The accelerated growth of Internet applications has increased the importance of the development of new network services, such as security, virtual private network (VPN), quality of service (QoS), accounting, and so on. All of these mechanisms generally require the router to be able to categorize packets into different classes called flows. The categorization function is termed packet classification.
An Internet router categorizes incoming packets into flows utilizing information contained in the packet header to lookup the predefined rule table in the router. A rule table maintains a set of rules specified based on the packet header fields, such as the network source address, network destination address, source port, destination port, protocol type and possibly other fields. The rule field can be a prefix (e.g. a network source/destination address), a range (e.g. a source/destination port) or an exact number (e.g. a protocol type).
The process of packet classification is presented in Fig. 1. When a packet arrives, the packet header is extracted first and then compared with the corresponding fields of rule in the rule table. A rule matching in all corresponding fields is considered a matched rule. The packet header is compared with every rule in the rule table, and the matched rule with the highest priority yields the best-matching rule. Finally, the router performs an appropriate action associating with the best-matching rule.
Table 1 is an example of a rule table. Here, the address fields are shown as 3 bits prefix. A star in address fields indicates a bit mask, and stars for entire entry indicates a wildcard which can be matched by any packet. The port fields are shown with range, and protocol field is shown with exact protocol type, including TCP, UDP and ICMP.
Rules are arranged in order of priority. Each rule has an associated action either Deny
or Pass. Consider a packet P with source address of 110, destination address of 010, source port of 4, destination port of 9 and protocol type of TCP arrives. In table1, packet P matches rule 1 and rule 4, where rule 1 has higher priority. Therefore rule 1 is the best-matching rule for packet P. According to the associated action of rule 1, packet P is denied.
Header Payload Figure 1: Packet classification process
Rule Source
Table 1: An example of packet classification rule table
Longest prefix matching [17] for route lookup is a special case of one-dimensional packet classification. Each rule is described by a prefix (address/mask pair). The length of the prefix defines the priority of the rule.
The d-dimensional packet classification problem (PC problem) is formally defined as follows. The rule table has a set of rules R = {R1, R2…, Rn} over d dimensions.
Each rule comprises d fields Ri = {F1,i, F2,i…, Fd,i}, where Fj,i denotes the value of field j in rule i. Each rule also has a cost (priority). A packet P(p1,p2…,pd) matches rule Ri if all the header fields pm, m from 1 to d, of the packet match the corresponding fields Fj,i in Ri. If packet P matches multiple rules, the minimal cost (highest priority) rule is returned.
The general packet classification problem can be viewed as a point location problem in multidimensional space [1]. Rules have a natural geometric interpretation in d dimensions. Each rule Ri can be considered a “hyper-rectangle” in d dimensions, obtained by the cross product of Fj,i along each field. The set of rules R thus can be considered a set of hyper-rectangles, and a packet header represents a point in d dimensions.
Point location in computational geometry involves from a set of non-overlapping objects (hyper-rectangles) finding the enclosing object that a point belongs to. The low bounds for point location problem in N objects with d dimensions, where d > 3, are either an time complexity with space complexity; or an
time complexity with )
(log N
Ο Ο(Nd)
) ) ((log −1
Ο N d Ο(N) space complexity. However, the packet
classification problem allows objects (rules) overlapping with each other. Therefore, packet classification problem is at least as hard as point location problem. A solution of packet classification problem either requires an enormous storage space or long search time. For example, let us assume that we would like the router to be able to process 1,000 rules of 5 dimensions. An algorithm with Ο(log4N) execution time
and space requires 10,000 memory accesses per packet. This is impractical with any current technology. If we use a
) Ο(N
) (log N
Ο time and space
algorithm, the space requirement becomes prohibitively large, in the range of 1, 000G bytes. The complexity drives us to use heuristic algorithm being a practical solution for the packet classification problem. By exploiting the characteristic in rule table, heuristic algorithms may break the performance low bound achieved in the point location problem.
) (N5 Ο
A good packet classification algorithm must classify packets quickly with minimal memory storage requirements. This study proposes a novel bit compression packet classification algorithm. This algorithm succeeds in reducing the memory storage requirements in the bitmap intersection algorithm [8], proposed by Lakshman and Stiliadis. The bitmap intersection algorithm converts the packet classification problem into a multidimensional range lookup problem and constructs bit vectors for each dimension. Since the bit vectors contain lots of ‘0’ bits, the bit vectors could be compressed. We compress the bit vectors by preserving only useful information but removing the redundant bits of the bit vectors. An additional index table would be created to keep tract of the rule number associated with the remaining bits.
Additionally, the wildcarded rules also enable more extensive improvement. The bit compression algorithm reduces the storage complexity from O( ) of the bitmap intersection algorithm to O(dN·logN), where d denotes the number of dimensions and N denotes the number of rules, without sacrificing the classification performance.
dN2
The rest of the thesis is organized as follows. Chapter 2 introduces performance metrics and related works for packet classification problem. Then, the basic idea of compressed bit vector (CBV) and the details of the bit compression algorithm are purposed in Chapter 3. We display experimental platform and the performance results in Chapter 4. And finally, the conclusion and future work is given in Chapter 5.