V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search." in Proceedings of the ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, 1999, pp. 135--146.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....classifiers, but the storage requirements are still high (up to 3MB) Also, we have as yet been unable to provide a mechanism for doing incremental updates to the data structure. Another algorithm called Tuple Space Search has been recently proposed for packet classification on multiple fields [2]. The scheme partitions the rules of a classifier into different tuple categories based upon the number of specified bits in the rules (a bit is specified in a rule if it is not a don t care bit) The scheme then uses hashing among rules within the same category. The main advantages of this ....

V. Srinivasan, S. Suri and G. Varghese, "Packet Classification using Tuple Space Search", Proc. ACM SIGCOMM 1999, pp 135-146.

....uses a version trie of tries designed for 2 tuple prefix matches. The Area BasedQuadtree proposed in [4] improves upon the Grid of Tries and exports to designers the tradeoff between lookup efficiency and insert efficiency by means of a configuration parameter. The tuple space search proposed in [25] decomposes a 5 tuple prefix matching problem into a series of exact match lookups performed using hashing. All the above algorithms can be implemented either in software or in hardware. The literature also contains several other algorithms specifically designed for hardware implementation ....

V. Srinivasan, Subhash Suri, and George Varghese. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM, pages 135--146, 1999.

....of the real time and the non real time packet handling. Typically, the determination of the application set (Network Application Profile, NAP) would be done o# line whereas the determined NAP would be used in conjunction with the flow and packet classification systems [10] 7] 11] 12] 13] [14] in real time fashion. The first explicit information on the sending application in the Internet appear at the TCP UDP source and destination numbers. An IP application is, in this work, defined to be a group of packets originating from an identical source (TCP UDP) port. Many of the new ....

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in Proceedings of SIGCOMM 99. 1999, IEEE/ACM.

....queue. Either per flow or aggregate queues may be used, depending on the QoS requirements. The filter database determines whether flows are aggregated or handled separately. To provide the required flexibility, a fast general flow classification algorithm is required, such as the one described in [39]. The design can be scaled in a couple ways. First, the number of ports can be increased by configuring the multistage interconnection network to have a larger number of stages. For the design in [7] a three stage network can support up to 64 ports and has an aggregate capacity of 154 Gb s, ....

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proc. of ACM SIGCOMM 99, Cambridge, MA, Sept. 1999.

....R matching P. We state that R matches P if for each R , P(R) P(R ) and for each C i , l i C i r i where [l i ;r i ] E(R) Current packet classification proposals may be divided into two main classes. The first one is dedicated to the classification of packets according to static policies [19], 20] 21] Static classification policies are policies where the rule set describing the classification policy almost never changes. Static policies can be opposed to dynamic policies where rules can be added or removed frequently. Both static and dynamic classification algorithms [14] 22] ....

V.Srinivasan, S. Suri, G. Varghese, Packet Classification Using Tuple Space Search, in proc. of the ACM SIGCOMM'99 Conference, September 1999.

....router mechanisms apply to individual packet flows, i.e. streams of packets that belong to the same application session. IntServ requires that routers perform the following operations on the packets of a flow: ffl flow classification, to identify the flow in which a packet belongs to [57, 100], ffl scheduling, to provide a certain delay deadline or rate to a flow [101, 102] ffl buffer management, to allocate a number of buffers to a flow [107] and ffl traffic shaping or policing, to control certain traffic characteristics (e.g. maximum burstiness) of a flow [20, 101] The ....

....the host network interface. In that case, the operating system or the application determines the macroflow that the flow belongs to. Or, the edge can be a router that connects a microflow aware network to a DiffServ network. The mapping from microflows to macroflows requires flow classification [100]. This is a relatively expensive operation, but since it is only performed at the network edges, where the number of microflows is much smaller, it is expected to not cause scalability problems. Microflows are aggregated in macroflows based on rules set by the network operator. For instance, a ....

[Article contains additional citation context not shown here]

V. Srinivasan, S.Suri, and G.Varghese, "Packet Classification using Tuple Space Search," In Proceedings ACM SIGCOMM, September 1999.

.... flows that are classified [16] Since a SV CSFQ router needs to maintain filters only for flows in Ve r fyTable and ContainTable, and since, as discussed in Section III, these tables are small, we can achieve high speed classification by using any of the algorithms proposed recently [16] 17] [18], 19] A simple hash table with the hash keys computed over the IP source and destination addresses would be sufficient for most practical purposes. Finally, flow verification and containment are constant time operations, and have a similar processing overhead as estimating the flow rate in ....

