Note: Descriptions are shown in the official language in which they were submitted.
CA 022247~6 1997-12-12
OPTIMIZATION AND RECOVERY TECHNIQUES IN IMA
NETWORKS
Field of the Invention
The invention resides in the field of the ATM inverse
multiplexing. In particular, it is directed to an ATM inverse
multiplexing technique which enables link selection to be performed at
any time according to one of a plurality of selection criteria.
10 Background of the Invention
It has been recognized that the T1/E1 rate (1.544/2.048 Mbit/s) is a
cost effective way of user access to an ATM network as well as
connection between ATM network switches. However, as ATM
technology for wide area networks is deployed more and more,
15 decmands for transmission links of a rate higher than T1/E1 are
increasing. Links of higher rates, such as T3/E3 (44.736/34.368 Mbit/s),
have been designed to meet these needs. However, the cost of T3/E3
links is still prohibitive in many cases and the ratio of cost versus
realistic utilization of the entire rate is not always attractive and fully
20 justified for new ATM end users and service providers. IMA (short for
inverse multiplexing for ATM) satisfies the need by using multiple
T1/E1 links which are grouped collectively to provide the service at a
higher rate.
Figure 1 and Figure 2 show two sample configurations in which
25 IMA is used. In this specification, inverse multiplexing for ATM and
ATM devices performing inverse multiplexing are collectively called
IMA. Figure 1 depicts a user access to a network through user network
interfaces (UNIs) and Figure 2 illustrates a link connection between
ATM switches through broadband inter-carrier interfaces (BICIs) or
30 private network to network interfaces (PNNIs).
Referring to the figures, the basic function of IMA device is to
work in pairs to take an ATM cell stream coming from the ATM layer,
send it over the multiple links by spreading cells over the available
links and ensure that the initial cell stream can be retrieved at the far
35 end. Thus the IMA preferably makes the ATM traffic transparent to the
ATM layer over multiple links in use. As far as the ATM layer is
CA 022247~6 1997-12-12
concerned, it should only see a pipe (can be considered as a virtual link)
whose rate is now the sum of the multiple link rates. It is assumed that
each link is run in clear-mode without the presence of intermediate
ATM nodes processing ATM cells. This means that there should be no
cell discard by any intermediate transmission equipment.
U. S. Patent No. 5,608,733, Mar. 4, 1997, Vallee et al uses ATM
sequence number cells indicating a specific round robin order of a
plurality of transmission links over which ATM data cells are
transmitted. The ATM sequence number cells also indicate whether or
10 not a destination is ready to receive ATM data cells in that specific
round robin order.
The present invention extends further a variety of
functionalities which are useful in inverse multiplexing.
15 Objects of the Invention
It is therefore an object of the invention to provide a method of
and apparatus for sending ATM traffic over a connection consisting of a
plurality of transmission links.
It is a further object of the invention to provide a method of
20 sending a series of ATM cells between IMA devices over a connection
consisting of a plurality of transmission links, transparent to the ATM
layer.
It is also an object of the invention to perform IMA over a
connection consisting of a plurality of transmission links which are
25 selected according to one of a plurality of selection criteria.
It is another object of the invention to provide a method of
sending ATM traffic over a connection which consists of a plurality of
transmission links and has been reconfigured after the connection
start-up.
It is a further object of the invention to provide a method of
handling link failure and link reconfiguration.
It is yet another object of the invention to provide a method of
handling link configuration or reconfiguration according to one chosen
set of a plurality of possible link setting criteria.
CA 022247~6 1997-12-12
Summary of the Invention
Briefly stated, the invention resides in inverse multiplexing
digital data. According to one aspect, the invention is directed to a
method of selecting any of a plurality of transmission links for a
5 connection. The method comprises steps of obtaining transmission
parameters of the transmission links and determining one or more
subsets of all transmission links whose transmission parameters meets
one or more preset requirements. The method includes further steps
of choosing one of a plurality of selection criteria and selecting one
10 subset of transmission links among all the subsets determined above
for activation in response to the chosen criterion.
Brief Description of the Drawings
For a more complete understanding of the invention and for
15 further objects and advantages thereof, reference may now be made to
the following description, taken in conjunction with the accompanying
drawings, in which:
Figure 1 shows a sample configuration involving IMA UNIs;
Figure 2 shows a sample configuration involving IMA BICIs or
20 PNNIs;
Figure 3 is a schematic illustration of multiplexing and de-
multiplexing of ATM cells over IMAs and links;
Figure 4 is an illustration of selection criteria.
25 Detailed Description of Plefelled Embodiments
Figure 3 shows how the ATM cells are demultiplexed and then
multiplexed over IMAs in one direction. At the (near end)
transmitting node, an IMA 10 takes a series of ATM cells from an ATM
layer device. It spreads ATM cells and transmits each cell over each of
30 N transmission links, N being a positive integer. The N transmission
links form a link group and there may be more than one link group
between the nodes. The order of transmission is in the round robin
fashion within the link group. At the (far end) receiving node, cells
from N links are assembled and sent to an ATM layer device by an IMA
35 14. This process is called an inverse multiplexing The same order
must be employed at this node to recover a proper sequence of cells.
CA 022247~6 1997-12-12
Both nodes must be aware of the round robin order which is to be
employed. Upon initialization, therefore, both IMAs send a series of
special cells (e.g. sequence number cells) in round robin fashion over
the links, e.g. T1/E1 links. This allows the receiver IMA at both nodes
to establish the sequence in which to read cells from the incoming links
as well as to adjust relative link delay.
Links within an IMA link group may have different delays
which may also vary in time. Thus, cells transmitted in a given order
may not be received in the same order at the far end. In fact, a cell may
10 be received significantly later than another cell that followed it in the
original cell streams.
