Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide
  • Today, all data networks are converging on IP transport because the applications have migrated towards being IP-based. The TDM-based voice networks have also begun moving towards IP. Video conferencing is also rapidly moving towards IP. When the different applications had dedicated networks, QoS technologies played a smaller role because the traffic was similar in behavior and the dedicated networks were fine-tuned to met the required behavior of the particular application. The converged network mixes different types of traffic, each with very different requirements. These different traffic types often react unfavorable together. For example, a voice application expects to experience essentially no packet loss and a minimal but fixed amount of packet delay. The voice application operates in a steady-state fashion with voice channels (or packet) being transmitted at fixed intervals. The voice application received this performance level it operates over a TDM network. Now take the voice application and run it over a best-effort IP network as VoIP. The best effort IP network has varying amounts of packet loss and potentially large amounts of variable delay (typically caused by network congestion points). The best-effort IP network provides almost exactly the opposite performance required by the voice application. Therefore, QoS technologies play a critical role to ensure that diverse applications can be properly supported in a multiservice IP network.
  • Properly Engineered network – sizing, secure Good equipment etc. - low latency, resilient, manageable etc. Adequate B/W: Handle existing traffic, overlay with peak Voice Traffic QoS – discussed further Ongoing monitoring – PVQM…
  • For each site pair determine number of voice “trunks” needed Call attempts per busy hour ( C ) Average call holding time in seconds (finance industry average – 192 seconds) ( T ) Erlangs per site = ( C x T ) / 3600 (An Erlang is a trunk used for 1 full hour = 36 CCS) Desired Grade of Service (assume P.01 – 1 call in 100 gets blocked)
  • A company has two sites connected by a leased-line WAN connection (PPP) operating at 128 kbps. Due to the potential use of 20% of link capacity for “zero-bit stuffing”, a safe assumption for link capacity is 102 kbps For design purposes, assume a maximum utilization of 70% (in this example, 90 kbps). This bandwidth has been sufficient for the current data requirements. The company believes that it only needs 70-80 kbps most of the time, with occasional traffic peaks up to the full capacity The company wants to support up to 4 simultaneous voice calls over the IP WAN network between the sites
  • Others can also do this (Netformx?)
  • Goals: Minimize Delay, Jitter, Packet Loss How? Good network design Adequate Bandwidth QoS Ongoing Monitor & Maintenance e.g. If you have a 256kb link, you would not put 3 x G.711 calls on the link
  • If a voice application is sent over a best-effort IP network, the following can occur: Voice packets experience variable, unpredictable amounts of delay Voice packets are dropped when the network is congested Voice packets can re-ordered by the network if the packets arrive out of sequence Nature of Voice Timing of delivery is paramount Very low end-to-end delay (latency), jitter Okay to lose / change some bits (except modem/fax) No time to recover from errors Nature of Data Timing less important than accuracy Accuracy of delivered information is paramount Errored frames tossed out at earliest detection, end devices recover lost, damaged packets, fix out-of-sequence delivery problems
  • What is Quality of Experience (QoE) and Quality of service (QoS) ? QoE is the fundamental determinant of performance for any technology and is the user perception of quality. QoS is a mechanism to help achieve QoE. Much has been written about QoS, and can be summarized into three primary jobs: To minimize end-to-end delay through the network (VoIP traffic to front of the line) To minimize the variability in end-to-end delay, that is, jitter (VoIP traffic to front of the line) To prevent packet loss (separate queues for the VoIP traffic) QoS doesn’t improve all network traffic performance. It simply gives a higher priority level between 0 lowest and 7 highest to administrator-specified traffic. This does not necessarily mean IP telephony traffic is top priority. Typically IP telephony traffic is given a level 6, and level 7 traffic is reserved for the network. QoS mechanisms can only be effective, however, over full duplex, point-to-point LAN and WAN links when there is one and only one device on each end of the link (most important for devices like L2 and L3 switches, frame switches, ATM switches, VoIP Gateways, etc.), and some QoS prioritization/scheduling mechanism is employed. The critical concepts are: Separate transmission medium in each direction Sending device has complete control over what is sent No other device can interfere with the transmission
  • MOS: Subjective method of rating quality of a conversation E-Model: Mathematical approach to measuring voice quality (implemented by Nortel…Telchemy Agent)
  • The simplest choice is to do nothing and hope that competition will continue to reduce bandwidth costs and therefore you can get a higher bandwidth connection for the same or marginally higher than your current costs. During times where there are high amounts of competition and in a strong economic climate, this may work to your advantage. However, the choices for high bandwidth last mile technologies are becoming more limited due to mergers and acquisitions of network service providers. Furthermore, many applications require predictable performance to ensure high customer satisfaction and these applications demand some form of QoS. There are may QoS technologies to choose from and unfortunately, no single QoS technology can provide the entire solution. Some examples of QoS technologies are given here and are: Differentiated Services The use of Class of Service in ATM Fragmentation and Multi-Class Extensions in PPP The Service Classes in Frame Relay Ethernet 802.1Q and 802.1p And Multi Protocol Label Switching Since IP is the predominant networking protocol, IP QoS can be achieved across the network using DiffServ technology.
  • Active Monitoring: Potentially adds to issue, by injecting further traffic onto already overloaded network
  • RTPStatShow / rTraceRoute can be issued from the Signalling Server
  • RTCP allows adjust voice coding algorithm based on network performance Can throttle non real-time traffic if thresholds exceeded
  • Proactive voice quality management ensures the premium quality of service (QoS) conditions that voice service demands. With Nortel VoIP service, you get clear, quality voice, equivalent to traditional networks. The solution continually monitors and reports network conditions in real time. Network administrators always have a current view of how the network is performing, and they can export and analyze performance data any time to ensure continued high levels of service.
  • Presentation

    1. 1. Pat Dempsey Head of Strategic Services NUI Galway Email: Pre-deployment Engineering for Voice over IP Solutions (IPT implementation in NUI Galway)
    2. 2. Objectives of this session <ul><li>Understand how IP network design can impact the quality and reliability of VoIP services </li></ul><ul><li>Understand the basic factors and design concepts for designing the IP network to support VoIP traffic </li></ul><ul><li>Calculate typical bandwidth requirements on the users IP WAN based on voice services requirements </li></ul><ul><li>What Tools are available to configure/troubleshoot </li></ul>
    3. 3. Voice Over IP Is a Unique Application - Demands Intelligent Handling 1 These applications are highly loss sensitive but loss is managed by TCP retransmissions Best Effort Traffic Email (store/forward) Client / Server Transactions Streaming media Video Conferencing IP Telephony APPLICATION Sensitivity to PERFORMANCE DIMENSIONS Low Low Low Low High High Jitter Low Low Med Med High High Delay Low Low-Med High 1 Low High 1 Low Med Low-High Med High Med Low Loss Bandwidth
    4. 4. Delivering Quality of Experience <ul><li>A satisfactory level of perceived voice quality is achieved through the following: </li></ul><ul><ul><li>a properly-engineered network </li></ul></ul><ul><ul><li>good network equipment and redundancy </li></ul></ul><ul><ul><li>adequate bandwidth for peak usage </li></ul></ul><ul><ul><li>use of QoS mechanisms </li></ul></ul><ul><ul><li>ongoing monitoring and maintenance </li></ul></ul>We will focus on these for the rest of presentation Design Guidelines / Traffic Engineering – next 4 slides
    5. 5. Design Recommendations for VoIP - typical <ul><li>The following slides are typical considerations when designing for VOIP </li></ul><ul><ul><li>Vendors only supports customers with Layer 2/3 switched networks (no shared media devices, cable-based, hub-based LAN) </li></ul></ul><ul><ul><li>L2 switch ports must be set to autonegotiate for VoIP devices </li></ul></ul><ul><ul><li>Goal of Zero Percent Packet Loss for VoIP </li></ul></ul><ul><ul><li>Use G.711 CODEC when possible </li></ul></ul><ul><ul><ul><li>Excellent Voice Quality </li></ul></ul></ul><ul><ul><ul><li>Bandwidth usually available in LAN and MAN </li></ul></ul></ul><ul><ul><li>Use G.729A or G.729AB to conserve bandwidth </li></ul></ul><ul><ul><ul><li>Take care to meet customer voice quality requirements </li></ul></ul></ul><ul><ul><ul><li>Watch out for multiple transcodings (multiple VoIP hops) </li></ul></ul></ul><ul><ul><ul><li>Be careful with VAD – subject to clipping effects </li></ul></ul></ul><ul><ul><ul><li>Centralised voice mail and music can be a call quality issue </li></ul></ul></ul>
    6. 6. Traffic engineering process - typical <ul><li>For site pairs, determine voice “trunks” needed </li></ul><ul><li>Calculate VoIP bandwidth demands </li></ul><ul><ul><li>Traffic Bandwidth Calculator / Vivinet Assessor </li></ul></ul><ul><li>Overlay VoIP traffic patterns onto physical network diagram </li></ul><ul><ul><li>Vivinet Assessor </li></ul></ul><ul><li>Size the required primary and alternate converged network links: </li></ul><ul><ul><li>Evaluate current traffic demand </li></ul></ul><ul><ul><li>Calculate, add in VoIP traffic demand </li></ul></ul><ul><ul><li>Evaluate various failure scenarios </li></ul></ul><ul><ul><li>Factor in desired headroom, unusable bandwidth </li></ul></ul>
    7. 7. Bandwidth Example <ul><li>Requirement: A company wants to support up to 4 simultaneous voice calls over the IP WAN network (128kbps) between two sites </li></ul><ul><li>If all 4 calls were simultaneously active, this would require 108.8 kbps (using a G.729 codec, 20 ms voice sample, and PPP overhead/frame) of the available 90 kbps of the 128 kbps link </li></ul><ul><li>This requirement exceeds the carrying capacity of the link and completely starves that data traffic </li></ul><ul><li>The solution is to upgrade the WAN connection bandwidth. A 256 kbps link is the minimum speed to provide 109 kbps for four G.729 VoIP calls, 80 kbps for data, and 20% availability for zero-bit stuffing </li></ul>
    8. 8. Is customer Network Ready for VOIP - Perform a Network Assessment <ul><li>Health Check – NUIG used NetIQ </li></ul><ul><li>Pinpoint mis-configurations prior to deploying a single phone </li></ul><ul><li>Can WAN links support G.711 or G.729? </li></ul>
    9. 9. What hurts VoIP Call Quality? <ul><li>Multiple transcodings of compressed voice </li></ul><ul><ul><li>Tandem hops, voice mail compression </li></ul></ul><ul><li>End-to-end delay </li></ul><ul><ul><li>Budget 250ms for G.711 </li></ul></ul><ul><ul><li>Budget 150ms for compression CODECs (G.729) </li></ul></ul><ul><li>Jitter – variable arrival interval between packets </li></ul><ul><ul><li>Late packets = Lost packets </li></ul></ul><ul><li>Packet Loss </li></ul><ul><ul><li>Our network likes to throw things away rather than forward damaged goods </li></ul></ul><ul><ul><li>Overloaded queue situations, device just can’t hang onto packet </li></ul></ul><ul><li>Goal: Design Network and PBX to minimise the effects of the parameters above </li></ul>
    10. 10. IP/Packet Networks – Why QoS? <ul><li>IP networks do not guarantee that bandwidth will be available for voice calls unless QoS mechanisms are used </li></ul><ul><ul><li>QoS to restrict delay, minimize packet loss </li></ul></ul><ul><li>QoS techniques can be applied to support VoIP with acceptable, consistent and predictable voice quality </li></ul><ul><li>QoS mechanisms refer to packet tagging mechanisms and network architecture decisions on the TCP/IP network to expedite packet forwarding and delivery </li></ul>
    11. 11. QoS versus QoE <ul><li>Quality of Experience (QoE) is subjective and relates to the actual perceived quality of a service by the user </li></ul><ul><ul><li>This applies to voice, multimedia, and data </li></ul></ul><ul><li>Quality of service (QoS) is an optimization tool designed to deliver a certain Quality of Experience (QoE) by ensuring that network elements apply consistent treatment to traffic flows as they traverse the network </li></ul>
