CN101419562B - Hardware PRI queue implementing method for balancing load and performance - Google Patents
Hardware PRI queue implementing method for balancing load and performance Download PDFInfo
- Publication number
- CN101419562B CN101419562B CN2008101629034A CN200810162903A CN101419562B CN 101419562 B CN101419562 B CN 101419562B CN 2008101629034 A CN2008101629034 A CN 2008101629034A CN 200810162903 A CN200810162903 A CN 200810162903A CN 101419562 B CN101419562 B CN 101419562B
- Authority
- CN
- China
- Prior art keywords
- priority
- entity
- directory entry
- processing unit
- queue
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000012545 processing Methods 0.000 claims abstract description 58
- 238000003780 insertion Methods 0.000 claims abstract description 17
- 230000037431 insertion Effects 0.000 claims abstract description 17
- 230000015572 biosynthetic process Effects 0.000 claims description 37
- 238000005755 formation reaction Methods 0.000 claims description 37
- 230000000052 comparative effect Effects 0.000 claims description 24
- 230000017105 transposition Effects 0.000 claims description 9
- 238000013519 translation Methods 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 7
- 241000288673 Chiroptera Species 0.000 claims description 3
- 238000012966 insertion method Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 11
- 238000012546 transfer Methods 0.000 abstract 2
- 238000010586 diagram Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an implementation method of a hardware priority queue capable of balancing load and performance. The method combines the present comparatively prevalent implementation methods of two hardware priority queues which are a parallel hardware priority queue and a transfer hardware priority queue respectively, solves the conflict between the load and the performance, gives full play to performance advantages of hardware parallelization and reduces load increase caused by the parallelization. By a modular design, the method causes the priority queue to have better extensibility, which ensures relative independence whether the priority or directory items are needed to be added without changing the structure of the whole queue. The method improves the parallelism and response speed of the priority queue by a parallel transpose queue of a lower layer processing unit, and causes the insertion of an entity to need only to be compared with all directory items in an upper queue of a first priority by an item-by-item transfer queue of the upper queue, which reduces the load.
Description
Technical field
The present invention relates to hardware priority cohort design field, relate in particular to the hardware priority formation implementation method of a kind of balanced load and performance.
Background technology
The world today, the application of computer technology has been pushed human civilization to a new height, not only in the scientific and engineering field, in people's the life, also has to have incorporated increasing computer equipment.And in a lot of embedded devices and the network equipment, it is very extensive that priority query uses.Because the speed advantage of hardware realizes priority query with hardware, becomes the selection of the quick queue operation of many requirements system.Many embedded real time systems, for example speech player etc. needs entity queue response fast.Many low delays that require, high-quality network system, for example the real-time voice exchange system needs the transmitting-receiving of network packet formation fast to guarantee.In order to require high in real time to these, require rapidly simple entity of scheduling or service, priority scheduling is one of its best selection scheme, and directly use hard-wired priority query, be undoubtedly one of approach of such system performance raising, how to design a rational and effective hardware configuration, realize priority query, become one of problem of studying on the engineering field.
The many embedded devices and the network equipment have been arranged in the market, had the hardware priority formation, its kind roughly comprises: fifo queue walks abreast and transplants formation and transmit formation item by item.
The realization of fifo queue is fairly simple, it is according to different priority, distribute directory entry, the entity of equal priority, to be placed in the same directory entry, the expression priority height that the directory entry position is forward is during the entity dequeue, the forward directory entry from the position is according to the tactful dequeue of first in first out.The advantage of this kind design is a simplicity of design, but shortcoming is because the number and the directory entry internal storage entity information number of directory entry are certain, causes extendability poor, is unfavorable for the expansion of follow-up function.
Parallel transposition formation has made full use of the strong characteristics of hardware parallel ability, and by hardware cell independently, relatively more all simultaneously hardware cells and the priority ratio of novel entities carry out the translation realization priority query of certain rule.The advantage of the hardware priority formation of this design is that speed is fast, hardware concurrency and extensibility are good, but shortcoming is also fairly obvious, owing to need simultaneously novel entities information to be delivered on all hardware modules, from the VLSI angle analysis, to cause the big shortcoming of hardware load, further cause the expansion of formation and the speed of operation to be restricted.
