseline & Validate Before Deployment

To work seamlessly, cloud-based UC solutions need to be optimised for your organisation and network. Network bandwidth, latency and packet loss can significantly vary from different parts of your enterprise network and directly affect user-experience. For example, bandwidth constraints, connection reliability or network configurations could impact some branch offices and not others. Understanding these scenario’s prior to deployment can help you proactively plan and assure reliable service. 

Traditional VoIP monitoring techniques like Call Detail Records (CDRs) and packet captures, while helpful in troubleshooting post-deployment, do not address pre-deployment benchmarking scenarios. They also provide limited visibility into VoIP performance outside of the enterprise network.

Consider a proactive approach to voice monitoring by actively monitoring and simulating VoIP calls within your network. As you build a continuous baseline of performance, active monitoring can provide insights into the health, availability, and quality of VoIP applications, across both internal and external networks. 

Approaches To Monitoring Performance

When it comes to monitoring the quality of a voice call, Mean Opinion Score (MOS) is the best indicator. However, MOS is not representative of the entire lifecycle of a VoIP call. The initial call establishment or Session Initiation Protocol (SIP) signalling is equally important to monitor. While degraded voice quality can lead to poor end-user experience, not being able to even make a call can be equally frustrating.

Consider monitoring SIP transactions along with Real-time Transport Protocol (RTP) quality for an effective VoIP monitoring strategy. Also, as the network path to a SIP server is independent to that of the RTP stream, understanding the network topology in each of these stages of call initiation and support will provide contextual data for troubleshooting VoIP.

Troubleshooting VoIP Quality Issues

As an application, VoIP is notorious for falling victim to underlying network inconsistencies. Packet loss and latency in the network can significantly affect voice quality, and service degradation is immediately recognised.

For example, Differentiated Services Code Point (DSCP) remarkings can induce network delays and increase latency that can severely affect voice quality. Used within a network, DSCP values are based on the operator’s own Quality of Service (QoS) policies. Because each operator sets its own QoS policies, DSCP values often change at borders between networks. With UCaaS deployments, this effect is enhanced as voice traffic transits multiple third-party ISP networks.

DNS misconfigurations, lossy interfaces or misbehaving network devices could also impact service quality. When voice quality dips, it is critical to understand how the network could have impacted the service degradation. Maintain visibility not just into MOS scores, but also the underlying network while monitoring and troubleshooting VoIP.

Collaborate To Reduce MTTR

Knowing exactly where a problem exists and the related cause analysis empowers teams to take immediate action. Whether the problem is internal to your network or external, chances are you need to work together with multiple internal teams, ISPs and UCaaS vendors. Collaborate by sharing diagnostic data with vendors or customers. When everyone sees and interacts with the same data, the mean time to restore is dramatically shortened.