SYSTEM FOR CONTENT BASED MESSAGE PROCESSING
The present invention provides a packet having a message and a processing priority associated with the message. The processing priority is dynamically changeable by a function operating on the message. The present invention also provides a method for associating a processing priority to a message by receiving the message, determining the processing priority for the message and associating the processing priority with the message such that the processing priority is dynamically changeable by a function operating on the message. In addition, the present invention provides a method for scheduling messages by receiving one or more messages and storing each message in a multidimensional processing queue based on a processing priority and an attribute associated with the message. Each queued message from the multidimensional processing queue is scheduled for processing based on an algorithm.
The present invention relates generally to the field of communications and, more particularly, to a method and apparatus for content based message processing.
BACKGROUND OF THE INVENTIONThe increasing demand for data communications has fostered the development of techniques that provide more cost-effective and efficient means of using communication networks to handle more information and new types of information. One such technique is to segment the information, which may be a voice or data communication, into packets. A packet is typically a group of binary digits, including at least data and control information. Integrated packet networks (typically fast packet networks) are generally used to carry at least two (2) classes of traffic, which may include, for example, continuous bit-rate (“CBR”), speech (“Packet Voice”), data (“Framed Data”), image, and so forth. Packet networks source, sink and/or forward protocol packets.
Congestion and Quality of Service (“QoS”) problems inside these networks have not been solved satisfactorily and remain as outstanding issues to be resolved. Although, message scheduling helps alleviate these problems, the efficient scheduling of work with thousands of entities (instances) is not a simple matter. At present, most message scheduling is based on the simplest technique for queuing packets for transmission on an internodal trunk of a fast-packet network: a first-in-first-out (“FIFO”) queue. However, FIFO queuing techniques do not address QoS parameters. This technique can also allow overload periods for digitized speech packets and for Framed Data packets, which results in a greater share of bandwidth being provided to one at the expense of the other; an undesirable result.
Another technique, head-of-line-priority (“HOLP”), may give data priority over speech, but does not solve the problem of data and speech queues affecting the QoS of each other and of CBR data fast packets under high traffic conditions. In HOLP, where speech fast packets are given a high priority, speech fast packets may affect the QoS of lower priority queues. Likewise, queuing schemes designed only for data do not solve the problems of integrating other traffic types, such as speech and CBR data.
Traditional packet data routers are constructed in software using a scheduler or Real Time Operating System (“RTOS”), which associates the processing priority of functions (protocols or other operations performed upon a packet including forwarding) with the task or process that the function operates under. Thus, a particular function has a priority and all packets processed by the function inherit that priority for the duration of their processing by the function. If the next processing step to which the packet is subjected is set at a different priority, then the packet inherits a different priority for the duration of this processing step. Priority is associated with the function applied to the packet rather than the packet itself. If all packets traverse the same set of functions, they have equal access to the central processing unit (“CPU”) and receive equal priority treatment.
If all packets had equal priority, this might be sufficient. However, due to the need to sell different QoS, and the needs resulting from multimedia (voice, video and data) carried upon the same network infrastructure, there is a need to assign different priorities to a packet, and vary resource allocation (especially processing time) and forwarding treatment on a per-packet basis.
SUMMARY OF THE INVENTIONThe present invention provides a system for assigning different priorities to a packet, and varying resource allocation (especially processing time) and forwarding treatment on a per-packet basis. The present invention is adaptable to accommodate new message types, multimedia applications and multi-service applications. It is flexible, with the ability to cater to a wide range of configurations and environments and improves the QoS of VoIP calls.
The present invention provides a packet having a message and a processing priority associated with the message. The processing priority is dynamically changeable by a function operating on the message. The processing priority can be associated with the message by attaching the processing priority to the start of the message, appending the processing priority to the end of the message or linking the processing priority to the message using pointers. The system for associating a processing priority to a message involves receiving the message, determining the processing priority for the message and associating the processing priority with the message such that the processing priority is dynamically changeable by a function operating on the message. This process can be implemented using a computer program embodied on a computer readable medium wherein each step is performed using one or more code segments.
