|
Home >
FAQs
Frequently asked questions about voice over Internet protocol and associated termsSomeday, all telephone systems will operate over data lines. Why? Because voice over IP (VoIP) is easier to manage, less expensive to install and maintain, and offers greater flexibility in connecting remote locations. With a network-based IP phone system, you can send voice and data simultaneously on the same line. You can say goodbye to many long-distance charges. Best of all, you can work smarter, using a host of cutting-edge
capabilities that blend voice and data to boost your productivity.
Latency
Jitter
Packet loss
Link bandwidth
Network utilization
Network analysis
Trace route
Collisions
Latency
Latency is the time it takes for a packet to travel from one end of the link to the other (or, latency is one half the round-trip time to the link address).
Back to top
There are two reasons to consider latency:
- Latency adds delay to voice communication. If the latency is small enough, neither party will notice the resulting delay; but latency above a certain point will cause a noticeable delay that can be annoying. Even longer latency times can make an IP conversation difficult; each party must wait to take turns talking to avoid “talking over” each other.
- High latency (i.e., times greater than 250 ms) often indicates a
poor IP connection. Be sure to retest the link during high-traffic times if
all the test parameters except latency appear to be satisfactory.
Back to top
Jitter
Jitter is the measure of the variation from packet to packet in round-trip time. This measure is calculated as the standard deviation of the individual packets’ round-trip time. Ideally, the round-trip time and
latency of all packets would be identical; however, in practice, this rarely occurs. Due to other data traffic or bandwidth constraints, some packets get delayed
and take longer to make the trip. This variation is jitter.
Back to top
The IP PBX and the Remote Phone can compensate for some jitter; but, past a certain point, the Remote Phone can’t wait any longer to play a packet. When it is time to play a late packet as part of the voice stream and that packet hasn’t arrived, an audio anomaly occurs. The actual distortion depends on the specific data stream received, but can vary from a slight warbling, to popping and clicking, or — in extreme cases — a crackling sound.
Back to top
Packet loss
On poor or overloaded IP connections, the amount of data traffic — i.e., the number of packets — may exceed the capacity of the connection. When this occurs, packets are discarded (“lost”) by the router or host computer at the point of congestion. Packet loss can occur also on wireless and microwave LAN and IP links, due to RF interference. On a high-quality IP connection, packet loss may occur only rarely;
however, on a poor connection, packet loss can occur often.
Back to top
Unlike data traffic, voice communication is very sensitive to packet loss. In IP data traffic, devices detect packet loss and simply retransmit the lost packets automatically; this process works so well for data traffic, users are likely to be unaware of significant packet loss on their IP connection. However, such retransmission of lost packets is not an option with voice over IP: the latency resulting from detecting and retransmitting a lost packet would cause the retransmitted packet to be unusable. Any lost
voice packet is lost for good; so, if packet loss occurs during speech, distortion will occur. Even a single lost packet can result in an audible pop or click. Significant packet loss will result in a crackling sound.
Back to top
- Obviously, zero packet loss is preferred. This can be achieved on high-quality IP connections.
- Packet loss of less than 1% will be acceptable with most users. The user may only occasionally notice the distortion due to packet loss of less than 1%.
- With packet loss of between 1% and 2%, the user will become increasingly aware of the distortion due to missing packets.
- Packet loss of 2% and higher will cause noticeable and definite distortion. This level of distortion will likely be unacceptable to most users.
Back to top
Link bandwidth
Link bandwidth is the measure of the amount of additional data that can be moved across the link in a second — i.e., what’s available for use by an IP Phone system after data traffic is taken into account. The unit of measurement is Kbps (kilobits per second).
The measured bandwidth will be less than the total bandwidth of the link by the amount of bandwidth consumed by the data traffic (i.e., M = Total — Data). It is useful to measure the link bandwidth during business hours, when typical or heavy data traffic is being experienced. For a new site, this additional bandwidth is what’s available for adding Remote IP
Phones.
Back to top
During a call to a Remote Phone, voice packets travel in a steady stream in both directions. There must be adequate capacity — bandwidth — in the link for each of those packets to arrive at the opposite end with no more than an acceptable amount of latency and jitter. When an IP link is
near-capacity or overloaded, some packets can be significantly delayed and discarded; as a result, a Remote Phone connection on such an IP link will suffer from serious audio distortion. Even a site with high-end broadband access may not have adequate available bandwidth, if that broadband access is heavily used.
