WO2004057814A1

WO2004057814A1 - Method, resource manager and computer program for using reserved resources for data traffic in an ip-network

Info

Publication number: WO2004057814A1
Application number: PCT/SE2003/001839
Authority: WO
Inventors: Joachim Johansson; Joakim NORRGÅRD; Johan Larsson
Original assignee: Operax Ab
Priority date: 2002-12-20
Filing date: 2003-11-28
Publication date: 2004-07-08
Also published as: AU2003282653A1

Abstract

The present invention relates to a method for using reserved resources for data traffic in an IP-network, wherein a receiving node comprises a classifier comprising filters with overlapping data flow separating parameters. The method comprises the steps of matching traffic with a first filter, metering said received traffic, based on a predetermined resource type, in order to establish if said received traffic fits into the reservation defined by the first filter together with the amount of resource, and repeating the preceding steps for a next filter, when the received traffic matches with said first filter and said traffic does not fit into the reservation. In addition, the present invention relates to computer program product and a resource manager for performing the steps of the above mentioned metod.

Description

Title

IP network

Field of the invention

The present invention relates to a method, a resource manager and a computer program product in an IP-network.

In particular, the present invention relates to usage of resource reservations in an IP-network when at least two reservations are overlapping.

Background of the invention A current networking trend is to provide "IP all the way" to wired and wireless units. Some current objectives are to simplify the network infrastructure, to support a wide range of applications, and to support diverse user demands on the communication service. To allow this, there is a need for scalable solutions supporting service differentiation and dynamic resource management in IP networks.

The primary goal when the Internet Protocols were designed was to provide an effective technique for interconnecting existing networks. One design trade-off made to enable the interconnection was to support only best-effort service at the network level and rely on endpoint functionality to obtain various levels of service. Best-effort service provides adequate support for traditional data applications that can tolerate delay, loss and varying throughput along the path. However, in networks carrying high loads of traffic, this type of service is often inadequate for meeting the demands of applications that are more sensitive to packet loss and delay e.g. telephony, video on demand, multimedia conferencing, etc. These types of applications require a more reliable resource allocation than what best-effort can offer.

Consequently, there are strong commercial reasons for network operators and equipment providers to offer Quality-of-Service (QoS) differentiation in IP networks. I.e., the users within a network are divided into different group depending on their priority, e.g. high prioritized users are offered more available resources (e.g. rate, delay, jitter, packet-size, toke-bucket parameters etc.) than users with lower priorities.

The scalability problems of per-flow QoS management in routers have resulted in the differentiated services architecture defined in Blake et Al, An architecture for Differentiated Services, IETF, RFC2475. The objective is to provide scalable QoS support by avoiding per-flow state in routers. The basic idea is that IP packet headers include a small label (known as the diffserv field) that identifies the treatment per-hop behaviour that packets should be given by the routers. The standard model is, however, limited to differentiated forwarding in routers and therefore the challenge lies in providing predictable services to end users.

The entity performing dynamic admission control is here called a resource manager and is further described in Wolf, L.C., Delgrossi, L., Steinmetz, R., Schaller, S., Wittig, H., "Issues of Reserving Resources in Advance", IBM European Network Center Heidelberg, TR 43.9503, 1995. The resource manager keeps track of the available transmission resources, e.g. bandwidth, delay, jitter, packet-size, token-bucket parameters etc, and performs admission control on incoming requests for resources from clients. The resource manager manages the resources within one domain. To perform the admission control the resource manager also stores a history of previously admitted resource reservations. The resource manager takes decisions to admit new requests for resources based on the total amount of available resources, the amount currently reserved by previously reservations and the amount of resources requested. The resources may or may not be scheduled over time. One request may involve admission control on multiple resource repositories that may consist of different types of resources. The most common type of resource managed is bandwidth.

There are specific requirements for resource management mechanisms. To provide service to end users, they must be aware of network resources and may schedule them for the committed service at any granularity (e.g. for a port range, for aggregate traffic between a pair of subnets, etc).

