CN1809019A - Method of implementing quick network message distribution based on adaptive cache mechanism - Google Patents
Method of implementing quick network message distribution based on adaptive cache mechanism Download PDFInfo
- Publication number
- CN1809019A CN1809019A CNA2005100111946A CN200510011194A CN1809019A CN 1809019 A CN1809019 A CN 1809019A CN A2005100111946 A CNA2005100111946 A CN A2005100111946A CN 200510011194 A CN200510011194 A CN 200510011194A CN 1809019 A CN1809019 A CN 1809019A
- Authority
- CN
- China
- Prior art keywords
- cache
- message
- cache entry
- distribution
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
This invention provides one method to realize rapid network sending method, which can realize message rapid matching to improve requirement efficiency without pre-process to support multi-dimension transferring items and gains updating. This method is based on buffer mechanism to realize rapid message and timing buffer to self adapt buffer to support transfer items gains adding and deleting. The invention method uses network communication space local relativity through buffer mask codes to control buffer particles to make the buffer cover the message to re-use distribution results.
Description
Technical field
The invention belongs to computer network and data communication technology field, be specifically related to a kind of method that realizes the quick network message distribution based on adaptive cache mechanism, this method can be used among the message dispensing device of the network equipments such as router, switch, fire compartment wall, intrusion detection, network auditing system and charging, load balancing and software, realize the quick match hit of message distribution item, the efficient of raising distribution inquiry.
Background technology
In the network equipment and software, usually need to process respectively according to the different messages of distribution set pair, distribution collect often larger, attribute is complicated and the matching operation type is various.Owing to need to distribute inquiry to each message, find its corresponding distribution item, so the efficient of message distribution guarantee to play a part aspect the entire system performance very crucial.In addition, need also to consider that message is distributed in the requirement of the aspects such as resource overhead, flexibility, extensibility and support distribution item incremental update.The method that realizes the quick network message distribution has become the important topic that academia and industrial quarters need to solve.
Summary of the invention
The purpose of this invention is to provide a kind of method that realizes the quick network message distribution, this method can realize that the Rapid matching of message distribution item hits, improves the efficient of distribution inquiry, and does not need the incremental update of preprocessing process, support multi-dimensional complicated distribution item and distribution item.
The object of the present invention is achieved like this: a kind ofly realize the quick network message distribution and regularly buffer memory is carried out the method that self adaptation is adjusted based on caching mechanism, and support the increment of distribution item to add and deletion.
Realize that based on caching mechanism the quick network message distribution is characterised in that:
At first set up a hash table, it is long to choose suitable table according to the scale of distribution collection and storage size.Each hash item comprises the head pointer of a cache entry chained list in this table.Each cache entry comprises following element: the pointer of next cache entry in buffer tag vector, a distribution array of pointers, comparison counter and the sensing chained list.Wherein the buffer tag vector by source IP address prefix, purpose IP address prefix, input equipment, protocol number, whether the fragment sign is formed;
Select then distribution to concentrate except single IP the most longways the mask of location prefix as the initial value of buffer memory mask;
Then extract the source IP address of message and purpose IP address and buffer memory mask and carry out AND operation, then with the input equipment of message, protocol number and whether information such as fragment sign form five-tuple message label vector.This message label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index;
To distribute the query counts device and add 1, judge that then the cache entry chained list of hash item indication is no for empty? if then carry out and add the cache entry operation;
If not, then the buffer tag vector of each cache entry in message label vector and the cache entry chained list is compared successively.If the buffer tag of cache entry vector is inequality with the message label vector, then the comparison counter of this cache entry is subtracted 1 and the cache invalidation counter add 1, forward next cache entry then to and proceed comparison loop.If also do not find its label vector cache entry (cache invalidation) identical with the message label vector when circulation finishes, carry out the cache invalidation operation; Otherwise, the comparison counter of cache entry is added 5, the distribution item that the pointer in the distribution item array of pointers is pointed and message are carried out matching operation successively then.If find corresponding distribution item (being cache hit), then finish this time distribution inquiry; Otherwise (being that cache hit lost efficacy) carries out cache hit inefficacy operation.
