R. Jain. A Comparison of Hashing Schemes for Address Lookup in Computer Networks. IEEE Transactions on Communications, 40(3):1570--1573, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... of Flow Caching for Layer 4 Switching Ye Tung and Hao Che Department of Electrical Engineering Pennsylvania State University State College, PA 16802, USA Abstract Next generation access routers and edge devices need to provide functlonalities for layer 4 packet forwarding and firewall security checks. Consequently, a challenging issue concerns how to achieve fast packet filtering and forwarding at low cost. This paper studies flow caching ....

....When a cache miss occurs, a full header filtering is performed for packet forwarding. This results in a cache miss penalty due to possibly large processing delay. Nevertheless, we note that the cache miss penalty would have little effect on the Quality of Service (QoS) of a connection for 0 7803 6494 5 00 10.00 2000 IEEE 135 the following reasons. A cache miss occurs only for the first packet of a packet burst. This packet can either be the first packet of a connection or the first packet after the connection has been idle for a duration longer than the flow cache timeout value. For ....

[Article contains additional citation context not shown here]

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Transactions on Communications, Vol. 40, No. 3, p. 1570, Oct. 1992. I9] http://www. internet2. edu

....on a combination of IP source field, IP destination field, protocol field, source port field, and destination port field. For the case of MPLS packets, this information may be hashed once at an ingress LSR and the output is stored in a particular label (e.g. bottom label) of a label stack. In [10][11] hashing based on the CRC polynomial was compared to other hashing functions, and was found to offer superior results. We will examine the performance of CRC hashing function in detail in Section IV A. We use m bits (where m = log 2 N ) of the hashed output bits to indicate one of the N ....

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Trans. Commun., pp. 1570-1573, Oct. 1992.

....better than the other algorithms. The only added cost of the Sequent algorithm over BSD is the memory required for the hash chain headers and the computation of the hash function itself. Memory is still decreasing rapidly in price, and efficient hash functions for protocol addresses are well known [Jai89, McK91]. One could imagine combining move to front with hash chains. However, better results can be obtained simply by increasing the number of hash chains. For example, if the number of hash chains in the above example is increased from 19 to 100, the average number of PCBs searched drops from 53 to ....

Raj Jain. A comparison of hashing schemes for address lookup in computer networks. Technical Report DEC-TR-593, Digital Equipment Corporation, February 1989.

....structures and algorithms used by a specific implementation of stochastic fairness queuing. 4.1 HASH FUNCTION The example implementation uses a hash function to map from sourcedestination address pair to queue index. This hash function must give a high information content, as defined by Jain in [Jai89], but must also allow perturbation such that address pairs that collide for one perturbation value are very unlikely to collide for a different perturbation value. Two hash functions were used in simulations. The first is the highlevel data link control (HDLC) procedure (ISO 3309 1979) cyclic ....

Raj Jain. A comparison of hashing schemes for address lookup in computer networks. Technical Report DEC-TR-593, Digital Equipment Corporation, February 1989.

....hash function would be CRC polynomials, which have good hashing properties. Unfortunately, calculating the CRC is easy when data arrives serially, but is more complex to do on a parallel word of data. There are, however, other solutions, such as XOR folding, that have also proved to work 8 well [5]. In our simulations we have used both kinds of functions with good results. The XOR folding is probably the preferred choice in an implementation, being very simple and easy to vary. 4.5 Performance The hash function can be implemented by standard programmable components such as Programmable ....

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Transactions on Communications, vol. 40, no. 10, pp. 1570-1573, 1992.

....the lowest arrivalrates. At each packet arrival, RED executes RED with different parameters for penalized and non penalized flows and updates counters measuring the arrivalrates of single flows and the total arrivalrate at the output port. 3 For the flow lookup the execution of a hash function [19] at each packet arrival can not be avoided with local algorithms. DRAFT DRAFT DRAFT DRAFT DRAFT DRAFT DRAFT DRAFT 11 3.4 RED State Machine Figure 10 RED state machine A flow is in one of four possible states: non existent: a flow is in state non existent if it has no ....