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in Proceedings of ACM SIGCOMM'99, Cambridge, MA, Sept. 1999, pp. 135--146.

....associate packets into a flow in the same data structure to keep the classification overhead as low as possible. Much research on how to e#ciently do high speed and low overhead classification in both software and hardware has been done in the area of designing firewalls and similar devices (e.g. [22, 52, 3]) and our architecture can take advantage of these advances. After classification and before the state machine is entered, the proper descriptionand flow specific memory pages must be located. Our design does not define how this is done in detail, but it is best done integrated with the ....

V. Srinivasan, Subhash Suri, and George Varghese. Packet classification using tuple space search. In SIGCOMM, pages 135--146, 1999.

....SWITCHGEN, a system for circuit synthesis and layout specialized for the domain of pattern matching circuits implemented in reconfigurable logic. We propose to use this system to implement high throughput pattern classification, for instance as part of a packet filter in a internetwork router. [2, 4] In that case, packet filtering rules become embedded in the topology of the circuit. The goals of the approach are throughputs on the order of 100M classifications per second with reconfiguration times (including all synthesis) on the order of 10 seconds. The main idea is to observe that a ....

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification using Tuple Space Search. In Proceedings of SIGCOMM'99, 1999.

....with the IP address 15:14:51:12 does not. It is worth noting that routing is just a particular case of packet classification, in which each filter is specified by only one field: dst addr. It should come as no surprise that the classification problem is inherently difficult. Current solutions [51, 66, 96, 97] work well only for a relatively small number of classes, i.e. no more than several thousand. This is because, as noted by Gupta and McKeown [51] the packet classification problem is similar to the point location problem in the domain of computation geometry. Given a point in an F dimensional ....

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM'99, pages 135--146, Cambridge, MA, September 1999.

....Label Switched Paths (LSPs) Service providers can perform this encapsulation at the ingress routers, and then use LSPs to implement Virtual Private Networks (VPNs) 6] or satisfy other quality of service (QoS) agreements with clients. At the ingress routers, packet classification [9] 11] [12] can be used to map packets into forwarding equivalence classes by examining packet headers. This aggregation (mapping into equivalence classes) also has the potential advantage of smoothing out the bandwidth requirement across many bursty streams. In addition, the service providers can use a ....

V. Srinivasan, S. Suri and G. Varghese. Packet Classification using Tuple Space Search. Proc. of Sigcomm, 1999.

....done with it. Web requests, for example, are identified as TCP packets destined for port 80. A packet classifier would take a packet, or packet header, as input and return as a result the target application. Much work has been done to design packet classifiers in both hardware [2, 11] and software [5, 8, 21]. Packet classification in network devices has often been implemented in custom silicon in the interests of high performance. However, the state of the art today in software based packet classification provides good performance and greater flexibility than ASIC solutions. In this study, we ....

....performance tradeoffs between hardware and software based classification. We use Engler s publicly available DPF packet filter[8] which uses dynamic code generation to create highly optimized software packet filters. While no longer the fastest software implementation described in the literature[5, 21], DPF has some nice properties including performance that is insensitive to the number of filters and good portability. We compare the performance of this software implementation to the performance of an ideal hardware implementation that classifies packets instantaneously. We believe this to be a ....

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification Using Tuple Space Search. Proceedings of the ACM Communication Architectures, Protocols, and Applications (SIGCOMM '99), 1999.

.... (when combined with previous techniques such as prefix expansion) and for new programmable network processors [4] We expect that it will also be useful for similar problems, such as packet classification and filtering, where hashing is commonly used as a subroutine to allow fast lookups [13]. The basic idea of the approach is to use multiple hash functions. The idea has been analyzed and developed in several recent theoretical works. We therefore specifically address how this approach can be used to improve performance on the real problem of IP lookups. In particular, we emphasize ....

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification using Tuple Space Search. In Proc. of SIGCOMM '99, pp. 135--146.

....and certain applications may even be generated automatically, the lookup database for a busy router may become quite large. Performance that scales well with the number of database entries is thus an important goal. We use the following classification algorithm that uses the tuple concept in [19]. Given an input packet, a trie based search is used to find in the lookup database the longest prefix matches for the IP source and destination address fields, respectively. All the database source (respectively destination) address prefixes that are prefixes of the lookup source (respectively ....