The matching operation of described distribution and message is about to the MAC Address, source IP address, purpose IP address, source port, destination interface, protocol number of message, whether the value of corresponding field compares in property values such as fragment sign, input equipment, output equipment, connection status and the distribution item.If these property values all drop in the scope of distribution item corresponding field value indication, then claim the message match hit this distribution item.
Does described cache invalidation operation promptly at first judge whether free buffer item in addition? if then carry out and add the cache entry operation; Replace the cache entry operation otherwise carry out.
Described cache hit lost efficacy and operated, and the counter that namely at first cache hit lost efficacy adds 1, jumped to then distribution and concentrated, and the distribution item corresponding with message inquired about in last pointer indication position backward from the distribution item array of pointers.In query script, if each attribute field value of message label vector all drops in the value institute how of certain each corresponding field of distribution item, then the pointer with this distribution item joins in the distribution item array of pointers of cache entry, until array is filled.
The operation of described interpolation cache entry, i.e. a newly-built cache entry and it is joined in the cache entry chained list, with the message label vector as the buffer tag vector.Jump to then to distribute to concentrate and start anew to inquire about the distribution item that is complementary with message.In query script, if each attribute field value of message label vector all drops in the value institute how of certain each corresponding field of distribution item, then the pointer with this distribution item joins in the distribution item array of pointers of cache entry, until array is filled.
Described replacement cache entry operation, choosing that cache entry that compares the Counter Value minimum in the cache entry chained list replaces, with the message label vector as the buffer tag vector, empty a distribution array of pointers simultaneously and also will compare counter O reset, jump to distribution then and concentrate start anew inquiry and the corresponding distribution item of message.In query script, if each attribute field value of message label vector all drops in the value institute how of certain each corresponding field of distribution item, then the pointer with this distribution item joins in the distribution item array of pointers of cache entry, until array is filled.
Regularly buffer memory being carried out the self adaptation adjustment is characterised in that:
At first calculate cache invalidation rate and cache hit crash rate according to distribution query counts device, cache invalidation counter, cache hit inefficacy counter;
Do you judge then whether cache invalidation rate or cache hit crash rate surpass threshold value? if the cache invalidation rate surpasses threshold value, show that then the buffer memory mask is too big, need to turn it down; If the cache hit crash rate surpasses threshold value, show that then the buffer memory mask is too little, need to transfer large.After adjusting the buffer memory mask, empty hash table, again add cache entry.
Supporting the distribution item increment to add is characterised in that:
At first extract the source IP address section that defines in the new distribution item of adding and purpose IP address field and buffer memory mask and carry out AND operation, and the input equipment that in distribution item, defines, protocol number and whether the information such as fragment sign form five-tuple label vector.This label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index;
Then each cache entry in this distribution item and the hash item indication cache entry chained list is compared.If each field value of the label of distribution item vector all drops in the value institute how of each corresponding field of buffer tag vector of certain cache entry, then scan from back to front the distribution item array of pointers of cache entry, the pointer of this distribution item is inserted in the array.
Support the deletion of distribution item increment to be characterised in that:
Distribution item to be deleted is done upper delete flag, and needn't reclaim the space;
Skipping these distribution items of doing mark when carrying out later on the distribution inquiry gets final product.