Filter A classifier classifies the traffic by means of a data flow separating filter and the classifier is preferably a part of a traffic conditioner. Furthermore, the traffic conditioner is generally implemented in a router, e.g. a DiffServ router. The classifier classifies the traffic into different classes depending on header values, contents, and attributes and port numbers that are relevant for the classification. The data flow separating filter separates the traffic into different flows by means of the data flow separating parameters, i.e. said header values, contents, attributes and port numbers. In general, there are a set of possible conditions including exact, prefix, range, masked and wildcard matches. A 5-tuple classifier is a common filter of the type multi- field classifier. This filter classifies based on five fields from the IP and TCP or UDP headers (destination address, source address, IP protocol, source port and destination port) . Traffic having parameters matching all these five fields of the filter is said to match the filter and said traffic belongs to the same flow. More general filters may include other fields and data from the , packets, wildcards on some fields and ranges or address-prefixes. Notice that more general filters introduce problems when the same traffic may match more than one filter. Thus, when the same traffic matches more than one filter, the matched filters are overlapping filters. The traffic may be a specific flow of data packets between two applications or all data packets between two sub networks. The data packets comprise a payload part and a header part. The payload part comprises the useful data information while the header part comprises control information such as the data flow separating parameters described above.

Each filter defines together with the resource type, e.g. a rate, a resource reservation.

Metering

A meter performs metering based on the above-mentioned classification of the filter in order to investigate if the traffic fits into the reservation. The meter is measuring a resource type, i.e. a temporal property of a traffic stream selected by a classifier, e.g. the rate of arrival. If the measured property, e.g. the rate does not exceed the reserved resource, e.g. the upper limitation of the rate, the traffic fits into said reservation and said traffic is generally marked with a Differentiated Services Code Point (DSCP). Excess traffic, i.e. traffic that does not fit into the reservation, in this case exceeding the upper limitation, is generally either dropped or marked with the default "best effort" treatment. The meter is preferably a part of the traffic conditioner. The traffic conditioner is an entity that is adapted to handle metering and classifying and the traffic conditioner is generally implemented in a router.

DiffServ router

The DiffServ model described above relies on that the traffic is classified and marked either by a trusted upstream node, e.g., a customer, or by the ingress routers to the Diffserv network. The conceptual model of a DiffServ router includes, among others, traffic classification, metering functions, actions of marking, dropping, continuing and multiplexing.

It is not a requirement to have DiffServ capable routers in a network that performs the classifying and metering function, other methods may also be used.

Overlapping filters When the ranges of all parameters of two or more filters are overlapping, the filters are overlapping. That implies that the resource reservations are overlapping if the resources associated to the respective filter are of the same type.

Overlapping filters are currently avoided as much as possible when configuring filters in routers. Filters are generally very specific which implies that each filter identifies a single flow but it also exist a few number of filters that are more general and categorise aggregates of data flows into a few number of predefined classes. Overlapping filters are considered to be a difficult problem, since a general filter may match the same traffic as a more specific filter. That may result in that one or more of the filters is never being used if support for the overlapping filters is missing. I.e. the "wrong" filter may be used, since it is not predetermined which of the filters should be matched first.

Therefore, much effort has been done to develop efficient ways to detect overlapping, "conflicting", filters, e.g. as described in Anja Feldmann and S. Muthukrishnan, Tradeoffs for Packet Classification, AT&T Labs-Research, In Proc.Infocom ,vol. l, pages 397 - 413, March 2000.

It should be noted that the terms metering, classifier, traffic conditioner, conforming, filter, reservation and overlapping filters are used in accordance with RFC 3290.

However, there exist situations where overlapping filters are required. In some cases, different applications and different users want to make reservations with different levels of aggregation. In one situation, one application requires resources for a single flow (a specific filter) while in other situations, a user wants to make a reservation for all traffic (a general filter) between two hosts or an administrator wants to make a reservation for all traffic between two networks (e.g. LANs or distributed company sites). Multiple resource reservations may also be aggregated into a more general aggregated reservation with a more general filter matching the aggregated traffic. In these cases, it is hence required to have a mixture of general and specific overlapping filters.