....that are prefixes of the lookup source (respectively destination) address will be marked during the search. Each marked prefix points to a set of tuples that contain the prefix in at least one of their entries. A tuple is a table of all database entries with given lengths in each dimension [19]. The intersection of the tuple sets marked by the source and destination longest prefix matches is then searched. Each tuple search first computes a hash key that is a concatenation of specified bit positions in each dimen . CPU scheduler Network scheduler Disk scheduler Memory ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proc. ACM SIGCOMM, Cambridge, MA, Sept 1999.

....Table 2. Benchmark characteristics. 4 of 12 schedule these concurrent operations such that each pipeline stage is occupied. Varying degrees of programmability may be used to implement each of these stages. Much work has been done to design packet classifiers in both hardware [2, 8] and software [3, 5, 16]. For this study, we assume that all packet classification is performed in dedicated hardware that is ideal in the sense that it is never the performance bottleneck in the system. The experiments discussed in this paper consider both hardware and software implementations of the packet storage ....

....than FGMT and SS by dynamically exploiting both instruction and thread level parallelism. For simple applications, such as IP forwarding, SMT and CMP can sustain network speeds exceeding 10Gbits second (e.g. OC12 links) Higher speeds may be attained with better IP forwarding algorithms, such as [16]. For computationally intensive applications, such as MD5, both SMT and CMP sustain network speeds approaching 1Gbits (e.g. gigabit Ethernet) There are a number of questions we plan to investigate in future work. First, we will augment some of the experimental architectures used in this study. ....

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification Using Tuple Space Search. Proceedings of the ACM Communication Architectures, Protocols, and Applications (SIGCOMM '99), 1999.

....queue. Either per flow or aggregate queues may be used, depending on the QoS requirements. The filter database determines whether flows are aggregated or handled separately. To provide the required flexibility, a fast general flow classification algorithm is required, such as the one described in [7]. The design can be scaled in a couple ways. First, the number of ports can be increased by configuring the multistage interconnection network to have a larger number of stages. For the design in [2] a three stage network can support up to 64 ports and has an aggregate capacity of 154 Gb s, ....

Srinivasan, V., Suri, S., Varghese, G. [1999]. "Packet Classification using Tuple Space Search," to appear in Proc. of SIGCOMM 99, ACM, Cambridge, Mass.

....algorithms that interpret filters as geometric shapes, and map the packet classification problem to some form of geometric point location problem. The latter refers to any other approach based on regular data structures such as trees and graphs. 8] and [2] belongs to the former, while [7] 4] [6], 5] and ours belong to the latter. 8] presents 2 algorithms. The first algorithm admits a hardware implementation but does not readily scale to a large number of filters. The second algorithm applies only to 2 dimensions, and it does not appear to easily generalize to higher dimensions. 2] is ....

....heuristic bit selection and the concept of a filter bucket, are unique to ours. Since no formal description was given for the DAG algorithm 10 , it is not clear how the algorithm scales to large number of filters. Again, the average case comparison with ours should be interesting. Both [6] and [5] represent new interesting approaches to packet classification. Recognizing the inherent difficulty of the problem, both try to exploit structure within a filter table to improve search performance. Specifically, 6] makes use of the observation that there is typically only a small number ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proceedings of ACM Sigcomm, pages 135--146, Cambridge, Massachusetts, August 30--September 3 1999.

No context found.

V.Srinivasan, S.Suri, and G.Varghese. Packet classification using tuple space search. In Proc of ACM Sigcomm'99, september 1999.

No context found.

V.Srinivasan, S.Suri, and G.Varghese, "Packet classification using tuple space search," in SIGCOMM, 1999.

....prefix of the packet header field this is useful for blocking access from a specified subnetwork. In a range match, the header values should lie in the range specified by the rule this is useful for specifying port number ranges. Ranges, however, can be converted into prefixes as shown in [1] [2]. Each rule ### has an associated action , which specifies how to forward the packet matching this rule. The action may specify if the packet should be blocked or if it is to be forwarded, it specifies the outgoing link on which the packet is to be sent, and perhaps also a queue within that ....

....of memory accesses required by an operation (the main limitation in modern computer architectures) and the memory size occupied by data structures (because it is important to fit into high speed memory) II. PREVIOUS WORK Packet filter classification has received broad attention ( 6] 1] 4] [2], 5] 7] 8] 9] 10] from previous work, it appears that the general problem is inherently hard (in a worst case sense) when the filters contain more than 2 fields. While Ternary CAMs [11] offer a good solution in hardware for small classifiers, they use too much power and do not scale ....