The process that this method is carried out the network message distribution is:
(1) carry out initialization operation: hash table of model, according to scale and the memory space of distribution collection
It is long that size is chosen suitable table, can be big according to 1/3rd or memory space of distribution item sum
It is long that little divided by the cache entry size 6 times are used as table.Each hash item in this table comprises the head pointer of a cache entry chained list.(can not surpass the cache entry sum) do not limit in the dynamically additions and deletions of cache entry in the chained list, number.Each cache entry comprises following element: buffer tag vector, distribution array of pointers, comparison counter and point to the pointer of next cache entry.Wherein the label vector by source IP address prefix, purpose IP address prefix, input equipment, protocol number, whether the fragment sign is formed.Select then distribution to concentrate except single IP the most longways the mask of location prefix as the initial value of buffer memory mask;
(2) extract the source IP address of message and purpose IP address and buffer memory mask and carry out AND operation, then with the input equipment of message, protocol number and whether information such as fragment sign form five-tuple message label vector.This message label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index;
(3) will distribute the query counts device and add 1, judge that then hash item indication cache entry chained list is empty? if then enter (10) operating procedure; If not, then enter (4) operating procedure;
(4) the buffer tag vector with each cache entry in message label vector and the cache entry chained list compares successively.If both are inequality, then the comparison counter of cache entry is subtracted 1 and the cache invalidation counter add 1, forward next cache entry then to and proceed comparison loop;
(5) if also do not find its label vector and the vectorial identical cache entry (cache invalidation) of message label during loop ends, then enter (8) operating procedure; Otherwise, enter (6) operating procedure;
(6) the comparison counter with cache entry adds 5, the distribution item that pointer in the distribution item array of pointers is pointed and message are carried out matching operation successively then, if be about to the MAC Address, source IP address, purpose IP address, source port, destination interface, protocol number of message, whether the value of the property value such as fragment sign, input equipment, output equipment, connection status and each corresponding field of distribution item compares---these property values all drop in the scope of distribution item corresponding field value indication, then claim the message match hit this distribution item;
(7) if find corresponding distribution (being cache hit), then enter (13) operating procedure; Otherwise (being that cache hit lost efficacy) enters (9) operating procedure;
(8) carry out the cache invalidation operation.Judge whether free buffer item in addition? if then change (10) operating procedure over to; Otherwise change (11) operating procedure over to;
(9) carry out cache hit inefficacy operation.The cache hit counter that lost efficacy is added 1, jump to then distribution and concentrate, the distribution item corresponding with message inquired about in last pointer indication position backward from the distribution item array of pointers.Change (12) operating procedure then over to;
(10) carry out the operation of interpolation cache entry.A newly-built cache entry also joins it in cache entry chained list, with the message label vector as the buffer tag vector.Jump to then to distribute to concentrate and start anew to inquire about the distribution item that is complementary with message.Change (12) operating procedure then over to;
(11) carry out the operation of replacement cache entry.Choosing that cache entry that compares the Counter Value minimum in the cache entry chained list replaces, with the message label vector as the buffer tag vector, empty a distribution array of pointers simultaneously and will compare counter O reset, jump to then to distribute to concentrate and start anew to inquire about the distribution item that is complementary with message.Change (12) operating procedure then over to;
(12) in the process of inquiry distribution collection, if each attribute field value of message label vector all drops in the value institute how of certain each corresponding field of distribution item, then the pointer with this distribution item joins in the distribution item array of pointers of cache entry, until array is filled;
(13) finish this time distribution inquiry.
The process that this method is carried out the adjustment of buffer memory self adaptation comprises the steps:
(1) regularly (time interval can be 20 seconds) calculates cache invalidation rate and cache hit crash rate according to distribution query counts device, cache invalidation counter, cache hit inefficacy counter;
(2) do you judge whether cache invalidation rate or cache hit crash rate surpass threshold value? if the cache invalidation rate surpasses threshold value, show that then the buffer memory mask is too big, need to turn it down; If the cache hit crash rate surpasses threshold value, show that then the buffer memory mask is too little, need to transfer large;
(3) behind the adjustment buffer memory mask, empty hash table, add cache entry again;
(4) finish this self adaptation adjustment.
This method is distributed a process of increment interpolation and is comprised the steps:
(1) extract the source IP address section that defines in the new distribution item of adding and purpose IP address field and buffer memory mask and carry out AND operation, and the input equipment that in distribution item, defines, protocol number and whether the information such as fragment sign form five-tuple label vector.This label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index;
(2) then each cache entry in this distribution item and the hash item indication cache entry chained list is compared.If each field value of the label of distribution item vector all drops in the value institute how of each corresponding field of buffer tag vector of certain cache entry, then scan from back to front the distribution item array of pointers of cache entry, the pointer of this distribution item is inserted in the array.
