Multiplex communications – Diagnostic testing – Determination of communication parameters
Reexamination Certificate
1998-02-17
2001-02-13
Rao, Seema S (Department: 2661)
Multiplex communications
Diagnostic testing
Determination of communication parameters
C370S413000, C370S473000, C370S522000
Reexamination Certificate
active
06188674
ABSTRACT:
BACKGROUND
In various communications networks user information is carried in packets that are routed from the source to the destination via switches. A switch buffers each packet received from upstream sources until a decision is made when and to which outgoing link the packet need to be forwarded. Since the buffer space is always finite, packets could be dropped within a switch when buffer overflow occurs or when packets with a lower priority class have to give way to those with a higher priority class. In applications where a particular quality of service is desired or promised, it is obviously important for a switch to monitor packet losses incurred within the switch. Moreover, routing protocols require this information to be advertised within a network or a group of nodes, to facilitate path selections and to avoid congestion.
A high speed switch can generally be represented by a multi-stage interconnection where an N plurality of ingress modules are connected to a central switch fabric, and the outputs of the switch fabric are connected to an N plurality of egress modules. Each ingress and each egress module couples K input links and output links, respectively. Links carry streams of packets, where the destination of each packet is specified in the packet's header.
High speed switches of such construction typically include buffering within each of the elements (ingress modules, switching fabric, and egress modules). A packet received from an incoming link is first buffered at the ingress module, where a table lookup typically takes place to find out the egress module which the packet is destined to. In such an implementation, a local routing header will normally be placed in the front of the received packet. The local routing header may contain a number of information fields that are necessary to route the packet through the switching fabric, including the identity of the output link to which the packet needs to be routed. The extended packet is then dispatched to the desired egress module through the high speed, where some local buffering may be necessary. At the egress module, the packet is buffered until the outgoing link becomes available to transmit the packet. The local routing header is removed in the egress module before the packet is transmitted.
In a hierarchical communication network, a set of interconnected switches, each having the construction described above, can be unified in a logical group of nodes. From the network point of view, this logical group of nodes itself represents a generalized switch with input modules, output modules and generalized switching fabric. The network itself is a generalized switch on the highest level of hierarchy. Each logical group of nodes can maintain an internal routing header.
In this configuration, the packet could be lost in any of multiple locations within the ingress module, the switching fabric, or the egress module. The loss can be caused by buffer congestion, priority control, usage parameter control or a random electrical glitch leading to mis-routing, delineation error, checksum error, etc.
It is of practical interest, as indicated above, to be able to know whether, and how many, packets are lost within a switch or a logical group of nodes. It is of even greater interest to associate a packet loss with a particular traffic flow within the generalized switch.
SUMMARY OF THE INVENTION
This disclosure presents a simple and effective approach to the local packet loss measurement that is oriented to packet flows rather than the loss sources and allows to achieve high precision in a distributed switching environment with multiple loss reasons and multiple loss locations. The loss measurements are achieved by selecting the particular traffic flow, or flows, to be measured, identifying those flows in the ingress modules, marking each of the flows, and observing the packets of the flows at the egress modules. More specifically, a packet flow to be measured is broken into blocks of packets, and the last packet in the block is marked in an ingress module. At an egress module, the packets are counted using the block size as the counting modulus, and receipt of a marking packet triggers an update of the number of lost packets.
REFERENCES:
patent: 5491697 (1996-02-01), Tremel et al.
patent: 5668801 (1997-09-01), Grunenfelder
patent: 5793976 (1998-08-01), Chen et al.
patent: 5822540 (1998-10-01), Caldara et al.
patent: 5978356 (1999-11-01), Elwalid et al.
Chen Xiaoqiang
Khotimsky Denis Andreyevich
Brendzel Henry T.
Rao Seema S
LandOfFree
Method and apparatus for packet loss measurement in packet... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for packet loss measurement in packet..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for packet loss measurement in packet... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2558787