How to Manage and Optimise an IP-PBX Service

In order to get the best out of VoIP whether it be from an on-premises or a hosted IP/PBX it is important to monitor and manage the network to ensure optimal performance. The network can be logically divided into the external network which is the WAN connection to the internet and the LAN which is the internal Ethernet network inside the firewalls and directly under your physical and administrative control. In this first part we will consider how we can manage and optimize the LAN.


Monitoring the LAN

The techniques we will need to monitor the LAN are dependent on the size of your companies LAN. Typically small medium businesses (SMB) will have a LAN that consists of only one or two subnets in a single office. A Large company however may have several subnets in a main office and several sub offices. The theory and the techniques we will use will be the same but for a large network you really should ease the burden by investing in a professional network management system.

There are several KPI’s (key performance indicators) we should track when monitoring the LAN. They are bandwidth utilization, latency, jitter, and packet loss. These are the most common indicators that something is wrong somewhere on the LAN. First of all however we need to explain how a LAN works, for only then can you understand what and then how we can monitor it.

A LAN will consist in its simplest form of a few computers connected by local Ethernet cables  (do not use Wi-Fi for voice) to a local switch. The switch will then be connected to an ADSL modem for an internet connection. If we are using physical IP phones we can connect the phone and the computer sharing the same desk location to a single cable going back to the switch. This is the typical design and the IP phone will have two RJ45 ports; one for the cable to the switch (the network) and another RJ45 port for attaching the laptop or computer.

When we have the computer and phone sharing the one network cable this reduces the number of ports required on the network switch. It also reduces the number of cables we have to run from the switch cupboard to the desks; and the fewer to keep tidy. However by sharing the same physical cable we are introducing potential conflicts between the data and VoIP bandwidth requirements. A common scenario would be if the PC got infected with a virus that started chucking out vast quantities of pings scanning the subnets this would have a detrimental effect on the voice quality.

In order to mitigate this potential conflict we need to segregate the data and voice traffic in different Vlans. On a physical phone this is not difficult we can just configure all the phones in the LAN to be on a different Vlan, for example configure Vlan 11 for VoIP, and just leave the data in the native Vlan 0. This will now segregate and logically isolate the traffic but there will still be a problem as the packets from each networks, data and voice, are still physically sharing the same cable, switch port, and the physical queues within the switch port. Therefore if there is flooding on the data network it will still swamp the physical queues causing voice as well as data packets to be dropped. In order to prevent this happening we have to implement Quality of Service.


Quality of Service (QoS)

QoS can be configured on the switches to tag voice packets with a higher priority. The way this works is that voice packets, just like data packets are carried inside Ethernet frames, which have along with the source and destination MAC addresses of the sender and the recipient a byte which determines the priority. The byte has three bits allocated for priority, which corresponds to eight possible settings, 0 – 7 with 0 being the lowest and seven the highest priority. However level 6 and 7 are reserved so in actual fact 0 is considered best effort and 5 the highest priority. This will give voice packets priority over data should the queues in the switches become congested, which is what we want.


Bandwidth Contention

Running data and voice over the same cable and switch port reduces hardware costs significantly and configuring Vlans and QoS can isolate and prioritize the different traffic classes. However we still have the physical bandwidth restrictions of the port to contend with. For example, if it is fast Ethernet Port at max 100Mbps, then no amount of Vlans are going to magically change that.

This is rarely likely to become congested except under a fault condition as 100 Mbps is a lot of bandwidth. However we do need to monitor and check our bandwidth utilization.  We also need to monitor and check the network delay, jitter and packet loss as these are the most common symptoms of network problems that can degrade voice call quality.


Checking the Key Performance Indicator’s

In order to be able to check what is happening on the wire we need to be able to either access the switch directly and then interrogate the switch management statistics for things like the number of packets dropped or the average delay, or download and install a copy of PRTG. This is a free network management system (NMS) tool which will automate statistic collection and graphically display the KPI’s we wish to monitor – making life much easier. For larger networks you may need to upgrade to the paid version or consider deploying a feature rich NMS like Solarwinds. Whatever, the main point is that PRTG and Solarwinds are simply data collectors and produce graphical displays, all the statistics they use is available on the switch and can be accessed directly if needs be.


VoIP Metrics

Maintaining high quality voice can be difficult as VoIP is sensitive to fluctuating network conditions.


Network Jitter and Delay

Delay or latency is simply the time it takes a packet to traverse the network from sender to receiver. Some networks have very low latency, fiber optic core networks for example, others have very long latency in the region of 15 seconds in the case of some satellite links. For satisfactory voice quality latency should for voice be less than 150ms one way. In practice latency is not a real show stopper so long as it is consistent. Jitter however is the variation in latency which can cause much more acute problems. Jitter is best described as the difference in response times that you see when running a ping test. It is that difference between echo replies that corresponds to jitter. On a LAN jitter should be very low, on the internet it can be very high. Jitter will begin to affect voice quality at > 30ms.


Packet data loss

Packet loss is the number of packets that fail to reach their destination. A single packet lost is known as a gap, a string of packets lost, a burst. Packet loss happens on a LAN for a variety of reasons, such as congestion, misrouted packets, flapping links due to poor cable connections, and mismatched duplex/half duplex or other port capabilities, being the most common.

By monitoring the status and the traffic characteristics of the LAN using a network management tool such as PRTG we can pro-actively react to anomalies in the KPI’s. Monitoring network conditions allows us to optimize the LAN to make it resistant to fluctuations in the network that would degrade the quality of the voice calls.

In part two we will see how to manage and optimize the Internet WAN link.



Leave a Comment

Your email address will not be published. Required fields are marked *