By supporting such more general aggregated reservations, application and user heterogeneity are supported. However, this leads to problems when one filter overlaps with another filter. The similar problem occurs if two different reservations using two separate filters with identical parameters. In this situation, a flow or parts of a flow may match more than one filter and the current common practise is to select one single filter to match to. E.g. in a DiffServ access router, all traffic that matches this selected filter is then measured to check if the traffic fits into the reservation, i.e. if the rate of the traffic is below the rate of the reservation. Traffic within the reservation is generally marked to receive the desired network service and excess traffic is often either marked or dropped.

Another way of handling overlapping filters is described in Hari Adiseshu, Subhash Suri and Guru Parulkar, Packet Filter Management for Layer 4 Switching, http:/ /www.ccrc.wustl.edu/~hari/packet-filter.ps, 1999, where the solution is to split conflicting filters into smaller non-overlapping parts. This will introduce extra filters in the routers and increases the complexity if resource reservations have to be inserted and removed dynamically. For each insertion and removal, many other filters may be affected and must be considered, either split, modified or removed. This may lead to extensive communication with the routers (a bottleneck), which also decreases the performance.

In Hari A. et al, Detecting and resolving packet filter conflicts, Infocom 2000, a method for detecting and handling overlapping filters is disclosed. Overlapping parameter areas are identified and replaced by resolve filters that have higher priority than the overlapping filters. The resolve filter covers the same area as the overlapping area. However, this method introduces also additional filters.

US 6341130 discloses a method for detecting overlapping filters and decomposing them into non-overlapping intervals. However, this solution introduces additional filters.

The RFC 3290 (Y. Bernet et Al, An Informal Management Model for Diffserv Routers, IETF, suggests that a precedence between overlapping filters must be established in order to uniquely classify a data flow by a single filter. This solution requires sorting of the filters and may lead to reinsertion of a part of the filters or of all filters in the router and thus decreasing the performance. Another problem with overlapping filters is to select which filter to match the traffic to if there are a plurality of overlapping filters. One way is to select the first inserted filter and another way is to select the most specific filter. Selecting the most specific filter may lead to problems if for example a general filter of 10 Mbit/s is inserted between two sites and if someone later inserts a specific filter of 10 kbit/s between said two sites, then the reserved traffic between the sites is limited to the later, but more specific reservation of 10 kbit/s. To sort the reservations in order to reach the desired match is even more difficult if the clients are allowed to specify ranges, e.g. source or destination port ranges. Furthermore, the filters may have the same range but overlapping port ranges and other fields in the filter may also have overlapping source or destination address prefixes.

Thus, an object of the present invention is to provide a method, a resource manager and a computer program product for using resource reservations when there are overlapping reservations in an IP-network.

Summary of the invention

The above-mentioned object is achieved by a method, a computer program, and by a resource manager according to the characterising part of the independent claims.

A method provided by the present invention, comprising the steps of a. matching data flow separating parameters of the received traffic with data flow separating parameters of a first filter of the at least two filters; b. metering said received traffic, based on said predetermined resource type, in order to establish if said received traffic fits into the reservation defined by the first filter together with the amount of resource, and repeating steps a and b for a next filter of the at least two filters, when the data flow separating parameters of the received traffic matches with the data flow separating parameters of said first filter and said traffic does not fit into the reservation, makes it possible to use resource reservations when there are overlapping reservations in an IP-network. A computer program product provided by the present invention, directly loadable into a memory of a node within an IP network comprising the software code portions for performing the steps of said method, makes it possible to use resource reservations when there are overlapping reservations in an IP-network.

A computer program product provided by the present invention, stored on a computer usable medium, comprising readable program for causing a computer within a node within an IP network to control the execution of the said method, makes it possible to use resource reservations when there are overlapping reservations in an IP-network.

A Resource manager provided by the present invention, adapted to configure a meter and a classifier for performing said method, makes it possible to use resource reservations when there are overlapping reservations in an IP- network.

Preferred embodiments are set forth in the dependent claims.

