QoS and Frame Relay

Vesa Kosonen
Helsinki University of Technology
Laboratory of Telecommunications Technology
P.O.Box 3000
02015 HUT, Finland
Email: [email protected]

Abstract it can be mentioned that there are many research projects

going on where frame relay is involved. As we can see
The purpose of this paper is to study quality of service frame relay is still in the headlines. What makes frame
properties of frame relay. In this paper we first relay as an interesting alternative for transport
concentrate on explaining the protocol architecture and technology? How is quality of service maintained in case
related matters. The frame format is studied in detail. of congestion? These are some of the questions that will
The behavior of frame relay network is under be discussed in this study.
consideration, too. How does the frame find its way to
the destination and what happens when there are errors? 1. A Short History of Frame Relay
Then we will study some properties unique to frame
relay. Finally we take a look how congestion control is In 1988 came out I.122 Recommendation entitled as
realized in frame relay to keep up with high quality of “Framework for Providing Additional Packet Mode
service. Bearer Services”. It introduced a new form of packet
transmission that has become one of the most significant
Introduction contributions of narrowband ISDN work. The title of this
recommendation was changed with the release of the
The rapid growth of Internet sets new requirements to 1993 version to “Framework for Frame Mode Bearer
the networks used for the services of Internet. Originally Services”, which we nowadays know as frame relay. The
best-effort type of Internet faces today demands for former term emphasizes the service being offered to the
quality of service. As new applications are coming into user, while the latter emphasizes the protocol that
existance the capacity of the existing networks has to be implements the service. In Table 1 can be found the
developed, too. There needs to be less errors, bandwidth ITU-T documents related to frame relay [2].
needs to grow manifold, delay and jitter should be as
small as possible, they should be able to endure Table 1. ITU-T Recommendations on Frame Relay
congestion, etc. In other words, transmission properties
of networks should be of high quality as they try to serve Number Title Date
the growing demands of the customers. I.122 Framework for Frame Mode Bearer 1993
Services
As an excample lets consider how to transmit voice over I.233 Frame Mode Bearer Services 1992
Internet Protocol (VoIP), which is one of the hottest I.370 Congestion Management for the ISDN 1991
topics of research today. Since voice is very sensitive to Frame Relaying Bearer Service
delay and jitter these properties of Internet has to be I.372 Frame Relay Bearer Service Network- 1993
developed before VoIP will gain support from the big to-Network Interface Requirements
audience. The new technology needs to offer something I.555 Frame Mode Bearer Service 1993
new and better to the existing telephone system and at Interworking
the same time try to maintain the good old level of Q.922 ISDN Data Link Layer Specification 1992
quality of the voice. for Frame Mode Bearer Services
Q.933 Signaling Specifications for Frame 1995
Frame relay is an old technology and it was initially Mode Call Control
developed to function as the transmission media for
narrowband ISDN. Since that time it has much Packet switching technology was developed at a time
developed and has today also many independent when digital transmission facilities exhibited a relatively
applications outside ISDN. It e.g. supports SONET/SDH high error rate compared with today’s facilities. As a
enabling higher bandwidth capabilities in the frame relay result there is a considerable amount of overhead built
network. No longer will users, service providers and into packet switching schemes (e.g. X.25) in order to
equipment vendors have to turn to other technologies compensate errors [2].
just to access greater bandwidth [1]. As another example