This method supports the process of distribution item increment deletion to comprise the steps:
(1) distribution item to be deleted is done upper delete flag, and needn't reclaim the space;
Skipping these distribution items of doing mark when carrying out the distribution inquiry (2) gets final product.
The characteristics of implementation method of the present invention are to have utilized the space local correlations of network service, and control cache granularity by the buffer memory mask, so that drop on the distribution Query Result that message in the spatial dimension that cache entry covers is preserved in can multiplexing cache entry, thereby improved utilization rate and the hit rate of buffer memory, spatial cache also is controlled at certain scale; According to the cache invalidation rate with hit crash rate the buffer memory mask is dynamically adjusted, when reducing, buffer efficiency can in time make a response.This method does not need preprocessing process, and applicability is good, is easy to realize that it is few not only to take memory space, and can significantly improves the efficient that distribution is searched.
Description of drawings
Below in conjunction with accompanying drawing the present invention is illustrated in further detail:
Fig. 1 is the operating procedure flow chart of realization network message distribution of the present invention;
Fig. 2 is the operating procedure flow chart that realization buffer memory self adaptation of the present invention is adjusted;
Fig. 3 is the operating procedure flow chart that realization distribution item increment of the present invention adds;
Fig. 4 is the operating procedure flow chart of realization distribution item increment deletion of the present invention;
Most preferred embodiment is described in detail
Below with reference to accompanying drawing of the present invention, most preferred embodiment of the present invention is described in more detail.
The present invention a kind ofly realizes the quick network message distribution and regularly buffer memory is carried out the method that self adaptation is adjusted based on caching mechanism, and supports the increment of distribution item to add and deletion.
Referring to Fig. 1, the process that this method is carried out the network message distribution comprises the steps:
(1) carries out initialization operation 1: hash table of model, scale and storage size according to distribution collection are chosen suitable table long (being the hash table entry sum), and it is long to be used as table divided by 6 times of cache entry size according to 1/3rd or storage size of distribution item sum.Each hash item in this table all comprises the head pointer of a cache entry chained list, takies 4 bytes.(can not surpass the cache entry sum) do not limit in the dynamically additions and deletions of cache entry in the chained list, number.Select then distribution to concentrate except single IP the most longways the mask of location prefix as the initial value of buffer memory mask.
The structure of hash table is as shown in the table:
Data message (arranging according to storage order) | Length (byte) | Implication |
The buffer memory mask | 4 | |
Distribution query counts device (query_times) | 4 | Be initialized as 0, the total degree of record distribution inquiry |
Cache invalidation counter (miss_times) | 4 | Be initialized as 0, record buffer memory inefficacy total degree |
Cache hit inefficacy counter (hitmiss_times) | 4 | Be initialized as 0, record buffer memory is hit the inefficacy total degree |
Hash table table long (hashtable_len) | 4 | The total number of hash item |
Hash table entry pointer (hashitems) | 4 | Point to the memory space at all hash table entry places, this memory space obtains according to the long dynamic assignment of hash table table |
The structure of hash item is as shown in the table:
Data message (arranging according to storage order) | Length (byte) | Implication |
Cache entry linked list head pointer (cachelist) | 4 | Point to the chained list of the cache entry composition of corresponding same Hash value |
The structure of cache entry is as shown in the table:
Data message (arranging) according to storage order | Length (byte) | Implication |
Source IP address prefix (sip) | 4 | Element in the buffer tag vector |
Source IP address prefix mask (sip_mask) | 4 ? | Element in the buffer tag vector |
Purpose IP address prefix (dip) | 4 | Element in the buffer tag vector |
Source IP address prefix mask (dip_mask) | 4 ? | Element in the buffer tag vector |