Back to top
Each remote network channel consists of a 22 Kbps packet flow in each direction. To prevent overloading the link, voice traffic should consume no more than 33% of the available bandwidth. The following table provides a guideline for the minimum bandwidth for the number of remote channels planned:
Number of remote
network channels |
Recommended
bandwidth (Kbps) |
| 1 |
66 |
| 2 |
132 |
| 3 |
198 |
| 4 |
264 |
| 5 |
330 |
| 6 |
396 |
| 7 |
462 |
| 8 |
528 |
| 9 |
594 |
| 10 |
660 |
| 11 |
726 |
| 12 |
792 |
If the measured link bandwidth is less than the recommended amount, you can expect voice quality problems. Also, a 30% or greater variation in the measured link bandwidth on multiple checks may indicate that the link is frequently in an overloaded condition. Voice quality may also be affected in such cases.
Back to top
Link performance problems typically result from one of the following:
- Connection bandwidth at one of the sites is inadequate for the site’s current amount of data traffic. Adding voice traffic will require upgrading the broadband access.
- The pathway through the ISP’s local service network adds significant latency or jitter. Upgrading service or switching ISPs may be required to resolve these problems.
- Connection is routed between different IP networks. If the sites at each end of the link are connected
to different ISPs, the ISPs may use different IP networks as their default entry point into the Internet; this may result in excessive hops and routing. The best approach in this situation is to have both sites use the same ISP.
Back to top
Network utilization
Utilization is the percent of time that the monitored network segment is occupied by data traffic — either voice data or computer data.
Back to top
A network, or network segment, is a common resource shared by the devices attached to the network. Only one device can transmit onto the network at a time. If a device, such as a PC, is ready to send data on the network, it must wait until the network is idle before sending data. If the utilization of the network is low (5% or under) a network device rarely waits for the network to be available. As the utilization increases, devices have to wait more often for other devices to finish using the network.
In computer data traffic, high utilizations (30% or over) can cause a slowdown in data transfer, because of the increasing amount of time the network’s PCs spend waiting for the network to be available.
High network utilization can be very detrimental to voice data. The voice data to an IP Phone, or IP PBX, consists of a stream of packets at regular intervals. If the IP Phones and IP PBX are forced to wait too long to transmit a packet, the audio stream may be interrupted at the receiving end. The result will be degradation in audio quality. Pops or clicks may be heard in this situation.
Back to top
Network utilizations below 30% are suitable for IP Phones and IP PBXs. Network utilizations between 30% and 40% may result in some audio distortion; network utilizations above 40% are more likely to result in audio distortion that will be unacceptable.
Network utilization is especially critical on 10 Mb networks. Due to the limited capacity of a 10 Mb network, a modest amount of additional load can put the network into an overloaded situation.
On evaluating a potential site, the added network utilization resulting from the installation of the IP Phone system must be considered in evaluating the network. Use network analysis to perform this evaluation.
Back to top
Network analysis
The mere fact that a prospect’s network currently supports the prospect’s computer data doesn’t mean the network also can support an IP Phone system. If the network is inadequate, you should include upgrading the network (or otherwise remedying the problem) as part of the installation proposal. Performing network analysis before proposing the system will help you avoid installing an IP Phone system and receiving complaints of the system’s poor audio performance, only to discover later that the problem
really was because the network hadn’t been properly upgraded to handle the additional traffic.
The expense of upgrading a network up-front is minor, compared to the trouble and effort of diagnosing an overloaded network, later, after a customer is experiencing problems.
Back to top
Probably not. For one thing, you must be conservative when evaluating a network for initial installation of an IP Phone system, because the customer is likely to add IP Phones, computers and other network devices in the future. And, that aside, only a few IP Phones can be properly supported on a 10 Mb network. Unless the network traffic is very limited, a 10 Mb network will probably not support an IP Phone system. Plan on upgrading a 10 Mb network unless there is a small number of users (fewer than 10) and no plans (and/or expectations) for growth.
Back to top
Two approaches to consider in addressing an inadequate network are to upgrade the network and/or eliminate unnecessary traffic from the network.
- Upgrading the network
— The goal of upgrading a network is to reduce network utilization (as a percentage) while still supporting all necessary data and voice traffic. In many cases, installing dual-speed 10/100 Mb multi-port Ethernet switches (in place of hubs) will accomplish this goal. A dual-speed switch will allow segments to run at different speeds, 10 Mb or 100 Mb. As an example of how this can help: some PCs’ NICs are capable of only 10 Mb traffic and, thus, may be slowing down
the entire network; but the dual-speed switch will allow those 10 Mb NICs to slow down only their segment, while other segments can run at 100 Mb. (Important: Plan on putting the IP PBX on a 100-Mb segment.) Installing an multi-port Ethernet switch will segment the network so that data and voice packets are transmitted on only the source and destination segments. This will significantly lower network-wide
utilization. - Eliminating unnecessary network traffic
— You may find that unnecessary protocols are generating excess network traffic. Windows loads three protocols by default: NetBEUI, TCP/IP, and IPX/SPX. Both IPX/SPX and NetBEUI will generate unneeded traffic when not used; however, TCP/IP won’t. Normally, either (a) TCP/IP will be the only protocol or (b) TCP/IP and one other protocol [such as IPX/SPX] will co-exist (in the latter case, TCP/IP will be used for external communications —
Internet access, e-mail, etc. — and the other protocols will be used internally for printer sharing and file transfer). Thus, e.g.: if the network is TCP/IP-only and you see that 10% of the packets on the network are NetWare and NetBEUI packets, you can immediately reduce the utilization rate by that amount by disabling those protocols on offending workstations.
Back to top
Trace route
As packets travel from one site to another, they typically pass through several routers (or host computers acting as routers). A trace route (such as the function by that name in the MS®-DOS command, TRACERT) queries each router along the path, starting with the first one, to determine the next step — or hop — along the route, and calculates the round-trip time to that router.
Back to top
Just as a chain is only as strong as its weakest link, so also will the performance of an IP connection be only as good as the poorest hop in the route. The number of hops in the route isn’t necessarily significant: if all the hops in the route are low-latency, high-bandwidth
connections, a route with 30 or more hops can provide an excellent link for
Remote Phones. However, just one poorly performing connection in a route
can cause a routing of only a few hops to be unacceptable for Remote Phone operation.
Back to top
Trace route information helps you isolate a weak link. After collecting the trace route information on a poorly performing link, review the round-trip times for each hop. Look for hops where:
- The round-trip time increases significantly from the previous hop.
- The minimum and maximum round-trip times differ significantly.
If one hop seems to be contributing significant latency or jitter, test the performance of this one location, specifically regarding jitter and bandwidth. If the performance to this IP address is significantly better than that for the entire link, the problem is probably at one of the higher-numbered hops; but, if the performance is no better, the problem is at one of the lower-numbered hops. Select another hop location and rerun the check. Work your way through the trace route to pinpoint which hop is causing the problem.
Back to top
Collisions
Collisions occur on a network when two devices start transmitting at exactly the same moment; the first device’s signals “bump into” or “collide with” those of the other device, hence the term collisions. When this happens, both devices abort the data transmission and wait for another opportunity to use the network. Other devices on the network segment will also sense the collision and may slow down the rate of transmitting data onto the network.
Back to top
Collisions happen occasionally on all networks become more frequent at high levels of network utilization; this is because, the more time devices spend waiting for the network, the more likely it is that another collision will occur with other devices also waiting to use the network.
A high level of collisions indicates that the network is overutilized. Much of the network bandwidth is being wasted during collisions and the amount of data that can be transferred is reduced. At extreme levels of collisions, measured utilization may be fairly low due to this effect.
Back to top
More than 100 collisions per second is an indication of a significant problem. Both voice and computer data performance will suffer from this level of collisions. The causes of a high level of collisions can include the following:
- Too many devices connected to a single segment.
- Devices on the network using too many different protocols.
- A malfunctioning device on the network.
Because a high level of collisions will result in poor voice quality from an IP Phone system, you must isolate and eliminate the cause of the high collision
rate before installing such a system.
Back to top
© 2007
DataVox Inc. | All Rights Reserved |
Terms of Use and Privacy Policy
713.881.5300 Main | 713.881.5353 Customer Service
FAQs |
