Congestion Caused In The Network

Published Date: 02 Nov 2017

Two major factors affect the voice quality, a) Loss of packets and, b) Delay in packets. Packet loss can cause voice clipping and skips. The standard codec algorithms in the industry used in Cisco’s Digital Signal Processor (DSP) can rectify up to 30 ms of lost voice. Here Cisco Voice over IP technology uses 20 ms samples of voice per VoIP packet. Hence, for the codec correction algorithms to be effective, a single packet could be lost.

The delay in Packet can cause either quality degradation in voice cause of end-to-end voice latency or packet loss if delay is adjustable. If end-to-end voice latency is too long, the conversation would sound like 2 parties talking on a radio, the delay is variable, and there is risk of jitter buffer over runs at the receiving end.

Excluding delays and drops is even more vital when including modem and fax traffic over IP networks. If packets get lost during modem or fax transmissions, the modems are forced to wait and synchronize again. By studying the reason of packet delay and loss, we can gain an understanding of why Quality of Service (QoS) is needed in most of the areas of the network.

INTRODUCTION TO QoS

We shall discuss about what is QoS and Why we need it. The description would be given in the following headings and the details below each.

What is the need for QoS?

While we make calls over the Intranet or Internet, there are factors that are need to be kept in mind for instance the main two factors are:

Lost Packets

Voice clipping and skips are caused cause of packet loss. Voice over IP (VoIP) technology uses 20-ms for voice payload for VoIP packet. Therefore, for codec correction algorithms be effective, only single packet could be lost during any given time

Delayed Packets

Packet delay could cause voice quality degradation cause of the end-to-end voice latency / packet loss if the delay is variable. If end-to-end voice inactivity is too long then the conversation began to sound like 2 parties talking on CB radio. If delay is variable, there is risk of jitter buffer to overrun at the receiving end. Eliminating drop and delay is even more imperious when sending fax and modem traffic over IP networks. If packets are loss during fax or modem transmissions, they are forced to wait to synchronize again.

Hence by examining the cause for the packet loss or delay in packets we can understand the issue and therefore the issue can be corrected. For doing this we need QoS, which basically is needed in all fields, not only IP Telephony.

Quality of the Network

Packets of voice could be dropped if the quality of network would be poor or if network is jammed, if too much variable delay is there in the network. Poor network value can lead to terms regularly going out of service, as lost logical or physical connections. As IP Telephony design and application is based on the assumption of the logical and physical network follows sound design methodologies and is very stable.

Congestion caused in the Network

Congestion in Network can lead to both dropping packets and packet delays variably. Voice packet are dropped from the network congestion and are frequently causing full transmit buffers of the egress interfaces in the network. As the links or connections reach cent percent utilization, the servicing of queues for those connections become full. New packets are discarded when new packets enter. The interface queue of Egress wait time or large series delays cause variable delays of the type.

Overview of Jitter and Delay

The time taken for a packet to reach the receiving end after transmitted from sending to the final point is called delay. End-to-End delay consists of 2 components:

Fixed network delay

Variable network delay.

Jitter is delta, difference, in fulfillment end-to-end delay valued of 2 voice packets in voice flow. Fixed network’s delay must be examined during the first design of IP Telephony network. The standard G.114 states "a one-way delay budget of 150 ms is acceptable for high voice quality". Egress queues that are congested and series delayed over a network interfaces could cause variable delay. Priority and Low-Latency Queuing (LLQ), queuing delay time equal series delay times as link utilization approaches cent percent. Series of delays is a function of link speed and size of packet. The bigger the packet the slower the speed of link, greater the series delay. If voice must wait for data packet to serialize, delay gained by voice is its own series delay plus the series delay of the data packet ahead of it. This is called jitter, the variation among packets is intended to arrive and when this actually is arrives. To recompense for the delay variations between voice packets in a discussion, VoIP uses jitter buffer to change the variations of delay into a value that is constant such as voice can be played smoothly.

Tools for QoS

The quality of the voce would be as clear as the quality of the network. Loss in packets, delays, and variation in delay, contribute to sullied voice quality. As congestion in the network can occur at any portion, the network quality is end-to-end issue. The tools that are used for QoS are a set of approaches to improve voice quality on data networks by decreasing dropped voice packets during times of network congestion and by minimizing both the fixed and variable delays encountered in a given voice association. The QoS tools can be categorized as:

Classification

Classification tools mark a packet or flow with a specific importance. This assigning establishes a trust border that should be enforced. This should take place at the network edge, usually on the wiring closet, within IP phones or voice end points themselves.

Queuing

Queuing tools assign a packet or flow to one of several queues, based on classification, for appropriate treatment in the network. When data, voice, and video are placed in the same queue, packet loss and variable delay are much more likely to occur. By using multiple queues on egress interfaces and placing voice packets into a different queue than data packets, network behavior becomes much more predictable.

