This document is intended to help understand and prepare your networks for implementation of VoIP services. The concepts discussed in this document are intended to help evaluate and prepare your network prior to deploying cloud-based VoIP services.
Pre-Deployment and Network Readiness Assessment
IP Telephony is very different from conventional data applications in that call quality is especially sensitive to IP network impairments.
Existing network and traditional call quality problems become much more obvious with the deployment of VoIP services. This means that LAN and network equipment may need to be upgraded and or replaced. Diagnostic tools are also needed when deploying and maintaining
Description of Network Components and Terms
By adhering to the recommendations provided in this document, clients will be able to prepare their network for deploying Hosted PBX services. While implementation models may vary slightly based on product, general suggestions will be provided that will help ensure
successful implementation and operation of services.
Internet Service Provider (ISP) – All VoIP services are IP based and rely heavily on the customer’s connection to the Internet. This connection is provided by the ISP and can range from dedicated fiber to cable or DSL.
IP Addresses – IP addresses are used to uniquely identify devices on a network and are classified as either “Public” or “Private” IP addresses. The ISP will typically distribute one or more public IP addresses to the WAN interface of the router or modem.
Modem – modems are provided with services such as cable Internet and DSL and is the DMARC between the ISP and the internal network. An Ethernet handoff is typically provided from the modem. Modems provided by the ISP actually contain router and/or firewall
functionality. When a separate router/firewall is used, the modem must be set to “Bridge Mode” to avoid issues with double NAT (Network Address Translation) that can severely impact VoIP traffic.
Router – A router is a device that routes packets between various IP subnets/networks. A router makes the decision to route traffic based on its routing table which is populated either by static routes, which are provided manually during configuration or by routing protocols which automatically exchange route information between routers based on network configuration and status. Most routers deployed will likely use static route and will simply route between the ISP on the WAN side of the router and the internal network on the LAN side of the router.
Switch – Switches are generally layer 2 devices that are deployed in the LAN (Local Area Network) environments and used to aggregate device connections on the LAN. Switches can be used to segment Ethernet networks and connect a large numbers of devices on a layer 2 to a layer3 network which is typically represented by the use of a router.
Bridge Mode – Putting a Modem/Router “combo device” (often supplied by ISPs such as AT&T or Comcast) in “Bridge Mode” effectively makes it work as a modem only. This will disable DHCP address assignment throughout your Local Area Network (LAN). You will need to use a separate router behind this device in this scenario.
While there are many different signaling and media protocols that can be used for VoIP traffic, all conversations regarding VoIP will assume SIP/UDP is being used as the signaling method and RTP/UDP is being used for all media. When working with VoIP services there are two
types of IP traffic that needs to be handled properly in order for call setup and voice communication to be successful.
SIP: SIP stands for Session Initiation Protocol. SIP is used for call setup, tear down, and all other signaling required for the management and manipulation of VoIP service. SIP primarily uses UDP port 5060 for transport. Typically, you will configure your firewall so that port 5060
is prioritized so that two-way SIP communication can be maintained.
RTP: RTP stands for Real-Time-Protocol and is the actual voice traffic that is transmitted once the call has been set up between two endpoints. Many calls can be going on at the same time from one site to another so a wide range of ports are typically available for RTP
traffic in order for each call to be designated a unique port. UDP ports 10,000 – 20,000 are used for RTP traffic and can be used when establishing rules to allow and prioritize VoIP traffic.
How VoIP Communication Works
The following represents a very basic, high level description of how VoIP communications work. This is being provided solely to illustrate why the recommendations for both network and device configuration are so important when preparing your network for VoIP services.
Registration – VoIP services are SIP based and rely on SIP registrations and authentication. When a device is connected to a network it will make an attempt to send SIP registration messages to the CoreDial cloud. Assuming the network is configured properly the registration
messages will be exchanged between the CoreDial cloud and the endpoint and registration will be established. Messages will periodically be exchanged between the CoreDial registration servers and your SIP devices in order to maintain this registration status. It is these periodic messages that keep connections open through the router / firewall and allows for consistent two-way communication. If for any reason your router/firewall closes this connection the phone will either become unregistered or will be flagged as unreachable by the PBX.
Call Set Up – In addition to establishing and maintaining registration status, SIP messages are also used to set up and tear down VoIP calls. When call set up is performed, both ends acknowledge set up of the call and established the appropriate ports for sending and receiving of the call media (RTP) traffic.
Voice Calls Once the call set up is complete the Hosted PBX and the device making or receiving the call will pass the actual call media back and forth as RTP traffic between the Hosted PBX and device. This is separate from SIP traffic and is why we need to prioritize and
manage both SIP and RTP when preparing our network to handle VoIP.
