Question 21/XI - Updating and enhancements of ISDN user-network interface
data link layer protocol

Considering

(a) that the 1988 (Blue Book) version of Recommendation Q.921 on LAPD
(ISDN user-network interface data link layer protocol) is firm and
stable, however, it may require updating in order to correct for
possible errors leading to misoperations;

(b) that there is need to define a new data link layer protocol, based
on LAPD, to satisfy the needs arising from:

-application of the D-channel in a symmetrical fashion (e.g., NT2 to
NT2),

- application to channels other than the D-channel (e.g., for data
transfer),

-application to Additional ISDN packet mode bearer services (Step 2 -
frame multiplexing and Step 3 - frame relay or frame switching, with
out-of-band signalling);

(c) that general signalling requirements and principles will be provided
from the study under the other Questions (e.g., service definition,
network architecture definition);

(d) that there may be a need for additional layer management functions
for both D- and non-D channel applications;

(e) that a desire was expressed to provide user guidelines for
implementors, especially for the implementation of the LAPD protocol on
the primary rate interfaces (23B+D, 30B+D),

What additional Recommendations andor enhancements to the
Q.920-Series of Recommendations are necessary?

The attached annex contains agreements reached during the previous study
period; it is structured as an expedient to allow easy comparison with
Recommendation Q.921, but the exact structure of a possibly new
Recommendation in the Q.920-Series is a matter for further review.

Study points should include:

1. symmetry (for user-to-user data transfer); use of CR bit;

2. structure of the address field;

3. need for HDLC additional frames (e.g., Test, Reset);

4. explicit congestion control (i.e., frame relay bearer service);

5. synchronization between the User (U) and Control (C) plane (e.g., by
primitives);

6. data link layer structure and modelling issues;

7. compatibilitycommonalityinterworking with Q.921 and possibly with
other HDLC elements of procedure;

8. definition of layer management functions and managed objects needed
to support these functions.

ANNEX

(to Question 21/XI)

A draft outline for an ISDN data link layer
specification for non-D-channel applications

Attachment 1

(to the Annex to Question 21/XI)

1. Dynamic window algorithm

The dynamic window algorithm is a way to control network congestion
(circuitswitched connections between two end terminals have no need for
this procedure). The algorithm modifies the transmitting datalinklayer
entity's send window when congestion is first detected, and again as
congestion decreases. The receiving datalinklayer entity does not
participate in the algorithm and does not require knowledge of the
sending datalinklayer entity's participation. Congestion in one
direction of a link is treated independently of congestion in the other
direction.

1.1 Operation

If a datalinklayer entity's transmit window parameter (k) is set to 1,
the working window parameter V(k) will always have the value 1 and the
algorithm need not be invoked. If the datalinklayer entity's k is
greater than 1, it uses a V(k) equal to k in the absence of congestion.

This congestion control algorithm is triggered by the loss of I frames.
A datalinklayer entity detects this loss:

-when it receives a REJ frame;

-when its T200 expires and it transmits a command with the Pbit set to
1, and subsequently receives an I frame response or supervisory response
in which the Fbit is set to 1, but in which the value of N(R) is less
than the current V(S).

When a datalinklayer entity detects either of these events, it invokes
the dynamic window algorithm, setting its V(k) to 1. As I frames are
then successfully transmitted and acknowledged, the transmit window size
V(k) is gradually increased until it returns to k, its value in the
absence of congestion. Several algorithms for control of the increase in
V(k) are possible. Prior discussions included a procedure in which the
rate of increase in V(k), i.e. the increase in V(k) per transmitted and
acknowledged I frame, decreases as V(k) approaches k. The procedure
described here is one in which the rate of increase in V(k) is constant.

Following reduction of V(k) to 1, when Nw consecutive I frames are
successfully transmitted and acknowledged, V(k) is increased by 1. For
each additional Nw I frames that are successfully transmitted and
acknowledged, V(k) is increased by 1. When V(k) reaches its maximum
value, k, the dynamic window algorithm ends.

1.2 List of system parameters and variables

In this algorithm, the following additional system parameters and
variables are defined:

1) Transmit working window (V(k))

The maximum number of sequentially numbered I frames that may be
outstanding (that is, unacknowledged) at any given time. In the absence
of congestion, V(k) is equal to the maximum number of outstanding I
frames (k). Initially, V(k) = k.

2) Dynamic window step size (Nw)

The number (Nw) of I frames that must be successfully transmitted and
acknowledged before the transmit working window is incremented in the
dynamic window algorithm is a system parameter. The default value is 5.

3) Information acknowledge counter (Ia Ct)

This counter contains the number of I frames successfully transmitted
and acknowledged since the last adjustment of the transmit working
window Ww, in the dynamic window algorithm.