1
With modern, high-speed telecommunications systems, • Flag (1 oct.): At the beginning and in the end there
this overhead is unnecessary since the amount of errors is a bit pattern 01111110 to tell the limits of the
has been dramatically lowered. Frame relay was frame.
developed to take advantage of the high data rates and • Address (2-4 oct.): Contains the Virtual Circuit
low error rates [2]. Frame relay was originally designed address and extra bits for congestion control.
to be used for speeds up to T1/E1 speeds (2 Mbps), but • Information (variable): Contains higher layer
has already been demonstrated to be practical for speeds protocol data. This field can contain any integral
up to 50 Mbps and even higher. number of octets. The frame relay standards suggest
that the minimum length should be 1600 octets in
Today the main usage of frame relay is to connect connection of LAN interconnection applications.
LAN/WAN networks. Other services include image Most frame relay services support a maximum
transfer, private line replacement and Internet access [3]. length of 4096 octets.
Frame relay has become very popular and many • FCS (2 oct.): The frame check sequence [3]
applications are being developed based on it.
Note the absence of a Control field in this frame. Since
2. Protocol Architecture frame relay treats all frames the same regardless of their
type and does not provide any guarantee of sequentially,
The protocol stack of frame relay is simple, only one and there is no reason for frame relay to ever examine a
a half layers. The protocol layers used are the Physical Control field. If one is present, LAPF’s core protocol
Layer and a subset of the Data Link Layer, called LAPF would consider it to be part of a higher-layer protocol
core (LAPF = Link Access Procedures to Frame Mode [3].
Bearer Services) which is defined in Q.922. LAPF is
based on and is an extension of LAPD, which is used in The address field contains information necessary for the
ISDN (Figure 1). operation of the frame relay service. This field contains
addressing information (DLCI) as well as congestion and
Data Link LAPD fairness indicators (C/R, EA, FECN, BECN and DE)
Layer LAPF core (Figure 3). The default length of this field is two octets.
Physical Physical Physical Three or four octet address field may be employed, too
Layer Layer Layer [3].
OSI ISDN Frame Relay 6 oct. 1 oct. 1 oct.
Figure 1. Comparing Protocol Stack of OSI, ISDN DLCI (high order) C/R EA0
and Frame Relay DLCI (low order) FECN BECN DE EA1
4 oct. 1 oct. 1 oct. 1 oct. 1 oct.
The Physical Layer is no different from any Physical Figure 3. Address Field Format – 2 octets (default)
Layer with definitions of how bits are transmitted. The
Data Link Layer on the other hand provides some of the Explanations to Figure 3:
same functions as defined in OSI model such as framing, • DLCI = Data Link Connection Identifier (10 bits)
addressing and bit error detection. The difference is that • C/R = Command/Response bit (1 bit)
there is no sequencing or no acknowledgements. In case • EA = Address Field Extension bit (1 bit)
of errors the frame is simply discarded. Since frame
• FECN = Forward Explicit Congestion
relay has no error correction it assumes that the network
Notification(1 bit)
infrastructure is relatively error-free (e.g. fiberglass).
• BECN = Backward Explicit Congestion Notification
Multiplexing is also performed at the Data Link Layer
(1 bit)
[3].
• DE = Discard Eligibility (1 bit)
2.1 LAPF Frame Format The Address Extension (EA) bits are used in accordance
The operation of frame relay for user data transfer is best with extended address fields. If EA is set to 0 it means
explained by beginning with the frame format (Figure 2). that more octets follow in this field. EA value 1 on the
other hand indicates that this is the final octet [3].
Flag Address Information FSC Flag
1 oct. 2-4 oct. variable 2 oct. 1 oct. 3. The Frame Relay Network
Figure 2. LAPF Core Formats Frame relay provides an unreliable, connection-oriented
service to the user. It means that a virtual circuit (VC)
has to be established before any data can be sent, but
network doesn’t give any guarantee that the data will be
delivered to the destination [3]. There are two options for

2
VC’s to choose from: Permanent Virtual Circuit (PVC)
or Switched Virtual Circuit (SVC). PVC is the original
VC, but SVC is gaining more popularity due to its
flexibility. Frame
Valid Discard