The present invention also provides a method for scheduling one or more messages. The one or more messages are received and then each message is stored in a multidimensional processing queue based on a processing priority and an attribute associated with the message. Each queued message from the multidimensional processing queue is scheduled for processing based on an algorithm. The attribute can be a virtual private network classification, destination software function, function index, a functionality type or other message attribute. The algorithm can be one or more algorithms, such as an exponentially weighted, non-starving, nested-round-robin, message-priority-based scheme or a weighted, non-starving, nested-round-robin, class-based scheme. This process can be implemented using a computer program embodied on a computer readable medium wherein each step is performed using one or more code segments.
In addition, the present invention provides a communications switch having one or more ingress cards, one or more signal processing cards, one or more control cards containing one or more processors and one or more egress cards. Each signal-processing card contains an array of digital signal processors. The switch also includes a switch fabric communicably coupling the ingress cards, the signal processing cards, the control cards and the egress cards, a TDM bus communicably coupling the ingress cards, the signal processing cards, the control cards and the egress cards, a multidimensional processing queue, and a scheduler communicably coupled to each processor and the multidimensional processing queue. The scheduler receives one or more messages, stores each message in the multidimensional processing queue based on a priority and an attribute of the message, and schedules each queued message from the multidimensional processing queue for processing based on an algorithm.
The above and further advantages of the invention may be better understood by referring to the following description in conjunction with the accompanying-drawings, in which:
While the making and using of various embodiments of the present invention are discussed in detail below, it should be appreciated that the present invention provides many applicable inventive concepts that can be embodied in a wide variety of specific contexts. The specific embodiments discussed herein are merely illustrative of specific ways to make and use the invention and do not delimit the scope of the invention. The discussion herein relates to communications systems, and more particularly, to processing messages within a communications switch. It will be understood that, although the description herein refers to a communications environment, the concepts of the present invention are applicable to other environments, such as general data processing.
The present invention provides a system for assigning different priorities to a packet, and vary resource allocation (especially processing time) and forwarding treatment on a per-packet basis. The present invention is adaptable to accommodate new message types, multimedia applications and multi-service applications. It is flexible, with the ability to cater to a wide range of configurations and environments and improves the QoS of VoIP calls.
Now briefly referring to
IP specifies the format of packets, also called datagrams, and the addressing scheme. Most networks combine IP with a higher-level protocol called Transport Control Protocol (“TCP”), which establishes a virtual connection between a destination and a source. IP allows a packet to be addressed and dropped in a system, but there is no direct link between the sender and the recipient. TCP/IP, on the other hand, establishes a connection between two hosts so that they can send messages back and forth for a period of time. IP network 110 receives and sends messages through switch 108, ultimately to phone 102 and/or fax 104. PCs 116 receive and send messages through IP network 110 in a packet-compatible format. Voice over IP (“VoIP”) is the ability to make telephone calls and send faxes over IP-based data networks, such as IP network 110. An integrated voice/data network 100 allows more standardization and reduces total equipment needs. VoIP can support multimedia and multi-service applications.
Now referring to the present invention and to
The priority/priority criteria 402 are associated with the packet/message 404 in such a way that reference to one allows reference to the other, they traverse the system together, and that functions operating upon the packet/message 404 have the ability to change the priority/priority criteria 402. Changing the priority/priority criteria 402 dynamically, when used in conjunction with the other aspects of this invention, creates a processing environment where the priority/priority criteria 402 of the message/packet 404 governs the work allocation or dispatching.