Advantages with the present invention are that it provides support for having filters with different levels of aggregation while minimising the problem of conflicting overlapping filters. Protection from "malicious" or misbehaving clients is provided, thanks to that a client is only allowed to add resources to the already reserved ones instead of replacing one reservation with another. In addition, the invention provides better support for application and user heterogeneity by allowing mixing of general and specific filters. Performance of managing filters in routers is also improved since reordering or splitting of existing filters is not required. Acquiring more resources for the same traffic is also performed -without requiring removing an existing filter. Each filter correlates to a single reservation (no splitting of filters) , which leads to less implementation complexity and thus higher robustness.

Another advantage is that the order of the filters is less important with the solution according to the present invention since traffic is matched to more than one overlapping filter, than in solutions according to prior art where only one match is considered. However, it may be advantageous to first match the traffic to the most specific filter and then continue with more and more general filters in order to obtain a high utilization.

Brief description of the appended drawings Figure 1 shows an IP-network schematically where the present invention may be implemented.

Figure 2 is a graph illustrating an example of how overlapping filters are handled according to the present invention.

Figure 3 is a flowchart of the method according to the present invention.

Detailed description of preferred embodiments

Figure 1 illustrates an IP network 100 where the present invention may be implemented. The network 100 comprises a plurality of routers 102a, 102b, endpoints (e.g. pc, IP telephones etc.) 104a-f and servers 103 connected to each other (not shown in figure 1). The routers 102 forward data traffic from a source 104a to a final destination 104f and the servers 103 comprise processing means for different types of operations. One or more resource managers 105 control the resources (e.g. rate, delay, jitter, packet-size- token-bucket parameters etc.) within the network 100 and are adapted to perform resource reservations. The resource manager is preferably located within a server 103 handling the resources of a plurality of routers. The resource manager 105 may also be distributed in a plurality of routers 102. The network comprises also traffic conditioners 106 and the traffic conditioners are in this example implemented in the routers 102. The traffic conditioners 106 comprise meters 107 and classifiers 108.

The IP network 100 in accordance with the present invention comprises one classifier 108 using at least two filters located in the same node, e.g. a router 102, within the network 100. The classifier 108 defines a resource reservation by means of data flow separating parameters e.g. header values, contents, and attributes of data packets, port numbers. Said data flow separating parameters are properties of the data packets, as described above. The classifier 108 uses at least two filters with overlapping data flow separating parameters according to the present invention. Said classifier 108 is preferably a part of a traffic conditioner 106, wherein the traffic conditioner may be implemented in a router 102. Moreover, the routers comprise preferably some DiffServ capabilities.

The classifier 108 defines for what type of traffic the resource reservation is performed and a rate or a set of other parameters intended for metering defines the type of resources and how much of this resource type that is reserved.

Traffic having data flow separating parameters matching data flow separating parameters of a filter in a classifier 108, i.e. the traffic matching the filter and at the same time, not exceeding the reserved amount of resources (e.g. rate) is said to be conforming and traffic exceeding the reserved amount of resources (e.g. exceeding the upper limitation of the rate) is said to be non-conforming. The conforming relates actually only to the metering function, i.e. if there are enough resources within the reservation, in terms of the rate or any similar parameter. The action that is applied to conforming traffic is herein called conform action and the action applied to non- conforming traffic is called exceed action. Examples of conform actions is mark and transmit packets. Examples of exceed actions are to drop traffic, mark packets or continue to check a subsequent reservation. The packets may be marked by traffic conditioner, which means that the marked packets can be recognized by the routers. That implies that the routers e.g. can prioritize selected packets. When the packets are marked in a DiffServ network, a DSCP is set.