Transmit priority query item by item, use the method compare item by item, the priority of novel entities only compares with first of formation, transmits item by item successively, up to finding correct position.This formation has well solved the excessive shortcoming of parallel transposition formation load, but owing to need compare item by item, has caused the formation performance decrease, simultaneously, compares item by item, need insert temporary module in each hardware cell, has increased the expense of hardware resource.
The implementation of above-mentioned three kinds of hardware priority formations respectively has relative merits, how to design a kind of hardware queue, can either bring into play the parallel and speed advantage of hardware, improve performance, can reduce the hardware load again, save hardware resource, become the focus of design hardware priority formation.
Summary of the invention
The object of the present invention is to provide the hardware priority formation implementation method of a kind of balanced load and performance.
The technical scheme that technical solution problem of the present invention is adopted is:
1. the hardware priority formation implementation method of balanced load and performance is characterized in that:
1) lower floor's processing unit is made up of some directory entries:
Each lower floor's processing unit, form by some identical directory entries, each directory entry is a hardware cell, comprise access hole, control signal and entity information I/O port are between the adjacency, by the data path identical with data bit width, and the compare result signal connection of a bit wide, each has write down entity information, comprises entity priority and other information;
2) lower floor's processing unit novel entities inserts, and according to parallel transposition lining up mode, adopts parallel mode relatively to realize:
When new entity arrives, the entity information of this entity will be sent to the entity input port of each directory entry simultaneously, simultaneously, control signal informs that each directory entry has novel entities to arrive, when directory entry finds to have novel entities to arrive, the entity priority of more own current record immediately and the priority of novel entities, if the priority of novel entities is higher than the entity of current record, and the comparative result of upper level directory entry output is higher than the current entity of its record for novel entities priority, then this directory entry reads in the current entity information recorded content of upper level directory entry, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage catalogue; If the priority of novel entities is higher than the current record entity, and the comparative result of upper level directory entry output is lower than the current entity of its record for novel entities priority, then this directory entry reads in novel entities information, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage directory entry; If novel entities priority is lower than the current record entity, then keep original entity information constant, simultaneously self comparative result is outputed on the compare result signal, as the input of next stage directory entry, once insert operation, the periodicity that needs altogether is 3 bats, first count input novel entities, and second count produces comparative result, triple time, according to comparative result, whole priority query carries out the translation of above-mentioned rule, to reach the series arrangement of priority;
3) go up layer queue and realize that by transmitting models of priority layer queue on each comprises lower floor's processing unit, between the last layer queue, according to priority, series arrangement:
Last layer queue is realized according to transmitting models of priority, between each level, connects by data path, control signal, and series arrangement according to the priority, its inside is lower floor's processing unit;
4) go up entity transmission between the layer queue, when the directory entry that occurs in previous stage upper strata inner queue lower floor processing unit exhausts:
When novel entities need insert, if in lower floor's processing unit of novel entities insertion position, priority query's directory entry less than, then insert according to lower floor's processing unit novel entities insertion method of above mentioning, need not anyly move between the last layer queue, in lower floor's processing unit of novel entities insertion position, priority query's directory entry is full, then need transmit entity information between last layer queue, at this moment, novel entities is inserted the position that needs insertion, simultaneously, will overflow the directory entry of the lowest priority of formation, move on the next stage in the layer queue, the entity information of the entity that transmits between the last layer queue and control signal are exported by the current layer queue of going up, layer queue is received signal on the next stage, and its inner lower floor processing unit begins the insertion of novel entities, if the directory entry of lower floor's processing unit less than, then insert successfully, operation stops; If, then need to continue and the last layer queue generation entity transmission of next stage again because the directory entry of lower floor's processing unit is full, carry out successively, stop up to operation;
5) whole piece priority query, be made up of 1 or many upper strata formations:
Whole piece priority query by 1 or many last layer queues, according to priority orders, is in series, and in the layer queue, comprises lower floor's processing unit on each, and lower floor's processing unit has comprised a plurality of directory entries;
6) the limit priority entity dequeue of constant time:
The entity information of limit priority, always leave in the highest highest directory entry of going up in the layer queue of lower floor's processing unit, when this entity dequeue, all follow-up directory entries are to higher level's translation, limit priority directory entry on the next stage in the layer queue will be moved in the last layer queue of upper level, and finishing the needed time of this process is constant;
7) the priority catalogue of balanced load and performance is inserted:
When novel entities inserted, because its need are given on the first order all directory entries of lower floor's processing unit in the layer queue, so the situation that the whole piece formation is given in load more simultaneously reduced; Simultaneously, the many methods of local parallel have been adopted in the insertion of entity, transmit priority mode, have saved time and all more required temporary hardware resource of relatively transmitting, are improved on the performance.