Since system hardware and software for carrying messages around a system vary, the present invention provides at least three ways to associate the priority or parameters that may be evaluated to produce a priority 402 to a message 404. The priority/priority criteria 402 may be attached to the start of the message 404 (
Referring now to
Now referring to
The main function of a packet network switch 600 is to relay user data cells from input ports to the appropriate output ports. When a call or communication is to be handled by the packet network switch 600, a network controller 118 (
Turning now to
For example, one possible configuration of the multidimensional queue 804 could be based on VPN classification in the X-direction, processing priority in the Y-direction, and first-in-first-out (“FIFO”) in the Z-direction. Moreover, each function can have a slot comprised of multiple dimensions. A fourth dimension can also be added to the multidimensional queue 804 by making it an array of three-dimensional queues, where each one is handled by one type of functionality. In order for the scheduler or dispatcher 802 of the multidimensional queue 804 to call the right functionality, a function index and a jump table can be used.
The multidimensional queue 804 can be characterized as an advanced queue structure that consists of multiple sub-queues bundled in a single receive queue wherein each sub-queue serves a set of messages 806. The messages 806 can be classified by their priority (first dimension) and message classification or service classes (second dimension). Priority sub-queues will be serviced according to one or more algorithms, such as an exponentially weighted round robin scheme. Within each priority there will be multiple sub-queues representing multiple VPN service classes. VPNs will be mapped into these service classes. Service classes themselves will have a weighting scheme among themselves so that different qualities of service can be provided. In this example, the multidimensional queue 804 is a two-dimensional queue, consisting of p*c monolithic sub-queues, where p is the number of message priorities and c is the number of VPN service classes. The multidimensional queue 804 itself is three-dimensional since the messages within each of the p*c sub-queues represent the third dimension (the depth of the sub-queues). The messages within each one-dimensional sub-queue are serviced in FIFO order.
The receiver function of scheduler 802 stores messages 806 in the multidimensional queue 804 (indicated by arrow 808) based on a processing priority or priority criteria and an attribute associated with the message 806. Note that multiple attributes may be used to determine where the message 806 is stored in the multidimensional queue 804. A special function can be used to insert the message 806 into the multidimensional queue 804. For example, this special function can use the function index, the VPN, the priority, and/or any other important criteria to insert the message 806 into the multidimensional queue 804. The dispatcher function of scheduler 802 pulls or schedules queued messages from the multidimensional queue 804 (indicated by arrow 810) for processing by the one or more processors 812 based on an algorithm. The algorithm may take into account operating criteria, such as historical operating data, current operating data, anti-starvation criteria, one or more of the message attributes as described above, or combinations thereof. For example, the algorithm may be an exponentially weighted, non-starving, nested-round-robin, message-priority-based scheme, or a weighted, non-starving, nested-round-robin, class-based scheme, or any combination thereof. Other suitable algorithms, depending upon the specific application, may be used in accordance with the present invention. The algorithm may also provide no more than a pre-determined number of consecutive messages to a function or processing entity within a time period. Once the scheduler 802 pulls or schedules a queued message, the scheduler 802 sends the message to the processor 812.
Now referring to
In addition, the present invention provides a communications switch having one or more ingress cards, one or more signal processing cards, one or more control cards containing one or more processors and one or more egress cards. Each signal processing card contains an array of digital signal processors. The switch also includes a switch fabric communicably coupling the ingress cards, the signal processing cards, the control cards and the egress cards, a TDM bus communicably coupling the ingress cards, the signal processing cards, the control cards and the egress cards, a multidimensional processing queue, and a scheduler communicably coupled to each processor and the multidimensional processing queue. The scheduler receives one or more messages, stores each message in the multidimensional processing queue based on a priority and an attribute of the message, and schedules each queued message from the multidimensional processing queue for processing based on an algorithm.
The algorithm used by the present invention can be a single algorithm or multiple algorithms that are selectively used depending on various operating criteria. For example, an exponentially weighted, non-starving, nested-round-robin message-priority-based scheme could be used. Weighted means higher priority messages are served more frequently than the lower priority messages. There is an exponential service ratio between successive priority levels. Non-starving means lower priority messages will eventually get served. Round-robin means servicing mechanism moves from one priority-level to the other in a round-robin fashion. The nesting gives the exponential service weighting, e.g. assume that there are three message priorities: High, Medium and Low. Also assume that the queues have messages in them at any given time. Then, the order and amount of servicing would be H-M-H-L-H-M-H and repeating in the same order. So, four High messages, two Medium messages and one Low message would have been serviced. Also, after one high priority message is serviced, it will take at most one more lower priority message service before another high priority message is serviced. The following illustration may serve better to explain the order of servicing.