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer networks", DECTR -503

....protocol. In comparison, the hashing based schemes can maintain per flow packet ordering and can be implemented without requiring any additional protocol support. Hashing has been widely used in indexing and searching [9] In the networking context, hashing based algorithms for address lookup [10], flow identification [11] and packet demultiplexing [12] have been proposed in the past. The use of hashing for network load balancing is not new. Some commercial router products have implemented simple hashing over the IP destination address to distribute traffic [13] In the OSPF Optimized ....

....bits of the destination address as an index of the outgoing link. This hash function has been implemented by router vendors. A. 2 Hashing Using XOR Folding of Destination Address XOR folding has been used in many hash functions, and has been shown to provide good performance in other applications [10]. We propose a hash function with XOR folding of the destination IP address. This hash function can be expressed as: H( Delta) D 1 Phi D 2 Phi D 3 Phi D 4 ) mod N where D i is the ith octet of the destination IP address. This approach utilizes more bits of the destinationaddress in ....

Raj Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Transactions on Communications, Oct. 1992.

....enough to change their entries as new connections are established and old ones released. While not as frequent as cell arrivals, connection procedures should be able to change switching table entries within milli seconds to avoid high latency. Static lookup table techniques, such as hashing [5] [10], 11] are therefore inappropriate. The simplicity and regularity of ubiquitous Random Access Memory (RAM) provides designers with low cost, fast, high density memories. Conventional single RAM lookup tables use the input number to address a location in RAM and retrieve from that location the ....

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks", IEEE Transactions on Communications, vol. 40, no. 10, pp. 15701573, Oct. 1992.

....of various lookup schemes. time in a rigorous sense; the source hashing lookup guarantees are independent of assumptions about the probability distributions of addresses or the details of the particular topology used. This is in stark contrast to many standard hashing schemes like XOR folding[Jai92] that are strongly dependent on the way addresses are assigned in particular networks 5 . The memory requirements for source hashing are a constant factor (a factor of two should be sufficient for most purposes) larger than the memory required for ordinary VCs; this extra factor is used to make ....

....VCI. These differences are explored more carefully in Section 2 and summarized in Table 1. It is hard to characterize the cost of lookups because they are so implementation dependent. For instance, hashing by exclusive OR folding of address octets seems effec tive and easy to implement [Jai92]. In [CJRS89] the cost of hashing (in the best case) is taken to be 25 instructions. However, we believe the lookup techniques described in this paper are important because: 1) Lookups are a critical part of cache processing and cache processing is becoming increasingly important for high speed ....

Raj Jain. A comparison of hashing schemes for address lookups in computer networks. IEEE Transactions on Communications, COM40 (10):1570--1573, October 1992.

....be implemented considering the network topology, for instance in gateways located at the edge of an isolated LAN. The use of a hash function is suggested for address lookup (i.e. matching of packets to their corresponding perconversation queues) Efficient hashing schemes are proposed in [10] and [18]. An interaction of RFCN and the BUC algorithm with the address lookup scheme has to be considered: Some hashing schemes may generate more than one queue per conversation at an output port in case of collusions. If this happens, the BUC algorithm has to calculate only one window based on the total ....

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Transactions on Communications, Vol. 40, No. 3, October 1992, pp. 1570-1573.

....lookup costs are important. Similarly, generic techniques to reduce packet processing are also important. It is hard to characterize the cost of lookups because they are implementation dependent. For instance, hashing by exclusive OR folding of address octets seems effective and easy to implement [Jai92]. In [CJRS89] the cost of hashing (in the best case) is taken to be 25 instructions. However, we believe the lookup techniques described in this paper are useful because: 1) Our techniques do not depend on traffic patterns. 2) The lookup costs accumulated over several layers do add up. Lookups ....

....datagram network ( Par93] one must either assign flow IDs to packets (which is not a solved problem) or infer a flow by hashing multiple address fields (e.g. source destination addresses) which exacerbates the lookup problem. An existing technique to reduce the cost of state lookups is caching [Par93, Jai92]. A generic technique used to reduce generic protocol processing is Header Prediction[CJRS89] These existing techniques depend on traffic patterns ( Jai86] and do not guarantee performance improvement. By contrast, our schemes use extra packet headers to provide performance improvements that are ....