Input equipment (indev) | 4 | Element in the buffer tag vector |
Protocol number (protocol) | 1 | Element in the buffer tag vector |
Whether fragment identifies (isfrag) | 1 | Element in the buffer tag vector |
Compare counter (meet_counter) | 2 | Be initialized as 0, if overflow then return 0 |
A distribution array of pointers (items) | 32 ? ? ? ? | Preserve the pointer of the distribution item of corresponding same cache entry, it is fixed that the order of distribution item in array comes by the precedence of concentrating in distribution; Preserve 8 pointers altogether, each pointer accounts for 4 bytes |
Point to next cache entry pointer (next) | 4 ? |
(2) hash table operation 2 is searched in execution.Extract the source IP address of message and purpose IP address and buffer memory mask and carry out AND operation, then with the input equipment of message, protocol number and whether information such as fragment sign form five-tuple message label vector.This message label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index.The operational formula of hash function is:
(sip+dip+indev+protocol+isfrag)%hashtable_len
(3) will distribute the query counts device and add 1, judge that then hash item indication cache entry chained list is empty? if then enter (10) operating procedure; If not, then enter (4) operating procedure;
(4) relatively cache entry operation 3.The buffer tag vector of each cache entry in message label vector and the cache entry chained list is compared successively.If both are inequality, then the comparison counter of cache entry is subtracted 1 and the cache invalidation counter add 1, forward next cache entry then to and proceed comparison loop;
(5) if also do not find its label vector and the vectorial identical cache entry (cache invalidation) of message label during loop ends, then enter (8) operating procedure; Otherwise, enter (6) operating procedure;
(6) carry out query caching item distribution item array of pointers operation 4.The comparison counter of cache entry is added 5, pointer distribution item and the message pointed that to distribute then in the array of pointers carried out matching operation successively, if be about to the MAC Address, source IP address, purpose IP address, source port, destination interface, protocol number of message, whether property values such as fragment sign, input equipment, output equipment, connection status compare with the value of distribution each corresponding field---these property values all drop in the scope of a distribution corresponding field value indication, claim that then the message match hit should distribution item;
(7) if find the distribution item (being cache hit) that is complementary, then enter (13) operating procedure; Otherwise (being that cache hit lost efficacy) enters (9) operating procedure;
(8) carry out cache invalidation operation 5.Judge whether free buffer item in addition? if then change (10) operating procedure over to; Otherwise change (11) operating procedure over to;
(9) carry out cache hit inefficacy operation 6.The cache hit inefficacy counter of cache entry is added 1, jump to then distribution and concentrate, the distribution item that is complementary with message is inquired about in last pointer indication position backward from the distribution item array of pointers.Change (12) operating procedure then over to;
(10) carry out interpolation cache entry operation 7.A newly-built cache entry also joins it in cache entry chained list, with the message label vector as the buffer tag vector.Jump to then to distribute to concentrate and start anew to inquire about the distribution item that is complementary with message.Change (12) operating procedure then over to;
(11) carry out replacement cache entry operation 8.Choosing that cache entry that compares the Counter Value minimum in the cache entry chained list replaces, put its label vector and be the message vector, empty simultaneously the distribution item array of pointers and will compare counter O reset, jump to then to distribute to concentrate and start anew to inquire about the distribution item that is complementary with message.Change (12) operating procedure then over to;
(12) executive logging query path operation 9.In the process of inquiry distribution collection, if each attribute field value of message label vector all drops in the value institute how of certain each corresponding field of distribution item, then the pointer with this distribution item joins in the distribution item array of pointers of cache entry, until array is filled;
(13) finish this time distribution inquiry.