Another algorithm that can be used is a weighted, non-starving, round-robin, VPN class-based scheme. Within each priority, there are multiple classes of service. For each class there is a maximum number of messages that can be serviced before the next class is serviced. The maximum number of serviceable messages assigned to each class defines the relative priority among those classes.
The algorithm may also provide a maximum number of messages that can be serviced during each scheduling period. Regardless of priority or service class a function is not given more than a pre-determined number of consecutive messages to be serviced. When it reaches the maximum, the dispatcher starts dequeuing messages for another function.
A service example with four priority levels and three classes will now be described. Service ratios for priorities: 8-4-2-1, i.e. for every eight priority-one messages serviced the task will serve one priority-four message. However, it will do this in a round-robin fashion so that priorities are interleaved. For example, assume that there are enough messages at each priority level and these number represent the priority level of each successive message being dequeued: 1-2-1-3-1-2-1-4-1-2-1-3-1-2-1, and so on repeating the same sequence.
Service ratios for classes: 10-6-3, i.e. for every 10 class-one messages serviced, the task will service 6 class-two messages and 3 class-three messages. These ratios are kept on a per priority basis so as to avoid starvation and imbalance among different classes. Within each class the higher priority messages will be serviced more frequently than the lower priority messages based on the service ratios for priorities. The maximum messages to serve consecutively for this function are 15.
Assume the following queue status to start with. The rows represent different priorities and the columns represent different classes. Each cell represents the depth of the sub-queue. For example, there are three messages in the queue represented by priority four and class three. For simplicity of this illustration, assume that no new message is inserted into these queues during servicing.
If this queue was serviced for the first time, the message at the head of priority-one, class-one sub-queue would be de-queued. So, after the first iteration, the queue depths would look like this (the change is shown in bold):
The following table illustrates the dequeuing from the subqueues at each iteration. The first column indicates the cell being serviced. The second column indicates the depth of the sub-queue after the service. The third column indicates the next message priority that needs to be serviced within this class. The fourth column indicates the next calls that needs to be serviced. The fifth column indicates remaining service quota for current class. Note that when remaining class quota reaches 0 or there are no more messages left in the current class, we move on to the next class. The seventh column indicates total number of messages that were served during this scheduling period. The first iteration would be:
At this point the maximum number of messages that can be serviced consecutively have been exhausted. Here is how the queue depths look like after the first consecutive run:
The second iteration would be:
At this point there is no message left in any of the subqueues, so the dispatcher would move onto the next function's queue. The exponential weighted priority servicing mechanism is not reset back to priority-one because the next message to service may be the lowest priority message in that class. This ensures that there is no starvation of low priority messages if the process happens to move out of a class whenever the class quota is exhausted.
Although preferred embodiments of the present invention have been described in detail, it will be understood by those skilled in the art that various modifications can be made therein without departing from the spirit and scope of the invention as set forth in the appended claims.
Claims
1. A packet comprising:
- a message;
- a processing priority associated with the message, the processing priority being dynamically changeable on a per packet basis by a function operating on the message, wherein the processing priority comprises one or more parameters per packet that are evaluated to produce a priority for the packet.
2. The packet as recited in claim 1, wherein the processing priority comprises one or more parameters that are evaluated to produce a priority for the message.
3. The packet as recited in claim 1, wherein the processing priority is associated with the message by attaching the processing priority to the start of the message.
4. The packet as recited in claim 1, wherein the processing priority is associated with the message by appending the processing priority to the end of the message.
5. The packet as recited in claim 1, wherein the processing priority is associated with the message by linking the processing priority to the message using pointers.