Bandwidth Utilization Most VoIP providers use the G.711 codec for RTP. This is essentially an uncompressed voice codec which uses approximately 87.2Kbps of bandwidth in each direction (upstream and downstream) on a network during an active call. Estimating the maximum number of concurrent calls and using the 87.2Kbps bandwidth utilization as shown below will help ensure that the ISP connection will be sized properly to handle VoIP traffic. Because of its low bandwidth requirements, G.729 is mostly used in Voice Over IP (VoIP) applications where bandwidth must be conserved.
Call Path A prepaid call path is defined as a call to/from somewhere on Jet-Dial’s CoreDial network to/from somewhere outside of the Jet-Dial CoreDial network such as a PSTN or cellular network. While all calls made from SIP devices on CoreDial service will use
bandwidth on the local network, not all calls will use call paths. It is important to make this distinction, as it is integral in the planning for network capacity.
The above diagram shows the difference between Bandwidth Utilization and Prepaid Call Paths in the Hosted PBX environment. Here you can see the difference between the bandwidth consumed for an extension to extension call at the same location when compared to the bandwidth consumed for a call from a single extension to/from the PSTN. Internal extension to extension calls do not consume any prepaid call paths as all voice traffic stays within the cloud network, while a call to/from the PSTN will use a prepaid call path.
By testing your internet connection, you will be able to determine (at least for a given slice in time) the quality of various network measurements that are important for VoIP quality. Those measurements and their description are provided below.
Packet Loss: VoIP is simply voice communication in IP packets. Minimizing packet loss on the network is a critical part of ensuring quality VoIP communication. When measuring network quality, you should ideally look for 0.0% packet loss but VoIP will typically provide quality
up to 0.75% packet loss. Jitter: Jitter is the variation in the time between voice data packets arriving to a destination. Jitter of 50ms (milliseconds) or less is good. If higher, voice quality could be adversely affected.
Latency: Latency is the time it takes for your voice data packet to reach its’ destination. Latency of 150ms (milliseconds) or less is good. If higher, voice quality could be adversely affected by echo and/or delayed voice delivery (resulting in one speaker interrupting the other).
MOS: MOS stands for Mean Opinion Score and is a scale for estimating the quality a VoIP call based on factors such as Jitter, Packet Loss, and Latency. MOS is a subjective measurement- a network should be able to maintain a MOS score of at least 3.6 or greater in order to
provide acceptable quality VoIP.
QoS: QoS is an abbreviation for Quality of Service. You must have good QoS to have good quality phone calls. QoS is determined by how much bandwidth you have, how good the connection is, and if calls get priority over other network traffic. Inter-Quest can help you measure the quality of a given network connection at a specific slice in time by using a special VoIP readiness test tool.
Recommended Router/Firewall Settings
Enable QoS to Prioritize VoIP traffic
Prioritize UDP port 5060 for SIP
Prioritize UDP ports 10,000 – 20,000 for RTP
Turn off/disable SIP ALG (whenever present)
NETWORK & CONFIGURATION CONSTANTS
- When both an ISP modem and a QoS router are used in the same network, the modem should be in “Bridge Mode” to avoid NAT occurring on both the ISP modem and QoS router. This will most likely cause VoIP problems. If you do not have access to the ISP modem in question you will need to contact your ISP.
- It is always a good idea to configure QoS (specifically prioritization for VoIP at the connection to the ISP). Even in an ideal situation where separate networks and separate ISP connections are used for voice and data, configuring QoS at the upstream router should guarantee quality in the event that a non-voice device is accidentally introduced to the voice network. The method you use to configure QoS will depend on the make and model of the router you are using.
- Give the highest possible priority to UDP port 5060 (for SIP traffic) and UDP ports 10,000 – 20,000 (for RTP traffic)
- Prioritize all traffic to and from CoreDial Registration Servers using URL (i.e. sip.east.sipregistration.com) NOTE: Using the IP address (instead of URL) of the Registration Server can cause issues as the IP address of the registration server can change over time.
- Prioritize traffic based on the network segment or the VLAN it is coming from in situations where you have created separate voice and data networks (either physically or logically).
Physical Separation of Voice and Data Networks
The best way to ensure that VoIP traffic doesn’t compete with traditional data traffic is to make sure that VoIP and data do not even cross over the same cable or equipment. By keeping Voice and Data traffic separate you remove any competition for resources and guarantee that Voice has unfettered access to all available resources.
Option 1a below shows a basic example where two completely separate networks are being used, one network for Voice and one for Data. With this example separate ISP connections are provided which is an ideal situation when deploying Voice services. The ISP connection is typically the source of bottlenecks in any network. Establishing a separate ISP connection for the voice network guarantees that Voice traffic has the best chance for minimal packet loss, jitter, latency, and delay. Also providing a separate ISP connection for voice also helps you right-size your ISP connection to adequately support your expected VoIP traffic.