[Article contains additional citation context not shown here]

V.Srinivasan S.Suri G.Varghese, "Packet classification using tuple space search," in Proceedings of ACM Sigcomm'99, september 1999.

....prefix of the packet header field this is useful for blocking access from a specified subnetwork. In a range match, the header values should lie in the range specified by the rule this is useful for specifying port number ranges. Ranges, however, can be converted into prefixes as shown in [1] [2]. Each rule R i has an associated action act , which specifies how to forward the packet matching this rule. The action may specify if the packet should be blocked or if it is to be forwarded, it specifies the outgoing link on which the packet is to be sent, and perhaps also a queue within ....

....of memory accesses required by an operation (the main limitation in modern computer architectures) and the memory size occupied by data structures (because it is important to fit into high speed memory) II. PREVIOUS WORK Packet filter classification has received broad attention( 6] 1] 4] [2], 5] 7] 8] 9] 10] from previous work, it appears that the general problem is inherently hard (in a worst case sense) when the filters contain more than 2 fields. While Ternary CAMs [11] offer a good solution in hardware for small classifiers, they use too much power and do not scale ....

[Article contains additional citation context not shown here]

V.Srinivasan S.Suri G.Varghese, "Packet classification using tuple space search," in Proceedings of ACM Sigcomm'99, september 1999.

....references. By making multiple classification algorithms publicly available we hope to encourage experimentation and improvements that can then be incorporated into revisions on the same web site. III. PRIOR WORK AND SUMMARY OF RESULTS The packet classification problem is inherently hard( 8] [9], 3] 7] 4] 10] from a theoretical standpoint. It has been shown [8] that in its fullest generality, packet classification requires either O(log N) time and linear space, or log N time and O(N ) space, where N is the number of rules, and k is the number of header fields used in ....

V.Srinivasan, S.Suri, and G.Varghese, "Packet classification using tuple space search," in Proc of ACM Sigcomm'99, september 1999.

....filter database or to use hardware, such as ternary CAMs (content addressable memory) or other ASICs that perform parallel linear search (e.g. 4] Such hardware solutions do not scale to large filter databases. Other solutions reported in literature that can be implemented in software (e.g. [9, 5]) are either slow or take too much storage. With the advent of software based routers (e.g. 7] which are typically aimed at the edge router space where classification is particularly important, it is necessary to find software techniques for fast firewall implementations. There is evidence ....

....it is necessary to find software techniques for fast firewall implementations. There is evidence that the general filter problem is a hard problem, and requires either O(N ) memory or #( log N) search time, where N is the number of filters and K is the number of classified fields [4, 9]. However recent research [5, 6, 10] indicates that such worst case behavior does not arise in real databases. Based on this observation, these papers introduce clever new techniques like pruned tuple search [10] and Recursive Flow Classification [5] that exploit the structure of existing ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification using Tuple Space Search. In Proc. of SIGCOMM'99, Sept. 1999.

....can be thought of as one dimensional packet classification. While several efficient solutions are known for the onedimensional IP lookup problem, the multi dimensional packet classification has proved to be far more difficult. While an time scheme is known for the IP lookup, Srinivisan et al. [1] show a lower bound of ### # for # dimensional filter lookup, where is the number of bits in a header field. In particular, this lower bound precludes the possibility of a binary search like scheme even for 2dimensional filters (say, IP source and destination pairs) In this paper, we ....

....Packet classification using ad hoc mechanisms like linear search through all filtering rules is too slow in practice and a significant source of bottleneck. Hence the problem has received some attention in last # years. In particular, the tuple space framework proposed by Srinivasan et.al. [1] and associated simulation results suggest significant reduction in search space, while keeping memory requirement almost linear. The tuple space is formed by distinct combinations of prefix lengths (#) in the filter set. For filters containing IP prefixes, maximum prefix length for fields is # # ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in Proceedings of SIGCOMM'99, 1999.

....connection between two different sites of a company. Such refined forwarding is part of the next generation Internet design, and falls within the broader scope of layer four packet classification, where packets are routed using arbitrary fields of the packet header [1] 9] 10] 12] 15] [16], 17] Routers capable of packet classification can implement many advanced services, such as firewall access control, Virtual Private Networks, and quality of service routing. In this paper we focus on a particular problem that arises in the context of using twodimensional routing tables many ....

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification Using Tuple Space Search. Proc. of ACM SIGCOMM, 1999.