DLCI DLCI
#762 #18 Host Known
DLCI Discard
A
Host C
DLCI

DLCI Figure 5. Nodal Processing in Frame Relay

#23 Host
DLCI DLCI
#515 #23 DLCI Thus, frame relay switches do not look at the user data
Host #917 packets, which makes the network transparent to all
protocols operating at levels above OSI level 2 [4].
Figure 4. Frame Relay Network
4. Some unique aspects of Frame
The frame relay network consists of a group of Relay
interconnected nodes (switches), which relay the frame
relay data across the network on the appropriate VC While frame relay protocol is relatively simple, it is very
(Figure 4). A frame relay switch uses only the DLCI different from the bit-oriented protocols employed in
(only local value) information contained in the frame X.25 and on the ISDN D-channel. This section explains
relay header to forward the frame across the network to some of the unique aspects of the frame relay protocols
its destination. The path through the network is [3].
transparent to the user. The DLCI does not include any
description of how the connection transverses the 4.1 Committed Information Rate (CIR) and
network, or routing topology of the network [4]. Class of Service Parameters
A frame relay network operates at an OSI Layer 2 router One of frame relay's main benefits is that it makes a pool
network. Each frame relay access node puts the routing of bandwidth available to many VCs. However, there is
information (destination DLCI) in the data link layer a danger that some applications consume all of the
(frame relay header) of the frame. In other words, the available bandwidth leaving nothing or only small
frame relay network nodes look only at the header and amount of bandwidth to other applications. To prevent
the FCS. This is important because the network will not this to happen frame relay has a value called Committed
be effected if the data is eg. encrypted [4]. Information Rate (CIR). The purpose of CIR is to
provide fair access to the network's bandwidth by all
The frame relay switch (node) uses a two step review user applications [3].
process to forward frames across the network (Figure 5):
Imagine that the available bandwidth is 64 kbit/s.
1. Integrity of the frame is checked using the Frame Application A needs much of the bandwidth most of the
Checker Sequence (FCS), and if an error is time and application B needs only a small amount of the
indicated, the frame is discarded. bandwidth every now and then. Application A is
assigned 48 kbit/s CIR and application B is assigned 16
2. The destination DLCI address is validated, and if it kbit/s CIR. This arrangement provides guaranteed access
is invalidated the frame is discarded [4]. for both of the applications [3].
All frames that are not discarded as a result of the FCS
But is this good use of resources? Application A has
or DLCI checks are forwarded. The frame relay node
only 48 kbit/s available even it would need a lot more
makes no attempt to correct the frame or to request a re-
while application B's 16 kbit/s stays unused. The
transmission of the frame. This makes for an efficient
solution to this problem is to allow application A to
network, but requires that the user end-stations assume
exceed its CIR. In our example it would mean that
responsibility for error recovery, message sequencing
application A is allowed to send traffic up to 64 kbit/s.
and flow control [4].
To protect the application B, however, all the traffic that
exceeds application A's CIR would be marked as discard

3
eligible by setting the DE-bit in the frame address field be reduced. Otherwise the Frame Relay nodes will start
(Figure 3). In this way the applications that need to send discarding frames to maintain internal buffer levels. If
traffic only now and then is guaranteed bandwidth while end-points do not reduce their transmission rate in this
the applications that need to send much traffic are situation, "packet storms" can result as stations
allowed to do so. Bandwidth across the access line never retransmit discarded data onto an already overloaded
goes unused as long as some applications have data to network [5].
send (Figure 6) [3].
Figure 7 shows the problem of congestion graphically.
During periods of no congestion, network throughput is
able to keep pace with the incoming traffic. At some
level of incoming traffic (point A), the network may
discard some frames due to localized congestion.
Network throughput decreases gradually once this period
of mild congestion begins (from A to B). As a result two
responses may occur. First, end users who have lost
frames will request retransmission, adding even more to
the traffic load in the network. Second, congestion in the
network results in delays at end-user equipment, causing
timeouts and, again, requests for retransmission. As a
result the incoming traffic will increase to point B, where
Frame 1 Frame 2 Frame 3 Frame 4 severe congestion begins. If this point is reached,
within CIR within CIR marked “DE” discarded at ingress
network throughput will degrade rapidly. To offset the
Figure 6. Relationship between CIR, DE and other negative effects of congestion, frame relay’s congestion
Class of Service Parameters. control strategy is to slow down the rate of incoming
traffic if the level reaches point A; in that way, incoming
Explanations to Figure 6: traffic level will never reach the potentially crippling
• Tc= the Committed Rate Measurement Interval, is point B [3].
the interval of time over which information transfer
rates are measured, normally 1 s. Congestion control is normally the function of the
• Bc=Committed Burst Size, is the maximum number Network Layer prototocol, but since frame relay doesn’t
of bits the network guarantees to deliver during the have one it is the function of the Data Link Layer.
time interval Tc under normal circumstances. Congestion control can be implemented in a frame relay
• CIR is defined as "the troughput rate that the network in the following ways:
network agrees to support under normal conditions" • use of the Explicit Congestion Notification bits
or CIR = Bc/Tc [bit/s] • the Consolidated Link Layer Management Protocol
• Be= Excess Burst Size, which is the maximum • Implicit Congestion Notification [3].
number of bits above the CIR that the network will
attempt to deliver during Tc (DE=1).
• EIR= Excess Information Rate, EIR=Be/Tc
• All the frames that exceed Bc+Be will be discarded.
• To avoid frame discards: Access Rate = Bc + Be

Since most data transmission is bursty in nature, it

actually means that most of the information is
transmitted in a relatively short period of time.