Network provisioning

Network Provisioning tools precisely calculate the requisite bandwidth needed for voice, the data traffic, video applications and required link management overhead such as routing protocols. While calculating a required bandwidth for operating voice over a Wide Area Network, it is important to remember that all the application traffic (that is, voice, video, and data traffic), when kept together, should equal not more than seventy five percent of the provisioned bandwidth. The remaining 25% is used for overflow and administrative overhead, such as routing protocols.

Connection to IP Phones

There are four ways to connect an IP phone to a network.

Single cable,

Multiple cables,

Soft Phone application (over a PC), and

Separate switches for voice and data.

The details of each of these methods such as duplex settings, virtual LAN and IP addressing, classification and queuing etc, will be discussed later under each methodology.

Figure 3.1 Different methodologies to connect IP Phones

Using a Single cable for connection to a PC and an IP Phone

Using a single cable for both the PC and IP Phone on the CISCO AVVID network is common with the many companies. The major reason for doing so is to make the connections simple by using lesser cables, saving installation time, cost saving over switch ports and wiring closest. With these cost savings comes requirements for additional switch features, particularly where QoS is concerned. Specifically, the requirements include correctly configuring the Ethernet link speed and duplex, Layer 2 Class of Service (CoS), and queuing on both the IP phone and the wiring closet Ethernet switch.

Fig: 3.1.1 Single Cable connection and QoS areas of concerns

Buffer congestion can arise through 100BaseT full-duplex to 10BaseT half-duplex aggregation. During periods of intense traffic, the half-duplex nature of the connection can lead to packet loss from deferred packets due to excessive collisions on the segment. Both the switch and the Cisco IP Phone, which uses a priority queue for voice, will always send voice traffic first. However, the high-speed video stream will also be sending as many packets as possible. When either the switch or the phone attempts to send the voice traffic, it can encounter collisions when attempting to transmit, thus resulting in deferred voice packets. IP addressing is also an issue that has to be kept in mind while we are configuring the IP Telephony system. The major concepts in IP addressing is to allocate proper subnet masks, Utilize the subnet mask properly, the IP address should be utilized to the fullest. Classifying, or marking, traffic as near to the edge of the network possible is always been the integral part of a Cisco design architecture. When connected by a single cable, the IP phone is the edge of the managed network.

Multiple Cables

Use multiple cables to connect the IP phones if any of the following conditions apply to Cisco IP Telephony network:

Connect IP phones that don’t have a 2nd Ethernet port attaching a PC.

Create physical parting between voice, data networks.

Provide in-line power easily to the IP phones

Limit the number of switches that need UPS power.

Limit the amount of CatOS upgrades needed in the network.

Limit the Spanning Tree configuration in the wiring closet switches.

Since there is no Personal computer behindhand the IP phone when you use multiple cables, port speed and duplex settings are not as critical as with a single cable

While it is safe to use the same configuration as with a single-cable connection (in case a PC is plugged into the second Ethernet port on the phone this configuration is not required. The recommended configuration for using multiple cables to connect IP phones to the Cisco AVVID network is to use a differential IP subnet and VLANs for IP telephony. Because the IP telephone and any data PCs are on separate physical cables, queuing on the IP phone is not expected. Nevertheless, because the IP phone is however a managed device, classification should yet happen on the phone or entrance Access switch port. This classification for VoIP packets could be treated in a mixture of ways, depending on which computer hardware is utilized in the wiring closet switch.

Soft Phone application (over a PC)

Some companies deploy IP telephony using the Cisco Soft Phone application. Many PC network interface cards (NICs) do not currently set 802.1P CoS bits for Layer 2 classification. Even if the PCs did set Layer 2 frame markings, the vast majority of network administrators would resist "trusting" a user's PC. Because of these factors, Cisco SoftPhone currently classifies voice packets only at the Layer 3 IP header. In fact, all voice bearer packets originating from the Cisco SoftPhone application are marked with an IP Precedence value of 5. Of course, this marking requires a wiring closet Ethernet switch that is Layer 3 enabled, with multiple queues, to correctly queue these voice packets. Currently,

this limits Cisco SoftPhone designs to PCs connected to Catalyst 6000 switches with a Policy Feature Card (PFC) installed. For use with SoftPhone, all wiring closet switch access ports should be set to 100BaseT full-duplex. IP addressing is not an issue in this case because the SoftPhone application runs on a PC. Cisco SoftPhone is a PC application, and it currently marks voice traffic only at the Layer 3 IP header. This requires the access switch to be Layer 3 aware because of the need to prioritize voice traffic before the first uplink to the distribution layer.