....references. By making multiple classification algorithms publicly available we hope to encourage experimentation and improvements that can then be incorporated into revisions on the same web site. III. PRIOR WORK AND SUMMARY OF RESULTS The packet classification problem is inherently hard( 11] [18], 7] 17] 6] 19] from a theoretical standpoint. It has been shown [11] that in its fullest generality, packet classification requires either time and linear space, or time and space, where is the number of rules, and is the number of header ....

V.Srinivasan, S.Suri, and G.Varghese. Packet classification using tuple space search. In Proc of ACM Sigcomm'99, september 1999.

....Paths (LSPs) Service providers can perform this encapsulation at the ingress routers, and then use LSPs to implement Virtual Private Networks (VPNs) G 00] or satisfy other quality of service (QoS) agreements with clients. At the ingress routers, packet classification [LS98, SVSW98, SSV99] can be used to map packets into forwarding equivalence classes by examining packet headers. This aggregation (mapping into equivalence classes) also has the potential advantage of smoothing out the bandwidth requirement across many bursty streams. In addition, the service providers can use a ....

V. Srinivasan, Subhash Suri, and George Varghese. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM, pages 135--146, Cambridge, Massachusetts, USA, September 1999.

....along specific Label Switched Paths (LSPs) Service providers can perform this encapsulation at the ingress routers, and then use LSPs to implement Virtual Private Networks (VPNs) 4] or satisfy other quality of service (QoS) agreements with clients. At the ingress routers, packet classification [5 7] can be used to map packets into forwarding equivalence classes by examining packet headers. This aggregation (mapping into equivalence classes) also has the potential advantage of Initial work performed at Washington University in St. Louis, partially supported by NSF grant ANI 9813723. An ....

V. Srinivasan, S. Suri, G. Varghese, Packet classification using tuple space search, in: Proceedings of ACM SIGCOMM '99, Cambridge, MA, USA, 1999, pp. 135--146.

No context found.

SRINIVASAN,V.,SURI, S., AND VARGHESE, G. 1999. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM '99 (Cambridge, Massachusetts, USA, Sept. 1999), pp. 135--146.

....the theoretical and practical contributions of this paper. 4 M. Waldvogel, G. Varghese, J. Turner, and B. Plattner 2. COMPARISON OF EXISTING ALGORITHMS As several algorithms for efficient prefix matching lookups have appeared in the literature over the last few years (including a recent paper [Srinivasan and Varghese 1999] in ACM TOCS) we feel that it is necessary to structure the presentation of related work using a taxonomy. Our classification goes beyond the lookup taxonomy recently introduced in [Ruiz S anchez et al. 2001] However, the paper [Ruiz S anchez et al. 2001] should be consulted for a more in depth ....

....times. For applications where search speed is much more important than update speed or worst case memory consumption, such as for Internet forwarding lookups, more aggressive search time optimization is required. To reduce the number of levels that need to be touched, Controlled Prefix Expansion [Srinivasan and Varghese 1999] selects a small number of prefix lengths to be searched. All database entries that are not already of one of these lengths, are expanded into multiple entries of the next higher selected length. Depending on the length of the strides # between the selected lengths and the prefix length ....

[Article contains additional citation context not shown here]

SRINIVASAN, V., SURI, S., AND VARGHESE, G. 1999. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM '99 (Cambridge, Massachusetts, USA, Sept. 1999), pp. 135--146.

....thought of as one dimensional packet classification. While several efficient solutions are known for the onedimensional IP lookup problem, the multi dimensional packet classification has proved to be far more difficult. While an O(log w) time scheme is known for the IP lookup, Srinivisan et al. [1] show a lower bound of w k 1 ) for k dimensional filter lookup, where w is the number of bits in a header field. In particular, this lower bound precludes the possibility of a binary search like scheme even for 2dimensional filters (say, IP source and destination pairs) In this paper, we ....

....Packet classification using ad hoc mechanisms like linear search through all filtering rules is too slow in practice and a significant source of bottleneck. Hence the problem has received some attention in last 2 years. In particular, the tuple space framework proposed by Srinivasan et.al. [1] and associated simulation results suggest significant reduction in search space, while keeping memory requirement almost linear. The tuple space is formed by distinct combinations of prefix lengths (w) in the filter set. For filters containing IP prefixes, maximum prefix length for fields is w = ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in Proceedings of SIGCOMM'99, 1999.