5. Congestion Control in a Frame

Relay Network
Frame Relay uses statistically multiplexed virtual
circuits within a single physical bearer. As we have seen,
each circuit may be configured to have a guaranteed
throughput (CIR-value) while allowing for bursts of
extra traffic (DE-marked frames). Frame Relay supports
Congestion Management, a feature whereby the network Figure 7. Throughput and Network Congestion
attempts to notify end-points that the network is
experiencing congestion and the volume of traffic should

4
5.1 Explicit Congestion Notification (ECN) 3. If a consecutive sequence of frames with the
BECN bit set to zero are received, then gradually
In frame relay there are two congestion bits in use increase the flow of frames [2].
(Figure 3 and Figure 8):
In general terms, for explicit congestion avoidance, the
• FECN - Forward Explicit Congestion Notification network alerts end systems to growing congestion within
• BECN - Backward Explicit Congestion Notification the network and the end systems take steps to reduce the
offered load to the network [6].
The FECN bit is set to notify the receiving end system
that the marked frame has encountered congestion. In 5.2 The Consolidated Link Layer
response to this, the receiving system should try to
reduce the flow of data from the sending system on this Management Protocol (CLLM)
frame relay connection. The mechanism for doing so There is a potential deficiency when using the ECN bits:
must be above the level of the frame relay bearer service, How does the source become informed about congestion,
which provides no direct flow-control facilities [2]. if there are no frames being sent back to it by the
destination? For this kind of situation there exists CLLM
User message. It is generated by the network and sent to the
source host to inform to reduce the number of frames
that it is sending. CLLM message is sent in LAPF
frames on DLCI 1007 and indicates the list of all
affected VCs as well as the cause and expected duration
of the congestion [3].

5.3 Implicit Congestion Notification (ICN)

Implicit signalling occurs when the network discards a
User frame, and this fact is detected by the end user at a
higher, end-to-end layer, such as Q.922 control protocol.
When this occurs, the end-to-end software may deduce
Figure 8. Use of FECN and BECN bits. that congestion exists. Once congestion is detected, the
protocol uses flow control to recover from the
In general terms, the receiving end system should use the congestion [2].
following strategy for each connection:
1. Compute the fraction of frames for which the
FECN bit is set over some measurement interval.
6. Conclusions
2. If more frames have the FECN bit set than have Frame relay provides an unreliable, connection-oriented
an FECN bit of zero, then reduce the flow of service to the user. It uses Data Link Connection
frames from the sending system. Identifiers in travelling through the network. Frame relay
3. If the congestion condition persists, institute was invented to utilize the new technology in the field of
additional reductions. transmission facilities. Frame relay’s advantage is light
4. When the congestion condition ends, gradually protocol structure which makes fast connections
increase the flow of frames [2]. possible. E.g. congestion control of frame relay operates
exceptionally at the OSI layer two.
The BECN bit, on the other hand, is set to notify the
receiving end system that the frames it transmits on this Mechanisms of maintaining the quality of service are
connection may encounter congestion. In response to tested especially in the situations of heavy traffic. The
this, the receiving end system should reduce the flow of important question is how does the protocol resolve the
data transmitted on that connection. congestion conditions. In case of frame relay network
the method is to reduce the incoming traffic by notifying
In general terms, the receiving end system should use the the end systems about the congestion. The notifying bits
following strategy for each connection: are built in the protocol structure of frame relay.

1. When the first frame with the BECN bit is Frame relay has become an interest of many research
received, reduce the information rate to CIR. projects and will be one of the most popular transport
2. If additional consecutive frames with the BECN technologies used in the Internet in the future, too.
bit set are received, then institute additional
reductions.

5
References
[1] http://www.frforum.com/6000/FRF_14PR.html
[2] Stallings, William: ISDN and broadband ISDN with
Frame Relay and ATM, Fourth edition, Prentice-Hall,
Inc. 1999.
[3] Kessler, Gary and Southwick Peter: ISDN: concepts,
facilities and services, Signature (4th) edition, McGraw-
Hill series, 1998.
[4] http://135.145.192.160:80/cgi-
bin/auth.pl?file=framerelay_security_guide.htm&session
=35f6c3c61b015f2f#2-1
[5] http://www.datacraft.com.au/whitepapers/f_relay.htm
[6] Stallings, William: High-Speed Networks, TCP/IP
and ATM Design Principles, Prentice-Hall, Inc. 1998.