6. A method for associating a processing priority to a message comprising the steps of:
- receiving the message;
- determining the processing priority for the message on a per packet basis;
- associating the processing priority with the message such that the processing priority is dynamically changeable by a function operating on the message, wherein the processing priority comprises one or more parameters that are evaluated per packet to produce a priority for the packet.
7. The method as recited in claim 6, wherein the processing priority comprises one or more parameters that are evaluated to produce a priority for the message.
8. The method as recited in claim 6, wherein the processing priority is associated with the message by attaching the processing priority to the start of the message.
9. The method as recited in claim 6, wherein the processing priority is associated with the message by appending the processing priority to the end of the message.
10. The method as recited in claim 6, wherein the processing priority is associated with the message by linking the processing priority to the message using pointers.
11. A computer program loadable into a memory to be read and executed by a processor, comprising:
- code segments adapted to associate a processing priority to a message, said code segments further comprising:
- a code segment for receiving the message;
- a code segment for determining the processing priority for the message;
- a code segment for associating the processing priority with the message such that the processing priority is dynamically changeable by a function operating on the message, wherein the processing priority comprises one or more parameters that are evaluated per packet to produce a priority for the packet.
12. The computer program as recited in claim 11, wherein the processing priority comprises one or more parameters that are evaluated to produce a priority for the message.
13. The computer program as recited in claim 11, wherein the processing priority is associated with the message by attaching the processing priority to the start of the message.
14. The computer program as recited in claim 11, wherein the processing priority is associated with the message by appending the processing priority to the end of the message.
15. The computer program as recited in claim 11, wherein the processing priority is associated with the message by linking the processing priority to the message using pointers.
16. A method for scheduling one or more messages comprising the steps of:
- receiving the one or more messages on a per packet basis;
- storing each message in a multidimensional processing queue based on a processing priority and an attribute associated with the message, wherein the processing priority comprises one or more parameters that are evaluated per packet to produce a priority for the packet; and
- scheduling each queued message from the multidimensional processing queue for processing based on an algorithm,
17. The method as recited in claim 16, wherein the attribute is a virtual private network classification.
18. The method as recited in claim 16, wherein the attribute is a destination software function.
19. The method as recited in claim 16, wherein the attribute is a function index.
20. The method as recited in claim 16, wherein the attribute is a functionality type.
21. The method as recited in claim 16, wherein the algorithm provides no more than a pre-determined number of consecutive messages to a function within a time period.
22. The method as recited in claim 16, wherein the processing priority comprises one or more parameters that are evaluated to produce a priority for the message.
23. The method as recited in claim 16, wherein the processing priority is dynamically changeable by a function operating on the message.
24. A computer program embodied on a computer readable medium loadable into a computer memory to be read and executed by a processor, comprising
- code segments adapted to schedule one or more messages on a per packet basis, further comprising:
- a code segment for receiving the one or more messages;
- a code segment for storing each message in a multidimensional processing queue based on a processing priority and an attribute associated with the message wherein the processing priority comprises one or more parameters that are evaluated per packet to produce a priority for the packet, and further wherein the processing priority; and
- a code segment for scheduling each queued message from the multidimensional processing queue for processing based on an algorithm.
25. The computer program as recited in claim 24, wherein the attribute is a virtual private network classification.
26. The computer program as recited in claim 24, wherein the attribute is a destination software function.
27. The computer program as recited in claim 24, wherein the attribute is a function index.
28. The computer program as recited in claim 24, wherein the attribute is a functionality type.
29. The computer program as recited in claim 24, wherein the algorithm provides no more than a pre-determined number of consecutive messages to a function within a time period.
30. The computer program as recited in claim 24, wherein the processing priority comprises one or more parameters that are evaluated to produce a priority for the message.
31. The computer program as recited in claim 24, wherein the processing priority is dynamically changeable by a function operating on the message.
Type: Application
Filed: Jul 28, 2008
Publication Date: Nov 27, 2008
Inventor: Fredrik Hammar (Molnlycke)
Application Number: 12/180,630
International Classification: H04L 12/56 (20060101);