The receiving IMA device must preserve cell order; to do this it
buffers cells received on each link and compensates for the differential
delays prior to recombining the cells into the single stream. The term
15 link delay synchronization (LDS) is used to indicate that a receive link's
differential delay, relative to its peers in the IMA link group, has been
measured and compensated for.
In the process of IMA group startup, while establishing
communication with the far end node, the IMA measured the
20 differential delay of the receive links. If a link's differential delay is
within a specified maximum, the link is declared to be in LDS, and it
can be accepted as an active member of the link group. If the link's
differential delay exceeds the specified maximum, the link is not in
LDS and is rejected as an active member. When both near and far ends
25 agree on which links are acceptable, they can start exchanging user cells.
The IMA continually monitors and compensates for the
differential delay of its links. If a link's delay changes such that it
exceeds the specified maximum differential delay, the link is declared
to have experienced loss of delay synchronization (LODS), and is
30 deactivated from the group.
When a change of link configuration occurs by a link being
added, removed or declared as being down, each node sends a series of
ICP (IMA Control Protocol) cells to allow the far-end node to reestablish
the sequence of cells to read from the incoming links.
A set of IMA link configuration procedures is accepted in the
industry, such as the IMA group startup which includes "Addition of a
CA 022247~6 1997-12-12
link to an IMA group", "Deletion of a link from an IMA group, and
"Deactivation of a link from an IMA group due to a link or IMA fault".
Referring to Figure 3 again, the IMA device provides
provisioning and monitoring of the IMA groups. In one arrangement,
5 up to 8 IMA groups are supported and one IMA group may comprise 1
to 8 ports, each handling one transmission link.
According to the above accepted procedures, the IMA device
provides the user with the following options;
(1) Selection of the maximum differential link delay to be
10 tolerated within the group (a default value of 25 milliseconds being
widely accepted); and
(2) Selection of the period of time to wait before attempting to
reactivate a failed link.
Because the receiving IMA equalizes the delay of all links active
15 in the IMA group, the IMA virtual link effectively exhibits a
transmission delay equal to that of the greatest delay link active in the
group. Thus the following trade-off must be made when selecting the
maximum differential delay to be supported: a larger value may permit
more links to be activated within the group, but admission of these
20 links may result in a larger effective transmission delay for the IMA
virtual link. Because the maximum differential delay to be supported
is provisionable, the user can select the best value for a given
application.
Selection of other transmission characteristics parameters is also
25 possible for provisioning transmission links, such parameters as bit
error rate, transit delays etc.
As mentioned above, there are more than one transmission
links in an IMA link group. At a connection startup or
reconfiguration, certain links are selected for activation based on
30 transmission characteristics of each link etc., such as those mentioned
immediately above. It is however recognized now that there may be
other criteria which can be used for selecting links for certain
conditions. When certain provisioned transmission links do not meet
minimum requirments, other selection criteria can be used for link
35 activation.
CA 022247~6 1997-12-12
As one example, according to one embodiment of the invention,
such selection criteria as "minimize delay" and "maximize bandwidth"
can be used when not all provisioned links in the group conform to the
maximum differential delay.
When the criterion is minimum delay, links are selected for
activation as follows. First the least delay link is selected as the
reference link. Then, all links with differential delay relative to the
reference link less than or equal to the provisioned maximum
differential delay are selected for activation. Any links with relative
10 differential delay greater than the provisioned maximum are rejected
and are considered to be in a Loss of Delay Synchronization (LODS)
condition. Using this criterion will ensure that the IMA virtual link
has the lowest effective transmission delay possible given the
maximum acceptable differential delay, but it may restrict the
15 throughput capacity of the IMA virtual link.
When the criterion is maximum bandwidth, links are selected
for activation as follows. Each link in turn is considered to be the
reference link, and the corresponding set of acceptable links is
determined. The acceptable links are those with differential delay
20 relative to the reference link less than or equal to the provisioned
maximum differential delay. The largest set of acceptable links is then
selected for activation; if there are multiple sets of the largest size the
one with the lowest delay link is selected. Any unacceptable links are
rejected and are considered to be in a LODS condition. Using this
25 criterion will ensure that the IMA virtual link has the greatest
throughput capacity possible given the maximum acceptable
differential delay, but it may result in a larger effective transmission
delay of the IMA virtual link.
In one embodiment, available criteria are presented to the user
30 for selection but default is set to the maximum bandwidth. The
selection of criteria and/or execution of link selection according to the
chosen criterion can be made at any time, either at the connection
startup or at reconfiguration.
Figure 4 provides an example illustrating the two selection
35 criteria. Assuming that the IMA group has been provisioned to accept
a maximum differential delay of 25 milliseconds, then two different
CA 022247~6 1997-12-12
sets of links would be activated depending on the chosen link selection
criterion. If the least delay is used, links 1 and 2 would be activated, and
the effective transmission delay of the IMA virtual link would be 40
milliseconds. If the maximum bandwidth is used, links 2, 3, and 4
5 would be activated, and the effective transmission delay of the IMA
virtual link would be 60 milliseconds.
Note that there is no requirement for the near end and far end
IMA groups to have the same provisioning. However, if provisioning
does not match, it may result in one end rejecting links that the other
10 end finds acceptable; these links will not be activated since a link must
be accepted by both sides in order to be activated.
The above embodiments use "maximum bandwidth" and
"minimum delay" as two possible criteria. It should be noted that
other transmission characteristics can be used as the selection criteria.
According to yet another embodiment, two ends negotiate for
the criterion to be chosen. In one case, one end can be a master and the
other a slave, in that the master imposes the set of criteria to be used.
Alternatively, the master specifies what links it wants selected. In this
way, cases where both ends choose disjoint sets can be avoided.