....of as one dimensional packet classification. While several efficient solutions are known for the onedimensional IP lookup problem, the multi dimensional packet classification has proved to be far more difficult. While an 99 time scheme is known for the IP lookup, Srinivisan et al. [1] show a lower bound of for dimensional filter lookup, where is the number of bits in a header field. In particular, this lower bound precludes the possibility of a binary search like scheme even for 2dimensional filters (say, IP source and destination pairs) In this paper, we ....

....Packet classification using ad hoc mechanisms like linear search through all filtering rules is too slow in practice and a significant source of bottleneck. Hence the problem has received some attention in last years. In particular, the tuple space framework proposed by Srinivasan et.al. [1] and associated simulation results suggest significant reduction in search space, while keeping memory requirement almost linear. The tuple space is formed by distinct combinations of prefix lengths ( in the filter set. For filters containing IP prefixes, maximum prefix length for fields is ....

[Article contains additional citation context not shown here]

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in Proceedings of SIGCOMM'99, 1999.

....composite and simple fil ters . In Section II we define the best matching filter problem. In Section III we describe related work. In Section IV we describe the entry pruned tuple search idea and describe how to blend it with a previous algorithm based on markers and precomputation presented in [7]. In Section V we describe our fil ter policy management architecture. In Section VI we describe our incremental update method. In Section VII we present our performance results. We conclude in Section VIII. II. THE BEST MATCHING FILTER PROBLEM The rules for classifying a message are called ....

....to only exact and prefix matches. However, other types such as ranges are possible. Studies [8] have shown that in real life databases about ### of rules have range specifications. Small ranges can be expanded into exact matches and large ranges can be expanded into an equivalent set of prefixes [7]. We let # denote the number of filters in a filter database # ### . Consider a filter database # ### with # filters. Given a header # , there can be several filters #### # in # ### that match # . With each filter #### # , there is an associated cost function #### # . The cost of a filter # is ....

[Article contains additional citation context not shown here]

V. Srinivasan, G. Varghese, and S. Suri, "Packet Classification using Tuple Space Search," Proceedings of ACM Sigcomm'99, vol. (August), 1999.

....Label Switched Paths (LSPs) Service providers can perform this encapsulation at the ingress routers, and then use LSPs to implement Virtual Private Networks (VPNs) 6] or satisfy other quality of service (QoS) agreements with clients. At the ingress routers, packet classification [9] 11] [12] can be used to map packets into forwarding equivalence classes by examining packet headers. This aggregation (mapping into equivalence classes) also has the potential advantage of smoothing out the bandwidth requirement across many bursty streams. In addition, the service providers can use a ....

V. Srinivasan, S. Suri and G. Varghese. Packet Classification using Tuple Space Search. Proc. of Sigcomm, 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search." in Proceedings of the ACM SIGCOMM Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, 1999, pp. 135--146.

No context found.

V. Srinivasan, Subhash Suri, and George Varghese. Packet classification using tuple space search. In Computer ACM SIGCOMM Communication Review, pages 135--146, October 1999.

No context found.

V. Srinivasan, Subhash Suri, and George Varghese. Packet classification using tuple space search. In Computer ACM SIGCOMM Communication Review, pages 135--146, October 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet classificationusing tuple space search. In Proc. ACM SIGCOMM, pages 135--146, 1999. http://citeseer.ist.psu.edu/srinivasan99packet.html.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proceedings of SIGCOMM 99. IEEE/ACM, 1999.

No context found.

V. Srinivasan et al., "Packet Classification Using Tuple Space Search," Proc. ACM SIGCOMM, Sept. 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in SIGCOMM 99, pp. 135--146, 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in SIGCOMM 99, pp. 135--146, 1999.

No context found.

V. Srinivasan, S. Suri, M. Waldvogel, "Packet classification using tuple space search", In Proceedings of ACM Sigcomm'99, September 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese, "Packet classification using tuple space search," in SIGCOMM 99, pp. 135--146, 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM '99, pages 135--146, Boston, MA, August 2001.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proceedings of ACM SIGCOMM'99, pages 135--146, Cambridge, MA, September 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet classification using tuple space search. In Proc. ACM Sigcomm, pages 135--146, Sept. 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese. Packet Classification using Tuple Space Search. Computer Communication Review, 29(4):135--146, October 1999.

No context found.

V. Srinivasan, S. Suri and G. Varghere, "Packet Classification Using Tuple Space Search," SIGCOMM, pp. 135-146, 1999.

No context found.

V. Srinivasan, S. Suri, and G. Varghese, "Packet Classification using Tuple-Space Search," Proc. SIGComm 99, ACM Press, New York, pp. 135-146.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC