The challenge of implementing UC in your network

Implementing a Unified Communication service like Microsoft Lync (now Skype for Business) seems, from an application point of view, quite simple. Take care of basic network connectivity, install a couple of Windows Lync servers using the step-by-step instructions you can find anywhere on the web. Setup some SIP trunks with your legacy PBX to enable the voice breakout part and deploy the client software, link and import the users from your Active Directory and here you go! Is it really that simple?

Well, from an installation point of view it actually is. Microsoft offers numerous tools like requirements questionnaires, planning and design wizards, implementation methodologies and templates to help plan, design and implement a Lync server topology. Using these tools, any smart Microsoft administrator will be able to setup a server farm with the required Lync server roles and hardware that matches the needs of your organisation.

However, an implementation of a server farm by your application staff that doesn't take an in-depth look at your network infrastructure won't guarantee a smooth adoption by your end-users. It will become even more challenging when the video and desktop or application sharing workloads are enabled to your users and the network is stressed to its limits.

Although your Lync implementation project might appear to have been completed within time and budget, your operations teams could face hidden costs and a variety of issues in the future.

This is because most data communications traffic on your network is not real-time sensitive or compensated for bad connections as it uses the TCP protocol which will send the packet again if it has not arrived intact. This leaves minor issues in the network unnoticed by your end-users. Media streams that carry voice and video data in UC applications are different. These streams need end-to-end Real Time Communications (RTC) settings and protocols like UDP in order to perform. The UDP protocol does not provide a check on arrived packets. There is no room for error whatsoever; lost packets are really lost and will not be send again which will degrade sound and video quality - something that will be immediately noticed by your end users.

A successful handover to your end-users is highly dependent on their perception and user-experience, which both greatly influence their willingness to adopt and use the service.

When you replace your “good old” and reliable PBX systems and legacy telephony services with, for example Microsoft Lync Enterprise Voice, you may struggle to retain the confidence of your end-users. Even when call and sound quality and reliability are technically the same or even better, every minor incident will be critically examined under a magnifying glass!

In many big organisations the IT governance for applications, networks and voice telecommunications are in different entities that don't work well together and even have their own budgets and objectives. Also many separate suppliers are involved which makes end-to-end monitoring, operations and fault finding a challenge. The successful introduction of Unified Communication services in your company needs these entities to work closely together and an overall strategy to transform your network from a traditional data network into a real time media network.

UC deployment assessment

Offering telephony-, video conferencing and desktop- or application sharing services via Unified Messaging applications are highly dependent on the quality and configuration of all elements in the network chain between the users.

This is caused by the fact that real-time audio/video services will stretch the capabilities of your network to the limit, it will for sure expose every weak link between the end-points, weaknesses you have never noticed before. You will be surprised.

In general, elements can be miss-functioning, become over utilised or are not configured to cope with media streams. Many concrete examples can be given; 

• Call Admission Control settings in Microsoft Lync that do not match the configured end-to-end network capacity or not matching your end-to-end Quality of Service settings for real-time media streams.
• Increased jitter and packet drops caused by over utilised WAN links.
• Unstable connections caused by badly configured, bad performing or overloaded WLAN segments.
• Latency issues caused by an inefficient geographical design or bad IP routing in the network.
• Slow session setup caused by bad DNS response.
• Firewalls or other security controls that are not dialled in for real-time media traffic and SIP trunks causing connection or functional errors.
• Even existing Ethernet interface speed/duplex configuration errors anywhere in the network path between the end-points can cause your Unified Communication service to be badly accepted by users and now with the increased load, even slow-down other business critical traffic on your networks as well.
• Last but not least, low performance client computers or bad quality audio/video peripherals can cause bad user experience.
• Many of these errors in the network are already there but often not noticed before because network utilisation was relatively low or real-time video/audio streams where not used before.

Reasons enough to perform a UC pre-deployment assessment on your network before you introduce Unified Communication services to your users. Since some time also Microsoft recognised that the success of a Lync implementation is highly dependent on the network infrastructure and they are now stressing assessments like this as part of a Lync implementation as being mandatory.

Of course, if you already deployed UC and face quality issues now, you can use assessments like these to diagnose where the problem is and to re-configure or update your network and systems accordingly.

A UC deployment assessment can be approached in several ways:

pre-deployment testing

If the UC solution is not in place yet, a non-intrusive assessment can be done by performing a paper due diligence exercise in order to capture site details, bandwidth for each link, QoS settings, users per site and expected usage of UC functions that are foreseen. Microsoft offers several tools and questionnaires to capture requirements, user profiles, expected usage, details of the current infrastructure and help design the required topology, server infrastructure and bandwidth requirements;

• Lync Server 2013 Planning Tool is a wizard that can generate a server topology as part of the design process.
• The Bandwidth Capacity Calculator is a spreadsheet template that can calculate bandwidth requirements in a multi-site environment by using pre-defined user profiles and workloads.

traffic simulation

As stated, in many big international organisations, network components that are in between the peers are managed by different entities or suppliers. In these cases it can be a challenge to get a full overview of configuration details and monitoring of each device needed to get usable assessment results.

In those cases it may be more effective to execute tests by simulating traffic and verify if the network behaves as expected from an end-to-end perspective.

When executing traffic simulation tests, the end-to-end network will be tested by generating real-time media traffic between peers up until congestion level. Only in this state the network will reveal hidden issues that will come to the users’ attention when using real-time audio and video applications. While generating traffic flows, monitoring can be done on the network components in the path to verify utilisation, routing, packet drops, jitter values, security settings and QoS settings. Simulation, traffic generation, monitoring and testing of real-time media flows can be done with many commercial- or open-source tools. Well known tools are IxChariot, NetResolve and JPerf. Be aware that doing this in a live environment may affect other operational traffic if not setup correctly or if the unexpected issues you were looking for come to the surface. In order to have a representative reading on top of other baseline traffic, testing should be done over a longer period of time, at least a week, during a normal business calendar.

post-deployment testing

In case Lync is already installed and operational but having quality issues with voice- or video connections, a post-deployment assessment can be done in different ways depending on the tools used.

If enabled in Lync, standard monitoring reports accessible via the Lync Control Panel can already show call statistics and detailed quality reporting and helps to identify between which clients or sites issues occur.

• The Lync PreCall Diagnostics tool is a very handy and easy to use tool that can setup a SIP call and monitor Network MOS, Packet Loss and jitter from a client system.
• By checking several sources on a single Lync server, the Lync Server 2013 Best Practices Analyzer is able to verify if settings for AD, WMI and registry are correct which may help to optimise the server settings.
• To check server health and performance, Microsoft offers PowerShell scripts that generate Key Health Indicators which can be generated by creating data collectors and logs on the to be tested Lync servers. These logs can be exported to a reporting template for further evaluation.
• For diagnostic purposes and to capture-, query- and assess logs generated by the Lync Servers and Lync Clients, built-in tools like the Central Logging Service, OCSlogger and Snooper are very valuable.

Advantage of using tools based on logs generated by the Lync application itself is that any issues found are directly linked to the Lync functional perspective and workloads.

Disadvantage is that external factors that affect Lync performance are not captured and a more in-depth analysis with specialised network monitoring tools will be needed to find the root cause of issues experienced.

To overcome this disadvantage, specialised network and application performance monitoring tools that look at Lync traffic as part of all other network traffic can be used.

• Pure network monitoring tools based on Cisco NetFlow or WireShark can recognise Lync flows and give real-time visibility of network utilisation, traffic patterns and usage of QoS policies.
• Application performance monitoring tools like Ixia, AppNeta and Dynatrace provide detailed insight of application transactions in the network, enabling analysis of separate Lync sessions for issues from the network perspective.

These tools make use of network probes or switch SPAN ports in the path and can also “see” other Layer 2 and 3 traffic and network utilisation that may affect performance such as packet drops, QoS mismatches, wireless authentication and roaming issues.

Consequently the best way to find, analyse and solve performance issues in your Lync environment is to use both Microsoft and specialised application performance tools.

It must be noted that any UC deployment assessment need skilled network engineers or consultants that understand both network technique and the UC application behaviour in order to execute and deliver a usable report.

Ask your network provider, if they also act in a role as managed service provider, to offer UC pre- and post-deployment services like this. They are also in a position where they can monitor and manage most traffic streams and devices and have the skilled staff needed that are both experienced with Unified Communication applications and network infrastructure.

Are you facing quality issues with audio and video after implementing Unified Communications services, how do you deal with it? Let us know!

Marcel van Wort, CISSP-ISSAP

Managing Consultant UC and Security

Orange Business