Raj Jain. A comparison of hashing schemes for address lookups in computer networks. IEEE Transactions on Communications, COM-40(10):1570--1573, October 1992.

....connections sharing the same address pair are treated first come first served. The conversation ID is used to access a data structure for storing state. Since IDs could span large address spaces, the standard solution is to hash the ID onto a index, and the technology for this is well known [9]. Recently, a simple and efficient hashing scheme that ignores hash collisions has been proposed [5] In this approach, some conversations could share the same state, leading to unfair service, since these conversations are served first come first served. However, this is attenuated by ....

Jain, R., A Comparison of Hashing Schemes for Address Lookup in Computer Networks , Tech. Rpt.-593, Digital Equipment Corporation , February 1989.

....set up delay. We use expected time in a rigorous sense; the source hashing lookup guarantees are independent of assumptions about the probability distributions of addresses or the details of the particular topology used. This is in stark contrast to many standard hashing schemes like XOR folding [Jai92] that are strongly dependent on the way addresses are assigned in particular networks. 3 The memory requirements for source hashing are a constant factor (a factor of two should be sufficient for most purposes) larger than the memory required for ordinary VCs; this extra factor is used to make ....

....traditional VCI. These differences are explored more carefully in Section 2 and summarized in Table 1. It is hard to characterize the cost of lookups because they are so implementation dependent. For instance, hashing by exclusive OR folding of address octets seems effective and easy to implement [Jai92]. In [CJRS89] the cost of hashing (in the best case) is taken to be 25 instructions. However, we believe the lookup techniques described in this paper are important because: 1) Lookups are a critical part of cache processing and cache processing is becoming increasingly important for high speed ....

Raj Jain. A comparison of hashing schemes for address lookups in computer networks. IEEE Transactions on Communications, COM40 (10):1570--1573, October 1992.

....The information in m consecutive bits of address checksums is shown in Figure 4. This also is a good hashing function. 6 Hashing Using Another Checksum Another popular checksum algorithm used to guard against memory errors in network address databases is [2] C = Mod 0 2 8 (4b[1] 2b[3] b[5] 4b[2] 2b[4] b[6] 2 1 6 0 1 1 Here, b[i] is the ith octet of the address and C is the 16 bit checksum. Since we are not aware of its name, we will call it the mod checksum. The information content of the bits of this checksum are shown in Figure 5. Notice that the ....

....not as good a hashing function as the Fletcher checksum even though it is more complex to compute. 7 Hashing Using XOR Folding The final alternative that we investigated is that of the straightforward exclusive or operation on the six octets of the address to produce 8 bits. C = b[1] 8 b[2] 8 b[3] 8 b[4] 8 b[5] 8 b[6] The information content of the bits in the XOR fold so obtained is shown in Figure 6. To our surprise, this function, which is so simple to implement, is an excellent hashing function. 8 Mask Size for an Address Filter In this section, we briefly address the problem of ....

[Article contains additional citation context not shown here]

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," DEC Technical Report, DEC-TR-593, February, 1989, 17 pp.

No context found.

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," DEC Technical Report, DEC-TR-593, January 1989, (Available from the author).

No context found.

R. Jain. A Comparison of Hashing Schemes for Address Lookup in Computer Networks. IEEE Transactions on Communications, 40(3):1570--1573, 1992.

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Transactions on Communications, 1992.

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Transactions on Communications, 1992.

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Trans. on Communications, 4(3):1570--1573, October 1992.

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Transactions on Communications, 40(3):1570-- 1573, October 1992.

No context found.

R. Jain, "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Transactions on Communications , Oct. 1992.

No context found.

. 14. Jain, R., "A Comparison of Hashing Schemes for Address Lookup in Computer Networks," IEEE Trans. on Communications, 40, 10, pp. 1570-1573

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Transactions on Communications, 1992.

No context found.

R. Jain. A comparison of hashing schemes for address lookup in computer networks. IEEE Transactions on Communications, 1992.

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