More Information
How much of the market has IP Telephony captured? And what is the forecast?
More than 50 percent of new stations shipped into enterprises in the United States are now IP. It is forecasted the pace will continue to accelerate so that the far majority of all new phones installed in 3 years will be IP. But many companies are holding back on a full commitment, citing factors of reliability and interoperability. Indeed many companies are finding that their location, or operation, calls for the migration to VoIP in stages. (Forrester Research)
What are the communication or transmission parameters that should be maintained for VoIP?
A lot of the discussions around VoIP Quality of Service focus on communication and transmission parameters. Among these, the best established parameter is overall latency. Latency is the time it takes a packet to transverse the path from where it originated to its destination. Latency should be 150 msec or less. Generally, anything more than this will be perceived by the end-user. Circuit routing, compression, packetization delay and buffering are just some of the factors affecting latency and a system engineer should determine a latency budget based on the path that the IP packets will take.
Packet loss is another parameter. "Bursty" packet loss is very disruptive to a voice conversation. Packet loss of less than 1 percent needs to be maintained. This sensitivity to packet loss would suggest that packet loss be a parameter that is monitored by the deployed management system.
Jitter, the variation in inter-packet arrival, is another parameter that is intrinsic to IP communications. Jitter can be managed by increasing or decreasing the delay in the jitter buffers keeping within the latency parameter. Typically a jitter buffer will be less than 40 msec and it should maintain the variation to less than 3 msec.
Other long-standing metrics for voice systems are the call set-up time and call success rate. Local call set-up should not exceed 3 seconds and toll calls should not exceed 5 seconds. This parameter should be tested after a new route is established from your IP PBX.
(For further information, one of the sources for this response is an article by Maurico Rosales titled "Business Class SLAs for VoIP Services" – Business Communications Review, July 2006)
What are the payoffs for an enterprise from installing an IP Telephony Platform?
The benefits include reduced long distance costs from site-site IP calling, easier adds, moves and changes, reduced telecom costs for remote workers, reduced voice infrastructure, reduced conferencing costs, better options on disaster recovery, and productivity gains from improved communication.
Perhaps the greatest benefit is the potential IP telephony has to accommodate business trends, increase business capabilities and facilitate applications. An example: the common control redundancy of the traditional TDM PBX is limited with co-located active and standby processors for call control. IP telephony systems, however, can be designed with primary and multiple secondary call control servers that can be geographically disbursed.
What is really needed to upgrade your data network for IP voice?
A network assessment is the starting point for a full deployment of VoIP. A limited deployment may be possible with some dedicated LAN equipment for common control and shared LAN infrastructure for end-points. But for a converged system the following assessments should be made:
- Review of the current network's switches and routers. Changes may be required to accommodate QOS by prioritizing VoIP traffic, plus power over Ethernet to power IP phones, and multiple virtual LANs for routing and handling of IP telephone traffic may also be needed.
- For remote VPNs encryption technologies and authentication can be provided by using IPsec of SSL.
- Review the current and planned traffic and bandwidth needs per site and port allocations based on engineering models considering busy hour, expected growth, peak file transfer periods, or nightly backups. During this analysis you will consider the planned use of voice codecs, considering trade-offs on bandwidth and voice quality.
- Review the WAN for use of VoIP between corporate locations. Does the network deploy QOS protocols if you are using MPLS or ATM? Or if you are using IP directly between end points is voice traffic getting prioritized? Is bandwidth adequate?
- Review backup power for all network components. Remember powering VoIP phones over the Ethernet means that local back up power (UPS) is needed in the remote closets.
- Review Network Management tools. It may be necessary to increase the tools you have to allow measurement of all VoIP endpoints.
- Lastly, a review of staffing and organization is required.
What QOS (quality of Service) mechanisms are most effective for handling converged voice traffic on both the LAN and the WAN?
For good voice quality every device in the data path of voice applications should be capable and configured to support acceptable QOS levels. This starts with the switches and routers and continues over Carrier networks that can support QOS SLAs. This will give voice traffic priority queuing; however, priority queuing will not overcome insufficient bandwidth. Proper capacity planning of the IP network, along with QOS provisioning, should minimize packet loss and jitter.
Bandwidth is greatly affected by the choice of codec used to form the IP voice packet and it is a very significant factor in perceived quality. Codecs such as G.711 are designated as "toll grade" with a voice quality MOS score of 4.1 out of a possible 5. Others, such as G.729, attempt to compromise between bandwidth and quality with a MOS rating of 3.9 but requiring only 8 Kbps for the voice channel instead of 64 Kbps.
VoIP gateways at the edge of the IP network can be configured to add additional gain/losses to optimize the overall loss plan. Depending on signal strength and the absolute delay associated with echo returns, echo cancellers in VoIP gateways can detect and eliminate echo.
Part of maintaining voice quality is to provide the tools to understand the network the voice IP packets will transverse. Network management tools should thoroughly measure all VoIP endpoints on the network and should determine availability, and measure packet loss, latency, and jitter. Traditional management functions for data networks should also be performed including auto-detect of network devices, continuous ping, traceroute, network scans, and SNMP.
It is recommended that sniffers, TFTP servers, and syslog servers are deployed at all major sites. Sniffers can be deployed to monitor specific voice-affiliated LAN ports or IP ports on the PBX and they can measure IP traffic within a specific period. TFTP servers are used to store configuration files and software images for network devices. Syslog servers accept the messages from routers and switches. They can all be used for trouble analysis and later for root cause analysis.
Which migration strategies offer investment protection without compromising features and functions?
For businesses that decide to introduce or migrate to IP-telephony the biggest decision is what to do with the existing circuit-based PBX. Should it be replaced entirely or should you migrate using most of the existing PBX's components?
A large part of this decision depends on the number of telephone sets that you would expect to keep as digital sets, since digital telephones and IP telephones would comprise about 50 percent of the system's cost. Reuse of a large number of phones would favor migration. Also a consideration of reliability may favor a migration plan because your IP phones would be dependent on the corporate LAN. Phones critical to the operation of the business can be kept off the LAN, whereas the majority of phones could be changed in phases to IP telephones.
Factors favoring a new IP system are phones and systems that are fully depreciated, planned upgrades to the LAN, or a new location.
For a medium to large system, migration would be the most cost effective means of introducing IP-telephony. Depending on the existing PBX there may be a good migration path to a converged IP telephony design using the majority of the existing system. For example, Avaya and Nortel have good migration paths.
As an example, the Avaya Definity models can be upgraded to the S8500 or S87XX systems by replacing the control carrier in the existing Processor Port Network with the respective S8500 or S87XX media servers. This allows customers to retain expansion port network carriers and circuit packs (Single Carrier Cabinets (SCC) or Multi Carrier Cabinets (MCC)) retaining circuit cards and digital or analog phones and other analog equipment. The customer can achieve a converged network by adding IP signaling interfaces (IPSI) to connect to the new S8700 or S8500 media servers, new media processors for TDM/IP connections and CLAN boards for IP ports to connect IP endpoints. In addition, media gateways can be used to support remote communication back to the media servers. There are many options on these gateways to design for the desired combination of IP, digital or analog endpoints.
What are the most common sources of trouble on IP-based voice networks?
Problems that are directly related to VoIP normally deal with communication and transmission parameters.
- If latency far exceeds acceptable values, phone conversations can be difficult when parties talk over each other as both start to speak at the same time. You may have witnessed this on a call being handled by a "Help Desk" in a foreign country.
- Due to jitter, the variability in packet arrivals, going outside of parameters can result in choppy or distorted conversations.
- Lastly, packet loss will result in very noticeable degradation of voice quality, with clipping of words, missing or incomplete parts of conversation, and white noise.
What level of performance or voice quality do I need to maintain on my VoIP network? Are there trade-offs?
The performance that is needed can be determined by how the VoIP circuit is used. For example, it may be acceptable for a voice circuit used for international calls to have a voice quality of "good" instead of "very good"; the same would apply to a site-to-site VPN to the warehouse. Whereas, "very good" may be the quality you want to maintain for calls to/from your corporate headquarters. Assuming that your WAN can meet general VoIP SLAs, the voice quality will become largely dependant on the codec selected for those circuits. Codecs present a trade-off between the bandwidth and voice quality.
Codec standards of G.711 and G.729 are commonly used. The G.711, which produces a "Toll Grade" voice channel of 64-kbps, is the best codec for voice quality. When you add the headers required on the VoIP packet the bandwidth needed for a single call is 80 Kbps. Whereas, the G.729 codec provides good voice quality and reduces the bandwidth required by reducing the voice channel to 8-kbps with a total bandwidth need of 30 kbps.
What staffing levels and skills are needed to support converged voice data networks?
Employees who have data network expertise often reside in different organizations than those with voice expertise. Since knowledge in both these areas is needed for the implementation of VoIP, successful convergence requires that these teams work together and share knowledge. Often an outside resource will be helpful in bringing these two groups together.
Contact TelTech for a free consultation. E-mail us at info@teltechcs.com or call 917-439-8202.