Separate switches for voice and data

You might want to connect the IP phones to separate switches in the wiring closet. This can avoid the need to upgrade your current data switches, and it serves to keep the voice and data networks completely separate. This type of installation is very similar to the scenario that uses separate ports on the wiring closet switch. Because there is no PC behind the Internet Protocol phone in this type of installation, port speed and duplex settings are not as critical as in other types of installations. The recommended configuration for connecting Internet Protocol phones to separate access layer switches on a Cisco AVVID network is to use separate Internet Protocol address space and separate VLANs for IP telephony. The second switch is the latest installed voice over IP only Ethernet switch runs on a single VLAN. Trunking is not necessary, IP phone and the Ethernet switch.

Design concept for a campus

Until recently, the conventional wisdom was that the Quality of Service would not ever be the issue in an enterprise of campus cause of the burstyness of the network traffic also the capability in the buffer overflow. QoS tools are needed to managing the buffer in minimizing delay, loss, and also the delay variation.

Transmitting buffers have the tendency in filling the capacity of fast speed campus network cause of the bursty nature in the data network combination within high volume of the smaller Transmitting Control Protocol data. Typically, these drops are a lot than the single packet at any given course.

Traffic Management and Control Management

In networks with high traffic loads, managing the delivery of control traffic is critical and ensures the positive user experience with VoIP. The Cisco Internet Protocol Phones use the Skinny Station Protocol in communicating with the Cisco Call Manager. When the Cisco Internet Protocol Phone gets off the hook, it would ask for Cisco Call Manager on what’s to be done. Cisco Call Manager then instructs the Cisco Internet Protocol Phone to play the dial tone. If the Skinny Client manages and controls the traffic, then the call is dropped or may be delayed inside the network, thus users then would experience would be affected. The same logic is applied to any signalling traffic to contacting the phones or the gateways.

Skinny Call Control Protocol

A Skinny Call Control Protocol uses TCP to connect with a Call Manager. This uses a Real-time Transport Protocol over the User Datagram Protocol transport for the traffic with the other Skinny client terminal. Skinny Call Control Protocol is the design as the communicating protocol for hardware and for other systems like embedded systems, with the significant PC’s and the memory constraints.

H.323 Gateway

It is recommended from ITU Telecommunication (ITU-T) standardization sector that would define any of the protocols in providing audio and visual communications on the packet network. H.323 is a standard that addresses the signalling and the control of calls, bandwidth control in point to point, multimedia control and transport, and the multiple points in conferences. It is mainly implemented for video and voice conference equipment. It is used with many applications which are real-time, for example NetMeeting and GnuGK. It is also widely deployed by the service providers and also by the enterprises for video and voice services over Internet Protocol network.

Media Gateway Control Protocol

It is the only implementation for any Media Gateway Control Protocol Architecture (MGCPA) for control over the media gateways on Internet Protocol network and then the public switch telephone network.

QoS parameters configuration

This includes setting up multiple queues on all ports, configuring access to the queues, setting thresholds for traffic drops, and connecting the switch to the distribution or core layer.

Here we would be discussing about switch – Catalyst 6000 series.

The following commands enable QoS on the access layer Catalyst 6000 by performing these functions:

Give the command in enabling the switch wise QoS.

Inform the designated ports that all the QoS association with the given ports would be done on each virtual LAN basis.

Instruct IP phones to reconfigure CoS from a given PC within the Internet Protocol phones Ethernet ASIC.

The incoming traffic from Layer 2 CoS classification should be accepted.

Create the access list that would accept the incoming Layer 3 classification.

Write all the access lists to the hardware.

The access list is then mapped to the auxiliary VLAN

Now to verify if QoS is applied or not we need to check it with the following commands

show port qos <mod/port>

This command shows the QoS settings for the specified port.

show qos info runtime <mod/port>

This command shows QoS runtime information for the specified port.

show mac <mod/port>

This command shows Media Access Control (MAC) information for the specified port.

show qos statistics l3

This command shows summary QoS statistics for all ports.

show qos stat <mod/port>

This command shows detailed QoS statistics for the specified port. Scheduling on the queues are done by Round-Robin (RR) method. If we enable QoS and do not modify any of the transmit queue mapping then the switch would performance might affect as the traffic would be assigned to queue 1.

Designing A Branch Office

The traditional branch office design for up to some users consisting of the branch router, an Ethernet switch. A router can handle Internet Protocol routing and Wide Area Network connectivity. The local computers are given connection by an Ethernet switch which is small for the traffic is small and that also would connect the router. The two areas for voice quality that should be taken care of within a branch office:

Voice quality across the WAN

Voice quality within the branch office

Here we address the branch office design, Internet Protocol addressing, and also the voice value inside a branch office. Typical branch office are designed for not only a single Internet Protocol subnets are used for the office. Changing this configuration is seldom feasible as doing so affects an enterprise wide routing outline. Hence, actual branch office design must understand three Internet Protocol addressing options for Internet Protocol phones:

Configure two VLANs by dividing the existing remote office IP address space into subnets, and use 802.1Q for trunking if the router supports trunking.