Referring to Fig. 2, the process that this method is carried out the adjustment of buffer memory self adaptation comprises the steps:
(1) regularly (time interval can be 20 seconds) calculates cache invalidation rate and cache hit crash rate according to distribution query counts device, cache invalidation counter, cache hit inefficacy counter.The cache invalidation rate
Computing formula is:
miss_times/query_times
The computing formula of cache hit failure rate is:
hitmiss_times/query_times
(2) do you judge whether cache invalidation rate or cache hit failure rate surpass threshold value? if the cache invalidation rate surpasses threshold value, show that then the buffer memory mask is too big, it need be turned down (moving to left 1); If the cache hit failure rate surpasses threshold value, show that then the buffer memory mask is too little, it need be transferred big (tape symbol moves to right 1);
(3) behind the adjustment buffer memory mask, empty hash table, add cache entry again;
(4) finish this self adaptation adjustment.
Referring to Fig. 3, this method is distributed a process of increment interpolation and is comprised the steps:
(1) extract the source IP address section that defines in the new distribution item of adding and purpose IP address field and buffer memory mask and carry out AND operation, and the input equipment that in distribution item, defines, protocol number and whether the information such as fragment sign form five-tuple label vector.This label vector as key assignments substitution hash function, is found corresponding hash item in hash table take operation result as index;
(2) then each cache entry in this distribution item and the hash item indication cache entry chained list is compared.If each field value of the label of distribution item vector all drops in the value institute how of each corresponding field of buffer tag vector of certain cache entry, then scan from back to front the distribution item array of pointers of cache entry, the pointer of this distribution item is inserted in the array.
Referring to Fig. 4, the process of this method support distribution increment deletion comprises the steps:
(1) distribution item to be deleted is done upper delete flag, and needn't reclaim the space;
Skipping these distribution items of doing mark when carrying out the distribution inquiry (2) gets final product.
So, implementation method of the present invention is to utilize the space local correlations of network service, and control the buffer memory granularity by the buffer memory mask, make and to drop on the distribution Query Result that message in the spatial dimension that cache entry covers is preserved in can multiplexing cache entry, thereby improved the utilance and the hit rate of buffer memory, spatial cache also is controlled at certain scale; According to the cache invalidation rate with hit failure rate the buffer memory mask is dynamically adjusted (adaptation mechanism that is buffer memory), when reducing, buffer efficiency can in time make a response, distribute the efficient of searching thereby improved whole system.In addition, this method does not need preprocessing process, and applicability is good, be easy to realize, and it is few to take memory space.
This method is used at the fire wall of applicant's development, has obtained good effect, has realized the purpose of invention.Method of the present invention has good practicability, has good popularization and application prospect.
Although disclose for the purpose of illustration most preferred embodiment of the present invention and accompanying drawing, it will be appreciated by those skilled in the art that: without departing from the spirit and scope of the invention and the appended claims, various replacements, variation and modification all are possible.Therefore, the present invention should not be limited to most preferred embodiment and the disclosed content of accompanying drawing.
Claims (10)
1. realize specifically may further comprise the steps the method for quick network message distribution based on adaptive cache mechanism for one kind:
At first set up a hash table, choose table length according to the scale and the storage size of distribution collection;
Select then distribution to concentrate except single IP the most longways the mask of location prefix as the initial value of buffer memory mask;
Extract the source IP address and the purpose IP address of message and carry out AND operation, then with input equipment, the protocol number of message with whether fragment identifies forms the message label vector with the buffer memory mask;
To distribute the query counts device and add 1, judge that then the cache entry chained list of hash item indication is no for empty; If then carry out and add the cache entry operation; If not, then the buffer tag vector of each cache entry in message label vector and the cache entry chained list is compared successively.
2. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 1, it is characterized in that: according to the comparative result of the buffer tag vector of each cache entry in message label vector and the cache entry chained list, if the buffer tag of cache entry vector is not identical with the message label vector, then the comparison counter of this cache entry is subtracted 1 and the cache invalidation counter add 1, forward then next cache entry to and proceed comparison loop.
3. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 2, it is characterized in that: if also do not find its label vector and the vectorial identical cache entry of message label during loop ends, be cache invalidation, carry out the cache invalidation operation; Otherwise, the comparison counter of cache entry is added 5, rule and message then that the pointer in the regular array of pointers is pointed are carried out matching operation successively.