Option 1b shows a slight variation to the network described above. With the network shown below, there are still physically separate voice and data networks at the Layer2 level on the customer LAN but those networks converge at Layer3 (IP Layer) when they are connected at
the router level. This helps to avoid collisions and congestion issues on the LAN but it does make configuration of QoS prioritization on the router a critical component of success as both voice and data will be competing for resources on the ISP connection.
Logical Separation of Voice and Data Networks
In a scenario where voice and data must traverse the same physical cabling, network equipment can be used to create a virtual separation of voice and data. This is done using VLANs and requires VLAN capable switches and VLAN capable routers.
As previously stated, the local ISP connection is typically the biggest point of contention in the network and therefore it is ideal to provide separate voice and data ISP connections.
Option 2a below shows a scenario where dual ISP connections have been installed but separate voice and data LANs do not exist. A VLAN capable switch is being used to create a logical separation between voice and other data traffic. This can be done even when user’s computers are connected to the network via the phones or “daisy chained”. This logical separation makes it easy for the switch to separate the voice and data traffic into separate physical uplinks as the traffic is passed to the voice or data router as appropriate.
Option 2b below shows an example of a network where there is no option at either the LAN or WAN level to physically separate voice and data traffic. In this example VLANs are still used to logically separate voice and data traffic until traffic is merged at the router and then
handed off o the ISP. Since voice and data are traversing all of the same physical components in this example, the only real advantages to using VLANs is 1) you have the ability to easily provide QoS further out in the LAN if switches in the network support QoS and 2) it is
easier to prioritize and manage voice traffic since it will all be on a specific VLAN and can be controlled separately from standard data. In a network scenario similar to the one depicted in option 2b ample bandwidth and properly configured QoS are critical factors in maintaining
Flat Networks (SOHO – Small Office Home Office)
The least customized type of network configuration is a standard flat (SOHO) network with no physical or logical separation of voice and data as shown in Option 3a below. It is not advised that this type of network configuration be used for networks with 10 or more SIP devices.
Bursty traffic patterns and other unforeseen data usage can impact voice traffic on the LAN as well as on the ISP connection. Any time this type of network configuration is used it is critical that a QoS router be installed and be properly configured as seen in Option 3b below. Also
the ISP connection should be right-sized to adequately handle both voice and data traffic, and that the ISP connection be stable and consistent.
Network Configurations to Avoid The following diagram represents common possible network configurations that need to be avoided in networks where VoIP service are deployed.
In this example, multiple routers have been strung together and as a result there would be multiple device handing out IP address via DHCP and multiple devices providing NAT between networks. This will almost always cause VoIP issues. Whenever a device is needed to provide
additional uplink ports into a network a switch and not a router should be used. Keep in mind that if the modem is not set to “Bridge Mode” in this type of network configuration it will act as a router and will cause the same type of issues as having two or more routers/firewalls
If you have followed the guidelines and configurations covered n the sections above, your network should be setup and ready for installation and operation of Hosted PBX services. Should you have issues with installation or ongoing operation the following are some basic steps to determine what might be causing and then resolve the issues you are experiencing.
Check Registration Status of Endpoint
As previously described, all CoreDial services are SIP based registration. Device registration can be confirmed in the following ways:
- Check the web interface of the phone or device in question and confirm that it is showing as “Registered”
- Log into the Hosted PBX Account Manager and check the SIP Peer Status under Services Tab -> Extensions (on left side menu) ->Extension List
Typical causes of failed registration are misconfigured SIP devices (phone, ATA, customer PBX in the case of SIP Trunk) or local network issues preventing the SIP device from accessing the network and sending SIP registration request to the Hosted PBX.
- Confirm Device Reachability (Hosted PBX)
There are certain circumstances where you may find that devices will register but then intermittent issues arise with the phone not appearing online or with inbound calls or transfers going straight to voicemail instead of ringing the phone. In most cases this is caused by the Hosted PBX inability to reach the phone or device in question. This can be confirmed by looking at SIP Peer Status in the Account Manager portal. If the
phone or device is not reachable you will see a red phone icon in the Extensions List. 99.99% of the time reachability issues are caused by a configuration issue on the customer router or LAN (Local Area Network and are the result of the router/firewall closing the session between the Hosted PBX and the phone. In order to help resolve this issue make sure that you have done the following 3 steps:
- Configure QoS on the router to minimize delay in responses from the phone to the Hosted PBX
- Configure the router with the settings we discussed earlier in this document
- Make sure you do not have multiple devices providing NAT in the path between the phone and the CoreDial cloud.
- Test an Extension or Device from another Location (This is one of the most important test you can run). If a device is having an issue at a particular location move it to another location on a different router and ISP connection in an attempt to isolate whether the
issue is the network, the ISP, the router, or the device itself.