Configure two VLANs by dividing the existing remote office IP address space into subnets, and use secondary IP addressing on the Ethernet interface of the router. Use a single IP address space and VLAN at each remote office.

Wide Area Network

Introduction

The main reason for preparing a lower cost ownership for migrating to converged voice,data and video network. A converged type of network can lessen the cost of an enterprise’s communication infrastructure and solid planning and also the design which is the required Cisco AVVID deployment. Nowhere is this fact more evident than when running VoIP over a Wide Area Network (WAN).

There are several other additional QoS tools which should also be kept in mind:

Link Fragmentation and Interleaving (LFI)

Traffic shaping

Network Provisioning

Call admission control

Link Fragmentation and Interleaving

In low-speed Wide Area Network connections, the requirement to have a mechanism for Link Fragmentation and Interleaving is necessary. A frame containing data could be sent to a physical wire at the series rate of the intersection. The series rates are the sizes of a frame divided by a clocking speed in the interface. Link Fragmentation and Interleaving tools are useful to divide larger data frames to a regularly size piece and this then interleaves the voice frames into a flow so as the delay at the end-to-end could be predicted precisely. This takes place on the jitter by preventing the voice traffic being delayed after large data frames.

Traffic Shaping

In Asynchronous Traffic Mode and Frame-Relay network, where physical access speed varied in between 2 endpoints, traffic shaping is used to prevent delay from congested network buffers caused due to these speed misalliances. Shaping of traffic is the tool that meters that transmits rate of the frames from the source router to the destination router. This type of metering is done at the value which is lower than a line or the circuit rate of a transmitting interface. This metering is done at the rate to reason for a circuit mismatch that is common in the current multiple-access, non-broadcast networks.

Network Provisioning

Provisioning network bandwidth would be a major component in designing the successful Cisco network. This then represents the least bandwidth required for a given link, and this should not excess 75% of the available bandwidth in the link. . The seventy five percent rules knows that some bandwidth is required for overhead traffic, such as in routing and 2nd Layer keep alive, and for additional application as such e-mail and Hypertext Transfer Protocol traffic.

Call admission control

It is a mechanism for ensuring that voice flows dose not exceed the ultimate provisioned bandwidth allotted for voice discussions. Later doing the controls to provision this network with the needed bandwidth to provision voice, data, and video, thus the importance to confirm that voice do not oversubscribe the share of bandwidth allotted to it. While most quality of Service mechanisms would be used to defend voice from data, admission of calling control would be used in protecting voice from voice.

Other QoS Tools for WAN

This section describes the following additional QoS tools, which can help ensure voice quality in WAN applications:

VoIP Control Traffic

TX-ring sizing

Compressed voice codecs

Compressed RTP (CRTP)

Voice Activity Detection (VAD)

VoIP Control Traffic

When allocating bandwidth for the IP WAN, do not overlook the Cisco Call Manager control traffic. In call processing design which is centralized, the IP phone uses a Transmission Control Protocol control connection to communicate with Cisco Call Manager. If not enough bandwidth is provisioned for these small control connections, callers might be adversely affected.

To ensure that this control and management traffic is important, Access Control Lists are used to categorize these on Layer 3 or 4.

TX-Ring Sizing

The RSP is a very inefficient QoS platform, especially with regard to modifying the TX-queue parameters. On all Point-to-Point Protocol (PPP) and Multilink PPP (MLP) links, TX-ring buffer size is automatically configured. On Frame Relay links, the TX-ring is for the main interface, which all sub interfaces also use. The default TX-ring buffer size is 64 packets.

Compressed Voice Codecs

To utilize as much of the limited WAN bandwidth as possible, VoIP uses codecs (coding-decoding algorithms) to digitize analog voice samples. Many codecs, such as G.729, can compress a 64-kbps call down to 8 kbps. These types of codecs, termed low-bit-rate codecs, are commonly used for voice calls across the WAN.

Compressed RTP

Compressed RTP (CRTP) compresses the 40-byte header of a VoIP packet to approximately 2 to 4 bytes. Compressed RTP works on a link-by-link basis and is enabled on Cisco routers using the ip rtp header-compression command.

Voice Activity Detection

In most transactions, only a single party is talking at the time. The VAD algorithm in the VoIP software examines the voice conversation, looking for these gaps in conversation. When a gap is discovered, no packets are sent, and the WAN bandwidth can be recovered for use by data applications.