4. according to claim 3ly realize the method for quick network message distribution based on adaptive cache mechanism, it is characterized in that: if find corresponding rule, namely cache hit then finishes this time distribution inquiry; Otherwise namely cache hit lost efficacy, and carried out cache hit inefficacy operation.
5. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 4, it is characterized in that: the matching operation of described rule and message, be about to the MAC Address, source IP address, purpose IP address, source port, destination interface, protocol number of message, the value of fragment sign, input equipment, output equipment, these property values of connection status and regular each corresponding field of attribute vector compares, if these property values all drop in the scope of regular corresponding attribute field value indication, claim that then the message match hit should rule.
6. the method based on the distribution of adaptive cache mechanism realization quick network message according to claim 4 is characterized in that: described cache invalidation operation promptly at first judges whether free buffer item in addition; If then carry out and add the cache entry operation; Replace the cache entry operation otherwise carry out.
7. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 4, it is characterized in that: described cache hit lost efficacy and operated, the counter that namely at first cache hit lost efficacy adds 1, jump to then distribution and concentrate, the rule corresponding with message inquired about in last pointer indication position backward from regular array of pointers; In query script, if each attribute field value of message label vector all drops in the value institute how of certain regular each corresponding field of attribute vector, pointer that then should rule joins in the regular array of pointers of cache entry, until array is filled.
8. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 4, it is characterized in that: described interpolation cache entry operation, be a newly-built cache entry and it joined in the cache entry chained list, with the message label vector as the buffer tag vector; Jump to then to distribute to concentrate and start anew to inquire about the rule that is complementary with message; In query script, if each attribute field value of message label vector all drops in the value institute how of certain regular each corresponding field of attribute vector, pointer that then should rule joins in the regular array of pointers of cache entry, until array is filled.
9. the method that realizes the quick network message distribution based on adaptive cache mechanism according to claim 4, it is characterized in that: described replacement cache entry operation, choosing that cache entry that compares the Counter Value minimum in the cache entry chained list replaces, with the message label vector as the buffer tag vector, empty simultaneously regular array of pointers and will compare counter O reset, jump to then to distribute to concentrate and start anew to inquire about the rule corresponding with message; In query script, if each attribute field value of message label vector all drops in the value institute how of certain regular each corresponding field of attribute vector, pointer that then should rule joins in the regular array of pointers of cache entry, until array is filled.
10. the method based on the distribution of adaptive cache mechanism realization quick network message according to claim 4 is regularly carried out the self adaptation adjustment to buffer memory and is characterised in that: at first calculates cache invalidation rate and cache hit crash rate according to distribution query counts device, cache invalidation counter, cache hit inefficacy counter; Judge then whether cache invalidation rate or cache hit crash rate surpass threshold value; If the cache invalidation rate surpasses threshold value, show that then the buffer memory mask is too big, need to turn it down; If the cache hit crash rate surpasses threshold value, show that then the buffer memory mask is too little, need to transfer large, behind the adjustment buffer memory mask, empty hash table, again add cache entry.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2005100111946A CN100444579C (en) | 2005-01-18 | 2005-01-18 | Method of implementing quick network message distribution based on adaptive cache mechanism |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNB2005100111946A CN100444579C (en) | 2005-01-18 | 2005-01-18 | Method of implementing quick network message distribution based on adaptive cache mechanism |
Publications (2)
Publication Number | Publication Date |