In accordance with the present invention, traffic matching a filter but exceeding the reserved amount of resources (e.g. rate) of the reservation, continues to search for another matching reservation. This implies that the traffic may search through a plurality of reservations for resources instead of applying an exceed action to the first matched reservation if said first matched reservation is exceeded. In accordance with the present invention, the traffic matching any of the reservations but being non-conforming to all may be treated differently from traffic matching none of the reservations. To exemplify, the traffic matching none of the reservations may be forwarded unmarked while the matching, non- conforming traffic may be dropped

Below is an example of the resulting router configuration when two overlapping reservations are performed by a traffic conditioner implemented in a router: filter 100 allow tcp from 10. 0 . 1 . 0/24 to 10 . 0 . 2 . 0/24 port range 500-

509 filter 101 allow tcp from 10.0.1.1/32 to 10.0.2.0/24 port range 501-

504 rate 64000 match 100 conform-action set-dscp-transmit ef

\ exceed-action set-dscp-continue ef rate 24000 match 101 conform-action set-dscp-transmit ef

\ exceed-action set-dscp-continue ef match dscp ef action drop

Wherein:

"Rate" is the rate of the packets, i.e. the resource.

"Match 100" means match filter 100. "Conform-action" is the action to be performed if the traffic is conformed.

"Set-DSCP-transmit" means that the traffic is marked with an EF (an DSCP), and transmitted. "Exceed-action" is the action to be performed if the traffic is matched, but non-conformed.

"Continue" means continue and investigate if the next filter matches and conforms to the traffic.

"Filter 100" is an example of a unique identifier of a filter. "Allow TCP from 10.0.1.0/24 to 10.0.2.0/24 port range 500-599" means that traffic matches filter 100 if there is TCP-traffic from network 10.0.1.0 to 10.0.2.0 and the used port numbers are within the interval 500 to 599. The filters in the example above are overlapping, since the port ranges of the filters are overlapping. However, overlapping filters may also have other overlapping parameters, or a plurality of parameters that have overlapping ranges. The two overlapping reservations are illustrated in Figure 2 where the port ranges are plotted against the resource type, in this case the reserved rates. The area denoted 200 shows the overlapping part.

In Figure 2, it is illustrated that traffic with the data flow separating parameter destination port 502 matches both reservation 1 and reservation 2. If the strategy is to match only reservation 2 for traffic with the port numbers between 501 and 504, which would limit the resources for all traffic on these ports to 24 kbit/s and the resources of reservation 1 would not be used for these ports. If instead, only reservation 1 is used for traffic with port numbers between 501 and 504, that would limit the traffic with port numbers between 501 and 504 to 64 kbit/s and the extra 24 kbit/s would never be used.

The method according to one embodiment of the present invention comprises the steps of first matching the traffic to reservation 2. If the resources of reservation 2 matches but the metering shows that reservation 2 is exhausted, i.e. is non-conforming, the method comprises the further step of matching the traffic to reservation 1. If the traffic conforms to reservation 2, a conform action is performed. If the traffic is non-conforming to both reservation 1 and 2, i.e. the metering shows that both reservations 1 and 2 are exhausted, a default exceed action is performed, e.g. exceeding traffic is dropped.

In one preferred embodiment, the traffic is matched first to the most specific filter and then continues with more and more general filters. This results in a high utilization, since the most specific reservation is utilized first.

However, in some situation it may be more advantageous to match the traffic to the most general filter first. An example of this situation is at a basic reservation between two different networks defined by the most general filter, when a client in the networks requires a guarantee for a certain amount of traffic even if the basic reservation is exhausted. Thus, the specific filter is only used when the basic reservation is exhausted.

Notice that the order of filters may still have some importance since there may be different situations depending on which reservations are exhausted before others. However, an advantage with the present invention is that the order of the filters is less important than in solutions according to prior art. That depends on that traffic may be matched to more than one overlapping filter even if a match is reached in the present invention while only one match is considered in the prior art.

Thus, the method in a general mode for using reserved resources for data traffic in an IP-network according to the present invention is illustrated in the flowchart of figure 3. The traffic comprises data flow separating parameters. A receiving node adapted to receive the traffic in the IP-network comprises a classifier that comprises a data flow separating filter provided with data flow separating parameters, wherein data flow separating parameters of the at least two filters are overlapping. A resource reservation is defined by the data flow separating parameters of one of the at least two filters together with an amount of a resource, e.g. measured in kbit/s, wherein the resource is a predetermined resource type, e.g. a rate . The method comprises the steps of:

301. Match data flow separating parameters of the received traffic with data flow separating parameters of a first filter of the at least two filters 302. Meter said received traffic, based on said predetermined resource type, in order to establish if said received traffic fits into the reservation defined by the first filter together with the amount of the resource.

When the received traffic matches with said first filter and said traffic does fit into the reservation: 303. Perform a conform action, e.g. mark and transmit packet.

When the received traffic does no match or matches with said first filter but said traffic does not fit into the reservation:

304. Repeat step 301 and 302 for a next filter of the at least two filters.

When the received traffic matches with at least one of said filters and said traffic does not fit into any of the reservations: 305. Perform a first exceed action, e.g. drop said traffic. When the received traffic matches with no one of said filters:

306. Perform a second exceed action. In one preferred embodiment the second exceed action differs from the first exceed action. When the received traffic matches with at least one of said filters and said traffic fits into any of the reservations, repeat step 303.

The method according to the present invention is implemented by means of a computer program product comprising the software code portions for performing the steps of the method. The computer program product is run on a computer located in a node, preferably a router or a server, within an IP network. The computer program product is loaded directly or from a computer usable medium such as a floppy disc, a CD, the Internet etc.

The resource manager 105 is adapted to configure the meter and classifier and other parts of the traffic conditioner 106 for performing the method according to the present invention. The traffic conditioner is preferably located in a router.

The present invention is not limited to the above-described preferred embodiments. Various alternatives, modifications and equivalents may be used. Therefore, the above embodiments should not be taken as limiting the scope of the invention, which is defined by the appending claims.

Claims

1. A method for using reserved resources for data traffic in an IP-network (100), wherein the traffic comprises data flow separating parameters, a receiving node (102) adapted to receive the traffic in the IP-network (100) comprises a classifier (108) comprising a data flow separating filter provided with data flow separating parameters, wherein data flow separating parameters of the at least two filters are overlapping, a resource reservation is defined by the data flow separating parameters of one of the at least two filters together with an amount of a resource, wherein the resource is a predetermined resource type, the method comprises the steps of: a. -matching (301) data flow separating parameters of the received traffic with data flow separating parameters of a first filter of the at least two filters; b. -metering (301) said received traffic, based on said predetermined resource type, in order to establish if said received traffic fits into the reservation defined by the first filter together with the amount of resource, the method is characterised in that it comprises the further step of:

-repeating (304) steps a and b for a next filter of the at least two filters, when the data flow separating parameters of the received traffic matches with the data flow separating parameters of said first filter and said traffic does not fit into the reservation.

2. Method according to claim 1, wherein the method comprises the further step of:

-treating (305,306) the received traffic, that matches data flow separating parameters of at least one filter but does not fit into any reservation, differently from received traffic that does not match data flow separating parameters of any filter.

3. Method according to claim 2, wherein the method comprises the further step of: -dropping said received traffic when said received traffic matches the data flow separating parameters of at least one filter but does not fit into any reservation, or

-forwarding said received traffic unmarked when said received traffic does not match the data flow separating parameters of any filter

4. Method according to any of claims 1-3, wherein the data flow separating parameters comprise at least a port number.

5. Method according to any of claims 1-2, wherein data flow separating parameters of the first filter is equal or more specific than data flow separating parameters of the second filter.

6. Method according to any of claims 1-5, wherein the resource type is a rate.

7. Method according to any of claims 1-6, wherein the receiving node is a router.

8. Method according to claim 7, wherein the router comprises Diffserv capabilities.

9. A computer program product directly loadable into a memory of a node within an IP network comprising the software code portions for performing the steps of claims 1-8.

10. A computer program product stored on a computer usable medium, comprising readable program for causing a computer within a node within an IP network to control the execution of the steps of claims 1-8.

11. A computer program product according to any of claims 9 or 10, wherein the node is a router or a server.

12. Resource manager (105) adapted to configure a meter (107) and a classifier (108) for performing the steps of claims 1-8.

3. Resource manager (105) according to claim 12, wherein the meter (107) and classifier (108) are parts of a traffic conditioner (106) located in a router.