Critical Success Factors in Design and Performance Management of UC Networks. March, PDF

Please download to get full document.

View again

of 20
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Film

Published:

Views: 3 | Pages: 20

Extension: PDF | Download: 0

Share
Related documents
Description
Critical Success Factors in Design and Performance Management of UC Networks March, 2008 Executive Summary The purpose of white paper is to provide the Business Decision Maker (BDM) with a non technical
Transcript
Critical Success Factors in Design and Performance Management of UC Networks March, 2008 Executive Summary The purpose of white paper is to provide the Business Decision Maker (BDM) with a non technical understanding of those factors that are critical to the successful performance of a Unified Communications (UC) solution with a focus on the voice application. Voice is the most widely implemented UC application to date. Today the competitive tension between application vendors, network vendors and a new wave of UC vendors has introduced a polarization of opinion on whether network management with attention to codec selection and Quality of Service (QoS), which concerns measurement of the treatment of the packets traversing a network including utilization, response time, latency (delays), delay variation, packet loss, jitter and availability, or application performance management with its focus on the unique VoIP Quality of Experience (QoE) requirements associated with differing business scenarios is the key to reliable operational performance. The truth of course is that both these aspects must be appropriately managed. And there is no one size fits all. The design and implementation of a UC solution must be targeted to each enterprise s unique needs if an optimal tradeoff of business requirements for enhancing individual and business process productivity, interoperability with legacy infrastructure and total cost of ownership is to be achieved. In achieving this balance the BDM must recognize that the productivity enhancing benefits of UC applications (e.g. UM, presence, integration with other business applications and simplicity of use) may be accompanied by performance considerations around bandwidth utilization and higher delay. However, with proper network design, attention to codec selection, QoS implementation in the network, and comprehensive performance management tools, it is possible to implement UC successfully. 1 Critical Success Factors in Design and Performance Management of UC Networks 1. Introduction Interest in and adoption of Unified Communications (UC) by businesses of all sizes is on the rise as they opt for the unique opportunity afforded by UC to integrate multiple forms of communications, both real time and asynchronous such as voice, video and conferencing, presence and messaging, as software applications into their value chain processes; thereby deriving significant competitive advantage from more efficient operations both internally and across their supply chain, as well as enabling markedly improved customer intimacy. Companies must address significant network design and communications application performance management concerns before the reality of UC meets the vision. Why? Real time applications such as VoIP drive more real time traffic onto networks. These real time streams are intermixed with non realtime traffic such as, or other back office applications. Now, VoIP is extremely bandwidth and delay sensitive. From an IP network perspective: for VoIP transmissions to be intelligible to the receiver, voice packets should not be dropped, re ordered, excessively delayed, or suffer varying delay (otherwise known as jitter). To ensure voice quality, current best practice is to classify data and voice traffic into different categories and give voice traffic priority handling across a shared data network backbone. Giving voice traffic priority handling minimizes delays and drops, and whenever possible, gives voice traffic predictable transmission performance. In other words, the underlying converged IP network provides the foundation for implementing intelligent services, Quality of Service (QoS) and security as well as resilience and connectivity. From a voice application perspective the other factors determining voice performance, such as echo, delay, speech level, noise level and speech distortion, must also be maintained within required operational boundaries. From an end user perspective both the network and voice application performance must be adequate to deliver satisfactory Quality of Experience (QoE). The goal in managing a converged network is to tune it so that many types of application data traffic can coexist and perform within business tolerances specific to the scenarios in which the applications are used. QoS mechanisms are necessary, as well as visibility into the underlying factors that affect end user QoE. The purpose of this white paper is to provide the Business Decision Maker (BDM) with a non technical understanding of those factors that are critical to the successful performance of a UC solution with a focus on the voice application. 1 Voice is the most widely implemented UC application to date. Today the competitive tension between application vendors, network vendors and a new wave of UC vendors has introduced a polarization of opinion on whether network or application management is the key to reliable operational performance. The truth of course is that both these aspects must be appropriately managed. And there is no one size fits all. However, as will be discussed below, with proper network design, attention to codec selection, QoS implementation in the network, and comprehensive performance management tools, it is certainly possible to implement UC successfully. 1 See the glossary at the end of this white paper for useful definitions. 2 2. Operational Requirements Voice performance is a critical issue for almost every business. Whether it is for internal communication or interaction with customers there is a minimum level of performance that must be achieved to ensure that productivity, business relationships, transaction rates and well being are not affected. Business workers rely on the telephone many hours out of the day, from collaborating with business partners and co workers to interacting with and helping customers and suppliers. Contact center agents spend the entire day on their telephones. Hence the first step to insuring an appropriately designed and managed UC system is to determine your business requirements. Key questions are: How will the system be used? How many calls per month (or day) are made out of your office? Are those calls to customers or internal employees? How many offices will you have on a system? Are there remote offices, mobile workers, or home workers to consider? How widely dispersed are they geographically? Different business scenarios have different performance requirements. Consider the following examples in Table 1 where, in each case, there is a level of performance below which the call quality would be viewed as unsatisfactory. Note, however that high fidelity telephony with a minimum MOS 2 requirement of is set to become common place with the introduction of UC. This will raise quality expectations in end users and reduce their tolerance for poor quality calls. Table 1: Voice Performance Requirements by Business Scenario Scenario A customer calling their bank A sales person working toward deal closure A call center agent (typical 8 hour shift) An occasional user of making an internal call Voice performance requirement Approximate min MOS requirement Notes on requirement high 4.0 Voice quality partly defines the customer s perception of the interaction. Customers become extremely stressed if it requires high effort to understand important and often unfamiliar information high 4.0 It is well known that sales people will deliberately choose a fixed line over a mobile connection to discuss transactions with customers. Better speech quality (often described as a clearer line) helps the parties to expend minimum effort hearing/understanding the dialogue high 4.0 High voice quality essential to maintain the well being of the agent and control staff churn etc medium 3.5 Occasional user will not be stressed by lower quality and interaction time is less critical A mobile call medium 3.25 The special utility provided by mobile phones is so high that people are prepared to accept incomplete coverage and low voice quality to have this utility. As mobile 2 Mean Opinion Score (MOS) provides a numerical indication of the perceived quality of received media after compression and/or transmission. The MOS is expressed as a single number in the range 1 to 5, where 1 is lowest perceived quality, and 5 is the highest perceived quality. 3 Based on wideband (7 khz acoustic bandwidth) opinion scale. 3 Long distance internet call network performance continues to improve several major mobile service providers are taking great strides to improve quality and market themselves on this basis. However, Over time users may grow less tolerant of lower quality. low 2.2 Almost complete cost avoidance makes users tolerant of almost unusable (poor) quality. This situation only remains true while the cost differentiation is sufficiently great and the operational/business risk is tolerable. 2.1 VoIP Performance Management Architecture While the performance requirement may vary between operational situations, there is always a performance level that must be met from a business operational view point. Efficient delivery of this performance level requires the capability to monitor and manage both network and application performance. For the traditional packet transport network there s a whole ecosystem around data applications. On the Service Assurance side you have HP s OpenView, Tivoli s NetCool, Computer Associates ehealth, etc. This operations support system infrastructure allows the network administrator an excellent vantage point from which to assess whether data applications are going to run adequately on the IP network. However, assessing voice application performance requires that you have additional information at your fingertips. You need to know, on a per session basis, whether the IP transport is service affecting for that session, implying a need to know things like the packet loss and jitter. You also need to know how these packet behaviors are distributed in time and how much power the user device has to error correct in order to know whether that particular network connection is going to affect the quality or not. You will also need to know several new things about the voice applications themselves: Was it noisy? Was there echo? Was the speech level sensible or was it too faint or too loud? Was the speech distorted? These are the things that you must know in order to ascertain whether the people involved were having a decent QoE or not. But these are things you can never learn by measuring just the packet transport. It s not that the IP data network tools out there today do a bad job of measuring those things. They don t measure them at all! And that leads to situations where you have unhappy employees and customers not able to converse successfully, while there is a solid row of green lights on the traditional IP network management tools. Consider this VoIP real comment from a deployment in the financial services sector. When the IPT system was only part deployed, we became very aware of the difference between quality of experience and quality of service. A small proportion of our call center staff were not satisfied with the call quality when using the IP telephony service, due to a number of factors including varying volume levels, or an echo or hiss on the line We purchased numerous quality of service tools, all of which indicated that our IP telephony system was working well. None of the tools could explain the mystery behind the small number of call quality issues that we were experiencing. We needed a tool to identify phone call experiences of customers as well as check the network infrastructure was working. 4 The answer is an integrated voice service assurance as depicted in Figure 1, below. Figure 1: Voice Service Assurance Architecture Voice Service Assurance Availability Management Performance Management Network Management Call Control Management Voice Performance Management Element Management Systems Network Performance Systems Fault Management Systems Vendor Management Systems Cisco Call Manager Avaya Call Control Nortel Voice Control Monitoring Based on User Experience Real time Passive Monitoring RTP Stream based IP Header ITU P.564 Payload Analysis Call Path Detail Alerting Real time Based on IP MOS Listening Quality Conversational Quality Tailored Thresholds Business Functions Gateways API to NMS and OSS Troubleshooting Drill Down to Specific Calls, Locations Diagnostics Analytics Root Cause Identify Voice/Network issues Find and Fix Reporting Custom Reports MOS Scores Performance CODEC Trend Analysis Historic Reporting Top N Problem Site Reporting Probes One notable candidate for the missing voice performance management component is Experience Manager developed by Psytechnics 4. Experience Manager is a performance management solution which can both measure the things you need to know about the IP network, but also can look inside the packets at the pair of wave forms for the two halves of the call. The measurements can be made either via embedded software at the endpoints or via a midpoint probe. The waveform analysis actually determines whether the call was noisy, had echo, whether speech levels were sensible, distorted and so on; providing that complete picture about whether you re achieving satisfactory performance with your real time voice communications applications. With Experience Manager you can set operationally relevant thresholds for different operational scenarios. If Experience Manager detects that you re not meeting your operational thresholds an alarm is raised. From that alarm you can go straight through to trouble shooting. The rest of product is all about drilling down to find specific faults, locations detail around what s gone wrong to inform a very efficient find and fix process. Another key requirement met by Psytechnics Experience Manager is the ability to provide performance management across hybrid (multi vendor) networks. Hybrid networks will be the norm in UC deployments as businesses combine multiple levels of UC application, access networks and core providers. A typical hybrid network would be a combination of Microsoft Office Communications Server 2007 (OCS) and Cisco Unified Communications and Collaboration applications. Experience Manager is 4 Psytechnics has supplied quality measurement software solutions to the communications industry worldwide. A spin off company from BT in 2000, Psytechnics has intellectual property rights in 6 of the ITU standards in the voice performance area. 5 vendor independent integrating seamlessly with the left side of Figure 1. It can readily be applied across different network technologies (including wireless), is integrated with OCS (with OCS end point data visible via the Experience Manager UI) and provides directly comparable standards based metrics for fault location and supplier management. Other vendors also address these issues although, primarily, from a network level view of performance. For example, Cisco Unified Operations Manager (CUOM) provides visibility into the network infrastructure, its performance, the applications used across the network, and the end points. CUOM uses open interfaces and numerous types of diagnostic tests to continuously monitor and evaluate the current status of both the UC infrastructure and the underlying transport infrastructure of the network. Key service and voice quality metrics include utilization, response time, latency (delays), packet loss, jitter, availability and MOS (as achieved by the IP Network). Cisco Unified Operations Manager does not deploy any agent software on the devices being monitored and thus is non disruptive to system operations. Information presented by a series of 4 dashboards (Service Level View, Alerts and Events, Service Quality Alerts and IP Phone Status) provides the network manager with a comprehensive view of the UC infrastructure and its current operational status. There are numerous practical causes of IP network and application performance issues in VoIP deployments. A few examples of each appear in Table 2. Table 2: Practical Causes of IP Network and Application Performance Issues Type of issue Cause and Effect IP network problems Packet loss and jitter Out of sequence packets, packet loss Delay, packet loss and jitter Application level problems Volume Levels (Loudness) Noise Echo Speech distortion Delay LAN congestion Leading to speech loss/distortion Diverse routing Leading to speech loss/distortion WAN QoS misconfigured Leading to speech loss/distortion Gateway pads, incorrect/faulty terminals, PBX faults, peering Noise floor from automatic gain control, faulty terminals/pbx/dsp, transcoding Poor edge devices (headsets, phones), faulty/under provisioned echo cancellers Poor/faulty edge devices (PC sound cards), transcoding, gateway/pbx DSP Low cost routing, failure of anti tromboning 5, excessive coding stages (i.e., transcoding) 5 Anti tromboning is the term used to describe the local re routing of media streams when a call is transferred over large geographical distances. 6 3. Hand in hand Partnership of QoS and QoE A well engineered and managed converged network is necessary though not sufficient to achieve operational voice performance. The goal in managing a converged network is to tune it so that many types of application data traffic can coexist and perform well. Good QoS policies must be in place to give priority to VoIP traffic over the TCP applications, which aren t very delay sensitive. The QoS for VoIP is mainly affected by latency, jitter (delay variation) and packet loss. It s bad business to have your voice traffic burdened by an excess of any of these effects. Latency, or delay, refers to the time it takes for a voice transmission to go from its source to its destination. As latency increases, it causes call participants to start interrupting each other because they believe the other person is finished speaking. A VoIP packet may be delayed for several reasons: Codec delays occur when the speech is encoded by the codec and the packet is created. Modern speech codecs operate on collections of speech samples known as frames. Each block of input speech samples is processed into a compressed frame. The coded speech frame is not generated until all speech samples in the input block have been collected by the encoder. In addition, many coders also look into the succeeding frame to improve compression efficiency. The length of this advance look is known as the look ahead time of the coder. Transport delay includes time to transmit the packets, buffer and queue them if needed, and move the packets from hop to hop through the network. Corporate QoS enabled IP networks use equipment with only about 25 to 100 microseconds of delay per hop. Without QoS, transport delay can be variable and high with congestion. Jitter buffers (used to compensate for varying delay) further add to the end to end delay, about 10 to 20 milliseconds (ms), and are usually only effective on delay variations less than 100 ms. Jitter must therefore be minimized. Transcoding delay occurs when one packet format is converted into another, or is formatted to cross the PSTN, like a current majority of inter enterprise calls. Such delays may be significant (several ms or more) across a multi hop call as multiple transcodings must be completed for end to end transmission. Propagation delay, the time taken for a signal to travel along its path from caller to called party can become relevant for off shore call centers, cross continent traffic, etc. For example, electrons travel through copper or fiber at approximately 125,000 miles per second implying that a fiber network stretching halfway around the world (13, 000 miles) induces a one way delay of about 70 ms. When these factors are combined, delay can easily become significant and there is a loss of synchronicity in the conversation and normal turn taking rules st
Recommended
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks