A practical guide to migrating to VoIP

One of the most crucial steps any organisation should take in deploying a solution for voice over IP (VoIP) is to plan for an Internet Protocol-oriented infrastructure. True, VoIP deployments vary significantly from one organisation to the next, and migrations don’t all follow the same path. But by understanding the principal elements of VoIP — network bandwidth and CODEC requirements, Quality of  Service (QoS) for networked voice traffic, standards such as the Session Initiation Protocol (SIP), etc. — organisations can
establish a specific, effective baseline for their VoIP solution deployment, and ensure the best possible system and network performance.
Critical phases of VoIP planning and infrastructure design
Since the mid 1990s when voice over IP was introduced, the IP industry has increasingly turned to open standards like SIP and recommended CODECs for network bandwidth in the effort to improve VoIP network readiness and security worldwide. At the same time to make migrating to VoIP more straightforward, IP vendors and service providers have continued to establish essential planning and design functions for a successful migration. Six of those key functions are discussed here.
Plan the right architecture for your particular VoIP deployment model
A single site or distributed locations? Also, is the migration phased, i.e., moving only selected systems, departments or sites to VoIP, or enterprisewide? Whichever model and migration approach, the planning goal is to structure your organisation’s cable plant design and data center resources sufficiently for VoIP call processing to all potential users. With regards to having all needed IP technologies in place and issues such as data center space, energy consumption etc, planning should also include a detailed inventory of existing and required architecture topology components. Among those components: each IP network device on the LAN or WAN/MPLS, plus gateways, routers, media servers, system servers (email, web services, speech recognition, CRM, databases, etc.), phones and other voice devices such as headsets, and business applications.
Shaheen Haque, Territory Manager, Turkey and Middle East , Interactive Intelligence Middle East
Understand the factors that impact voice (call) quality
As voice transmission travels over an IP-based data network, the clarity and quality of the call can be negatively impacted by delay, echo, and jitter. Delay stems from the amount of time it takes a VoIP voice packet to be created, sent across the network and converted back into sound, while echo results from delay(s) in any point of the voice packet process. Jitter occurs when voice packets arrive at an interval greater than they’re sent. Overall, echo becomes more noticeable as delay increases, and jitter is more prevalent when an IP network provides different waiting times for voice packet transmissions, or varying levels of latency. Another factor affecting voice quality is VoIP signaling/signal loss associated with delay.
When planning your network for VoIP, note that delay has the most impact on voice quality since it precedes echo and jitter. To achieve the best potential quality for a VoIP call, a general guideline is that one-way delay should not exceed 150 milliseconds. A range of 150-400 milliseconds is acceptable for higher one-way delay ranges, provided system administrators are aware of the increased transmission time and its impact.
However, any count above 400 milliseconds is considered unacceptable. To measure delay in total, a best practice is to determine end-to-end delay for Real Time Protocol (RTP) packets on the network without using “pings,” which are deliberately small info packets sent from one computer to another via the network being evaluated. And while some network vendors routinely use pings to garner “favorable” quality readings, pings aren’t subject to Quality of Service (QoS) controls for network bandwidth and the latency in voice packet transmissions. Fortunately with QoS, ongoing enhancements have minimized the latency that can hamper voice quality in VoIP configurations, and many telecom and Internet service providers use QoS measures to improve their VoIP network service.
Analyse and prepare your network for voice and data
Analysing a network’s voice and data traffic volumes and planning the appropriate capacity for VoIP isn’t something a company’s IT team does on a routine basis. Therefore, actual network prep is usually better left to a vendor or consultant certified in network assessments. To determine where your network is and where it needs to be for a VoIP migration, an analysis for voice/data network readiness, traffic capacity and ongoing reliability typically encompasses the following systematic valuations:
Voice network
·         Voice load measurement
·         Network traffic during normal and peak (busy) hours to avoid congestion
·         Voice/data transport selection (using Erlang B or Erlang C tables; see
·         Traffic engineering
·         Voice circuits between sites, identified for multi-site configurations
·         Voice traffic and traffic cost assessmentsAn inventory of PBXs and voice mail boxes
·         Quality of Service (QoS) with regard to switches and routers that prioritize voice traffic
·         Network reliability and voice (call) security
Data network
·         Voice/data bandwidthQoS features selection
·         WAN media types, identified
·         Traffic patterns
·         Data network costs
·         IP telephony considerations
·         Routers: Homo or heterogeneous? Modular? Voice-enabled?
·         Switches: Homo or heterogeneous? Voice-enabled?
·         PBXs and other telephony equipment: Product life cycle? IP-enabled?
·         Address schemes: RFC 1918 compliant or public addresses? Dynamic Host Configuration Protocol (DHCP) scope design?
A network analysis should further study factors like disaster recovery and E-911 service, call recording, quality monitoring, and how deploying VoIP capability to new branch offices and remote and mobile users might affect network bandwidth. Your network analysis should simulate VoIP traffic on the network to measure capacity and evaluate traffic characteristics, Quality of Service (QoS), congestion, reliability and other potential issues. By this, your organisation can make needed changes and reasonably assure network success before launching its VoIP initiative.
Determine CODEC and bandwidth needs
Defined, CODEC is the COmpression/DECompression that voice-based data packets experience when they’re converted from analog form to digital signals for VoIP. CODEC factors can originate in a PBX/IP PBX phone system and be shared by analog phones, or take place in phones themselves. The International Telecommunications Union establishes various CODECs as recommendations and standards for VoIP planning — G.711, G.726, G729, G723.1 are currently the most used — with CODEC selection typically driven by network design (LAN, WAN, MPLS). Determining actual percall bandwidth consumption depends on IP header size, voice payload size, and voice packets per second, or sampling rate. In general, each VoIP voice packet in a call transmission contains 40 bytes of IP overhead, and overall for CODEC bandwidth with overhead, combined WAN data (voice, video and data) should not exceed 75% of available link bandwidth if planning to optimize the network for VoIP.
To achieve the best possible voice transmission quality, it’s critical to incorporate the right CODEC and reach a maximum theoretical Mean Opinion Score, or MOS (see the following chart). In the MOS scale, 1 is interpreted as unintelligible for a VoIP call and 5 is considered ideal, although compression and other factors in an IP-based voice/data network make it virtually impossible to reach a score of 5. At the high end, a maximum MOS score of 4.2 to 4.4 is considered more realistic.
Default Data Rate
Datagram Size
Combined Bandwidth for 2 Flows
Default Jitter
Buffer Delay
Maximum MOS
64.0 kbps
20 ms
1.0 ms
174.40 kbps
2 datagrams (40 ms)
64.0 kbps
20 ms
1.0 ms
174.40 kbps
2 datagrams
(40 ms)
32.0 kbps
20 ms
1.0 ms
110.40 kbps
2 datagrams
(40 ms)
8.0 kbps
20 ms
25.0 ms
62.40 kbps
2 datagrams
(40 ms)
6.3 kbps
30 ms
67.5 ms
43.73 kbps
2 datagrams
(60 ms)
5.3 kbps
30 ms
67.5 ms
41.60 kbps
2 datagrams
(60 ms)
The truth about MOS and analyser tools to measure voice quality
Different people rarely interpret a call’s clarity the exact same way, and arriving at a true Mean Opinion Score to determine voice quality is similarly subjective — if not maddening. To measure quality via the VoIP voice packet transmissions over a network, many professionals in VoIP circles recommend using time-synchronized analyzers. Yet questions persist about the accuracy and consistency of such analyzers and the methodologies behind them. One study, for instance, cites that the same voice packet trace run through two different analysers produced MOS scores of 3.0 and 3.8 for the same call, whereas another “good quality” call was rated 2.8 and one considered “worse” came in at 3.4.
The truth is, analysers — and analyser vendors — almost never use the same algorithm to compute MOS. If you do turn to an analyzer tool, be sure the vendor provides actual test data for the analyzer, collected under real call scenarios and validated by a third-party tester. More so to establish a true baseline of MOS scores and avoid score discrepancies for your organisation’s VoIP environment, rate the quality of calls under network conditions specific to your own business, such as peak hours and call loads.
Determine QoS priorities and the appropriate methods/policies
Another key component of VoIP network planning is deciding where your organisation’s QoS priorities lie. Using the example of a multi-site centralised call processing deployment model, where call processing originates from a central site and reaches multi-site locations via SIP tie lines to a WAN (or MPLS), QoS can reside in network points for campus access, campus distribution, the WAN, and branch locations. Noting those points and having determined QoS priorities, the next steps are to characterize the data network, implement QoS policies and monitor the network’s operational load. QoS priorities themselves should consider how your network will be used and what level of network service is required (integrated services, differentiated services for guaranteed latency/delivery, best effort). Priorities must also consider all network applications. Characterising the data network requires dividing traffic into classes for voice, video and things like financial applications, E-business applications, point-of-sale transactions, back-ups or server synchronization, database transactions, and web surfing, file sharing, and quake. Once QoS is prioritized and characterized, implementing your QoS policy comes down to coding the actual priorities in your organisation.
Address security needs and potential issues
Any organisation that handles confidential information on an IP-based network must make securing calls and data an ongoing priority. Fortunately, the security mechanisms now available for IP technologies are some of the most stringent ever, and new standards are constantly being deployed to make security even more concrete. Among these standards, the Session Initiation Protocol (SIP) is highly accepted worldwide for its rigorous message encryption and user authentication in a VoIP environment, in large part because SIP is regulated by the Internet Engineering Task Force (IETF) for IP communications security.
In addition to SIP, two security standards to note for their encryption capability are Transport Layer Security (TLS) and the Secure Real-time Transport Protocol (SRTP). TLS is based on the Secure Sockets Layer (SSL) standard and extends two distinct layers of security for an IP-based network. The first layer is the TLS Record Protocol, which ensures a private network connection via symmetric encryption. The second layer is the TLS Handshake Protocol, which provides authentication between an IP application server and a client using digital certificates. Encryption using the TLS and SRTP standards has become a best practice for protecting calls that travel over an IP-based communications network, especially when used in conjunction with other safeguards such as virtual private networks, virtual LANs (VLANs), access lists, and voice traffic authentication.
Another valuable layer of voice messaging security in a VoIP environment is the Internet Protocol security (IPsec) protocol, a framework of open standards that leverage cryptographic security services to protect communications traveling over IP networks. Comprehensively, IPsec supports network-level peer authentication, data origin authentication, data integrity, encryption for data confidentiality, and replay protection. (Microsoft is a true IPsec believer, having implemented IPsec in much of its Windows product lineup via standards developed by the IETF IPsec working group.)
Take advantage of every possible security method when planning a move to VoIP and your organisation will be well protected.
The more you understand upfront about VoIP and how it works, the more straightforward your organisation’s migration to IP communications will be. And by knowing how the details of VoIP can affect system and network performance both positively and negatively, you’ll be better prepared to optimize VoIP performance throughout your organisation after deployment.
Previous ArticleNext Article

Leave a Reply


The free newsletter covering the top industry headlines

Send this to a friend