    12. 12. Measuring QoE: MOS and the E-Model <ul><li>Mean Opinion Score (ITU P.800) </li></ul><ul><ul><li>Subjective call quality measurement perceived by the user </li></ul></ul><ul><li>E-Model (ITU G.107) </li></ul><ul><ul><li>Transmission planning tool for estimating user satisfaction </li></ul></ul><ul><ul><li>Objective measurement </li></ul></ul><ul><ul><li>E-model output: R value </li></ul></ul><ul><ul><ul><li>Under 60 is not acceptable </li></ul></ul></ul><ul><ul><ul><li>Over 94.5 is unattainable in VOIP </li></ul></ul></ul>Adapted from Diagram by Roger Britt, Senior Eng., Nortel Average quality scores over the duration of a call may not reflect end users perception of call quality R-Value User Satisfaction MOS Not Recommended Nearly All Users Dissatisfied Many Users Dissatisfied Some Users Dissatisfied Satisfied Very Satisfied 0 50 60 70 80 90 94 100 1.0 2.6 3.1 3.6 4.0 4.3 4.4 5.0 Toll Quality
    13. 13. What are the Choices for QoS? <ul><li>There are several ways to deliver QoS, including the following: </li></ul><ul><li>Network QoS Technologies </li></ul><ul><ul><li>Ethernet 802.1Q/802.1p </li></ul></ul><ul><ul><li>IP Differentiated Services (DiffServ) </li></ul></ul><ul><ul><li>ATM CoS </li></ul></ul><ul><ul><li>PPP Fragmentation and Multi-Class Extensions </li></ul></ul><ul><ul><li>MPLS for Traffic-Engineered Paths </li></ul></ul><ul><li>VoIP Application QoS Technologies </li></ul><ul><ul><li>Codec Selection </li></ul></ul><ul><ul><li>VAD / Silence Suppression </li></ul></ul><ul><ul><li>Call Admission Control / Bandwidth Management </li></ul></ul><ul><ul><li>Packetization rate </li></ul></ul><ul><ul><li>Jitter buffer size </li></ul></ul>Some QoS technologies are end-to-end
    14. 14. QoS Management: Ongoing Monitoring <ul><li>Passive Monitoring </li></ul><ul><ul><li>Source code integrated into endpoints (i.e. Telchemy Agent in Phone) </li></ul></ul><ul><ul><li>Software performs real time, in-call quality calculation </li></ul></ul><ul><ul><li>Metrics can be obtained at end of call or mid call </li></ul></ul><ul><ul><li>Alerts in real time for voice quality degradation </li></ul></ul><ul><li>Active monitoring </li></ul><ul><ul><li>NetIQ performance endpoints generate synthetic voice traffic </li></ul></ul><ul><ul><li>Useful for ongoing assessment of network and troubleshooting </li></ul></ul>
    15. 15. Phone Diagnostic Capabilities <ul><li>Ping and Traceroute </li></ul><ul><ul><li>The administrator can execute the Ping or Traceroute command from a specific endpoint with any arbitrary destination, typically another endpoint or Signaling Server. </li></ul></ul><ul><li>IP Networking statistics </li></ul><ul><ul><li>The administrator can view information on the packets sent, packets received, broadcast packets received, multicast packets received, incoming packets discarded, and outgoing packets discarded. </li></ul></ul><ul><li>Ethernet statistics </li></ul><ul><ul><li>The administrator can view ethernet statistics (for example, number of collisions, VLAN ID, speed and duplex) for the IP Phone on a particular endpoint. The exact statistics will depend on what is available from the IP Phone for the specific endpoint. </li></ul></ul><ul><li>UNISTIM statistics </li></ul><ul><ul><li>The administrator can view RUDP statistics (for example, number of messages sent, received, retries, resets, and uptime) for the IP Phones. </li></ul></ul><ul><li>Real time Transport Protocol statistics </li></ul><ul><ul><li>The administrator can view RTP/RTCP QoS metrics (for example, packet loss, jitter, etc.) while a call is in progress. </li></ul></ul>
    16. 16. Real Time Protocol RTP and RTCP <ul><li>Real-time transport protocol (RTP) </li></ul><ul><ul><li>Provides end-to-end delivery for voice and video on top of UDP </li></ul></ul><ul><ul><li>Maintains packet sequence </li></ul></ul><ul><li>Real-time transport control protocol (RTCP) </li></ul><ul><ul><li>Specified in same IETF standard, RFC 1889 </li></ul></ul><ul><ul><li>Monitors and controls information of the RTP session (not an independent protocol) </li></ul></ul><ul><ul><li>Separates flow - RTP port number +1 </li></ul></ul><ul><ul><li>Transmits packets as a percentage of session bandwidth (min. of every 5 seconds) </li></ul></ul>
    17. 17. <ul><li>Passive voice quality monitoring notifies network managers of quality degradation in real-time, expediting problem resolution </li></ul><ul><li>Proactive thresholds identify problems before they are perceptible to the user and impact end-user productivity </li></ul><ul><li>Granular statistics supply accurate metrics for troubleshooting and SLA delivery </li></ul><ul><ul><li>Jitter, latency, packet loss, jitter buffer discards </li></ul></ul><ul><ul><li>Accurate MOS and R-value </li></ul></ul>RTCP XR IETF RFC 3611 - Focus on End User Experience RTCP XR: Real-Time Control Protocol eXtended Reports CODEC IP Phone CODEC IP Phone RTCP XR IP Network RTCP XR
    18. 18. Why Traditional Management Approaches Don’t Work for IP Telephony <ul><li>Can’t provide end-to-end information from a user perspective </li></ul><ul><ul><li>Traditional approaches to managing IP networks capture the wrong metrics: throughput is not relevant for real-time traffic and jitter vs. discard rates </li></ul></ul><ul><li>Doesn’t see transient problems </li></ul><ul><ul><li>Current metrics (such as RTCP) are too coarse: per call statistics and average packet loss rates are not detailed enough to capture the transitory nature of the impairments </li></ul></ul><ul><ul><li>Can’t combine per-call average metrics </li></ul></ul>Traditional Approaches Do Not Work for Real-Time Traffic Do not accommodate transient effects and end user perspective
    19. 19. Proactive Voice Quality Management Closed-loop Monitoring and Troubleshooting for Actual Voice Quality in Real Time Voice Quality Metrics Support Real-Time Notifications on Quality Degradation Route Analysis Problem Isolation Voice Monitoring System Gateway Call Server RTCP XR SNMP Signalling
    20. 20. Where it all fits in! Meridian 1 Components of IP Telephony Systems Media Gateway’s have always been a part of the core TDM PBX. Formally referred to as IPE shelves in a Meridian 1, Digital Cards/ Analog Cards and Trunks reside here. Media Gateway Call Server The Call Server has been in existence since the inception of the PBX. Acting as the “brains” of the PBX, it provides all of the core telephony features and functionality. Signaling Server The signaling server was introduced to provide the IP intelligence to register, manage, and direct IP components. CS1000M Once the IP Components have been added and the software is upgraded to Rel. 3 or higher, the system is referred to as a Communication Server 1000M or CS1000M.
    21. 21. Migrating an Existing Location to support IP Existing Meridian Option X1 PBX + + Administration (Digital) Courtesy (Analog) New Software (Release 4.5) Migrated CS1000M = Administration (Digital) Migrate all previous features/services to support analog, digital and IP. IP Enabled CS 1000M Supporting IP and all previous services Courtesy (Analog) Executive (IP) Signaling Server Signaling Server
    22. 22. Flexible Telephony Deployment in NUI Galway: <ul><li>We have choice: TDM, Hybrid IP with new multimedia applications) </li></ul>CS 1000E Call Servers Signaling Server(s) Media Gateway IP Phones (up to 15,000) Analog/Digital Phones IP Phones (up to 15,000) Analog Phones IP IP & Digital CallPilot, Digital Meridian 1 Signaling Server(s) =central dialplan IP phone services CS1000M Digital Phones Analog/Digital Phones Branch Media Gateways LAN LAN WAN PSTN