CN102055670B - Method for improving message classification rule capacity of hardware by coordination between software and hardware - Google Patents

Abstract

The invention provides a method for improving message classification rule capacity of hardware by the coordination between software and hardware. In the hardware, a large-capacity SRAM (static random access memory with a capacity reaching several hundred MBs) is adopted to store message classification rules in a hash list format; and by establishing an expanded copy of the hash list of the hardware in the software and allowing the software to be responsible for constructing and maintaining the message classification rules, a hardware message classification system can configure a large number of rules. The use of the SRAM and the hash list makes the management of the rules in the SRAM quick and convenient and ensures the matching efficiency of the rules in the hardware; meanwhile, the capacity of the SRAM storage medium is high, so more rules can be stored.

Description

A kind of method that improves hardware message classification rule capacity by software-hardware synergism

Technical field

The present invention relates to network data processing field, be specifically related to a kind of method that improves hardware message classification rule capacity by software-hardware synergism.

Background technology

In general hardware message classification system, message classification rule is generally used CAM (access to content memory) to preserve, and is convenient to hardware and carries out high speed coupling.But in some network flow monitoring system, often need jumbo accurate classifying rules, the capacity of CAM generally only has several MB, is difficult to meet the demands.

In network environment, different pieces of information bag adheres to separately different classes of according to some principles, and these principle of classification are exactly message classification rule.In traditional hardware message classification system, generally adopt CAM (access to content memory) to preserve these rules, so that hardware carries out high speed coupling.Hardware, by interface routine, is searched the rule in CAM, and is mated, and according to matching result, message classification, for the message of classifying, carries out feature processing.

In traditional hardware message classification system, process chip is connected with CAM (access to content memory), and this CAM has the advantages such as speed is fast, flexible, but capacity is generally smaller, only has several MB.But along with the requirement of system, rule capacity constantly increases, need the rule of storage usually to reach hundreds of MB, so just need jumbo storage medium to store these jumbo rules.Process chip is directly connected with CAM (access to content memory), by interface function, obtains CAM access handle, directly rule is write in CAM space.Because CAM capacity is little, general common algorithm gets final product simple-to-maintain.For large capacity, use CAM to solve.

Development along with application demand, there is the shortcoming that capacity is inadequate in prior art scheme, because a lot of systems need jumbo accurate classifying rules, should realize efficient rule match, meet again its capacity requirement, often need jumbo storage medium to meet the requirement of storing, in existing technical scheme, intelligence is satisfied with efficient coupling, and cannot solve the problem that needs jumbo accurate rule classification.

Summary of the invention

The object of the invention is to solve the shortcoming of above-mentioned prior art scheme, in hardware, adopt jumbo SRAM to preserve message classification rule, form storage with hash table, by set up the copy of the expansion of hardware hash table in software, by software, be responsible for structure and maintenance packet classifying rules, make hardware message classification system can configure jumbo rule.

A method that improves hardware message classification rule capacity by software-hardware synergism, is characterized in that: comprise the following steps:

A, at region of memory of main frame application, wherein a part is used for storage rule, another part is used for storing hash table;

B, by interface library, obtain hardware address, by mapping function, obtain SRAM buffer pointer, and initialization SRAM, be then each rule generation hash keyword, and according to hash keyword, rule added in main frame rule list and SRAM;

C, increase or during deletion rule, interface library, by hash function, is searched the position at the hash keyword place of this rule correspondence, searches and whether has conflict, if there is not conflict, directly this rule is put in rule list and rule conflict chain when needs; If there is conflict, this rule is dosed to a position of conflict chain place chained list;

D, interface library upgrade the storage rule in host memory, after storage rule upgrades in host memory, then according to the rule list after adjusting, upgrade the rule list of hardware end in SRAM;

E, when packets need matched rule, by hash function, search the rule in SRAM, according to hash keyword, at hash key position and collision position matched rule.

The present invention, according to regular characteristic, owing to having used SRAM and hash table, not only can fast, conveniently manage the rule in SRAM, has guaranteed the rule matching efficiency on hardware, simultaneously because the capacity of SRAM storage medium is larger, can store more rule.

Accompanying drawing explanation

Fig. 1 is work schematic diagram of the present invention

Specific embodiments

The present invention shows to search feature fast by hash, sets up a regular hash table in Installed System Memory, shows to realize the efficient coupling of hardware message classification by hash.

Concrete scheme is as follows:

(1) in host memory, apply for a region of memory, wherein in this piece region of memory, a part of storage rule, part storage hash table.

(2) by interface library, obtain hardware address, by mapping function, obtain SRAM buffer pointer, and initialization SRAM, be then that each rule generates hash number, and according to hash number, rule added in main frame rule list and SRAM.

(3) when needs increase, deletion rule, interface function is by hash function, and the hash that searches this rule correspondence counts the position at place, searches and whether has conflict, and if there is no conflict, directly puts this rule in rule list and rule conflict chain into; If there is conflict, this rule is added to a position of conflict chain place chained list.

(4) in order to prevent, in the process of operation rules, affect the message classification in SRAM, first interface library upgrades the storage rule of host side, after having upgraded host side rule, then according to the rule list after adjusting, upgrades the rule list of hardware end in SRAM.

(5) when packets need matched rule, by hash function, search the rule in SRAM, according to hash number, in hash numerical digit, put and collision position matched rule.

In the network security private server that the present invention has produced at dawn, use, prove that can realize SRAM stores large capacity rule, has improved the regular quantity of system.

Claims (1)

1. by software-hardware synergism, improve a method for hardware message classification rule capacity, it is characterized in that: comprise the following steps:

A, at region of memory of main frame application, wherein a part is used for storage rule, another part is used for storing hash table;

B, by interface library, obtain hardware address, by mapping function, obtain SRAM buffer pointer, and initialization SRAM, be then each rule generation hash keyword, and according to hash keyword, rule added in main frame rule list and SRAM;

C, when needs increase, interface library by hash function, is searched the position at the hash keyword place of this rule correspondence, searches and whether has conflict, if there is not conflict, directly this rule is put in rule list; If there is conflict, a position of this rule being added to the conflict chain of this rule;

D, interface library upgrade the storage rule in host memory, after storage rule upgrades in host memory, then according to the rule list after adjusting, upgrade the rule list of hardware end in SRAM;

E, when message needs matched rule, by hash function, search the rule in SRAM, according to hash keyword, at hash key position and collision position matched rule.