---|---|
CN1809019A true CN1809019A (en) | 2006-07-26 |
CN100444579C CN100444579C (en) | 2008-12-17 |
Family
ID=36840717
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNB2005100111946A Expired - Fee Related CN100444579C (en) | 2005-01-18 | 2005-01-18 | Method of implementing quick network message distribution based on adaptive cache mechanism |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN100444579C (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101059958B (en) * | 2007-03-15 | 2010-12-15 | 上海交通大学 | Coding and decoding method for speedup the quantitative data with the buffer memory |
CN102970150A (en) * | 2011-09-01 | 2013-03-13 | 日电(中国)有限公司 | Extensible multicast forwarding method and device for data center (DC) |
CN104468157A (en) * | 2013-09-16 | 2015-03-25 | 上海斐讯数据通信技术有限公司 | Distribution method |
WO2019010703A1 (en) * | 2017-07-14 | 2019-01-17 | 华为技术有限公司 | Method for reading and partially writing data, and relevant apparatus |
CN110149248A (en) * | 2019-06-06 | 2019-08-20 | 杭州商湾网络科技有限公司 | A kind of method of express statistic analysis router traffic |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1129281C (en) * | 2001-10-17 | 2003-11-26 | 华为技术有限公司 | Method of raising IP message forwarding speed utilizing slow storage technology |
US7325059B2 (en) * | 2003-05-15 | 2008-01-29 | Cisco Technology, Inc. | Bounded index extensible hash-based IPv6 address lookup method |
-
2005
- 2005-01-18 CN CNB2005100111946A patent/CN100444579C/en not_active Expired - Fee Related
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101059958B (en) * | 2007-03-15 | 2010-12-15 | 上海交通大学 | Coding and decoding method for speedup the quantitative data with the buffer memory |
CN102970150A (en) * | 2011-09-01 | 2013-03-13 | 日电(中国)有限公司 | Extensible multicast forwarding method and device for data center (DC) |
CN104468157A (en) * | 2013-09-16 | 2015-03-25 | 上海斐讯数据通信技术有限公司 | Distribution method |
CN104468157B (en) * | 2013-09-16 | 2020-01-17 | 上海斐讯数据通信技术有限公司 | Distribution method |
WO2019010703A1 (en) * | 2017-07-14 | 2019-01-17 | 华为技术有限公司 | Method for reading and partially writing data, and relevant apparatus |
CN110149248A (en) * | 2019-06-06 | 2019-08-20 | 杭州商湾网络科技有限公司 | A kind of method of express statistic analysis router traffic |
Also Published As
Publication number | Publication date |
---|---|
CN100444579C (en) | 2008-12-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1863169A (en) | Route searching result cache method based on network processor | |
CN1940922A (en) | Method and system for improving information search speed | |
CN1809019A (en) | Method of implementing quick network message distribution based on adaptive cache mechanism | |
CN1863142A (en) | Method for providing different service quality tactics to data stream | |
CN101035062A (en) | Rule update method for three-folded content addressable memory message classification | |
CN1664790A (en) | Method and apparatus for increasing data storage capacity | |
CN1929447A (en) | Method and device for searching address prefixion and message transfer method and system | |
CN1404591A (en) | Apparatus and method for performing high-speed IP route lookup and managing routing/forwarding tables | |
CN1878139A (en) | Three-layer forwarding method, device and ARP information table updating method | |
CN100342377C (en) | Method of raising officiency of data processing | |
WO2020160017A1 (en) | Efficient and flexible load-balancing for clusters of caches under latency constraint | |
CN101043695A (en) | Method for storing and maintaining user equipment information in mobile communication system | |
CN101079042A (en) | System and method for quickly inquiring about black and white name list | |
CN1744563A (en) | Method for realizing strate gic route in Ethernet switch | |
CN1852261A (en) | Method for maintenance of neighbor entry | |
CN1968212A (en) | Hardware-based differentiated organization method in stream classification | |
CN101047649A (en) | Method and equipment for transmitting data flow | |
CN1815997A (en) | Group classifying method based on regular collection division for use in internet | |
CN103973571A (en) | Network processor and routing searching method | |
CN1208725C (en) | Method and device for high-speed buffering of buffer | |
CN1859208A (en) | Method and system for managing TCAM route list | |
CN1719769A (en) | Method for classifying received data pocket in network apparatus | |
CN101030897A (en) | Method and apparatus for matching mode in invading detection | |
CN1207878C (en) | Routing table compression method | |
CN1845531A (en) | Data forwarding controlling method and apparatus |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20081217 Termination date: 20140118 |