The present invention compares with background technology, and the useful effect that the present invention has is:
This design is a kind of parallel transposition formation hardware concurrency advantage that combines, and the little advantage of formation load relatively item by item, by parallel transposition and the method that relatively mutually combines item by item, design the hardware priority formation, thereby reach the characteristics of balanced load and performance.
(1) modular design makes the favorable expandability of priority query, no matter is to need to increase priority or increase directory entry, and is all relatively independent, need not change the structure of whole formation.
(2) utilize the parallel transposition formation of lower floor's processing unit, improved the concurrency and the response speed of priority query.
(3) formation of transmission item by item of layer queue in the utilization makes the insertion of novel entities, only needs that all directory entries compare in the last layer queue with the first order, has reduced load.
Description of drawings
Fig. 1 goes up the layer queue synoptic diagram.
Fig. 2 is the processing unit figure of lower floor.
Specific implementation method
Relate to relevant symbolic interpretation in the method:
Compare
Self: the comparative result of entity input port and current record, its value are that the current input entity priority of 0 (Low) expression is higher than current record priority, or the current input entity priority of 1 (High) expression is less than or equal to the entity priority in the current directory item.
Compare
Left: the comparative result of directory entry upper level directory entry, its value is the priority that the current input entity priority of 1 (High) expression is higher than entity in the upper level directory entry, or the current input entity priority of 0 (Low) expression is less than or equal to the entity priority in the upper level directory entry.
The hardware priority formation implementation method of a kind of balanced load and performance, its general structure is made up of according to the priority orders polyphone layer queue on several, on each in the layer queue, lower floor's processing unit is arranged, this unit is made up of some directory entries, and directory entry adopts parallel transposition formation mode to realize.The formation block diagram of balanced load and performance as shown in Figure 1.Fig. 2 is lower floor's processing unit internal frame diagram.Directory entry and lower floor's number of processing units are according to hardware resource and concrete user demand decision, and the operation committed step is as follows:
1) lower floor's processing unit is made up of some directory entries:
Each lower floor's processing unit is made up of some identical directory entries, and each directory entry is a hardware cell, comprises access hole, control signal and entity information I/O port.Between the adjacency, by the data path identical with data bit width, and the compare result signal of a bit wide connects.Each has write down entity information, comprises entity priority and other information;
2) lower floor's processing unit novel entities inserts, and according to parallel transposition lining up mode, adopts parallel mode relatively to realize:
The initialization of last layer queue and lower floor's processing unit, at first, connected in series with layer queue on each, come the formation stem, be the five-star layer queue of going up.In the lower floor's processing unit of each, directory entry all places disarmed state, and the expression directory entry be a sky, for the directory entry of sky is special directory entry, and the comparative result Compare of itself and entity priority
SelfBe 0.When new entity arrived, the entity information of this entity will be sent to the entity input port of each directory entry simultaneously, and simultaneously, control signal informs that each directory entry has novel entities to arrive.When directory entry finds to have novel entities to arrive, the entity priority of more own current record immediately and the priority of novel entities, if the priority of novel entities is higher than the entity of current record, and the comparative result of upper level directory entry output is higher than the current entity of its record for novel entities priority, then this directory entry reads in the current entity information recorded content of upper level directory entry, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage catalogue; If the priority of novel entities is higher than the current record entity, and the comparative result of upper level directory entry output is lower than the current entity of its record for novel entities priority, then this directory entry reads in novel entities information, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage directory entry; If novel entities priority is lower than the current record entity, then keep original entity information constant, simultaneously self comparative result is outputed on the compare result signal, as the input of next stage directory entry.Once insert operation, the periodicity that needs altogether is 3 bats, first count input novel entities, and second count produces comparative result, and triple time, according to comparative result, whole priority query carries out the translation of above-mentioned rule, to reach the series arrangement of priority.Concrete comparison procedure, its entity information and priority will be sent to five-star going up in the layer queue, and the lower floor's processing unit in the last layer queue is input to its inner directory entry, the comparison that walks abreast with what it walked abreast.Each directory entry independently compares the priority of self record entity and the priority of input entity, obtains Compare
Self, in conjunction with the Compare that obtains from the upper level directory entry
LeftValue transplants according to following rule: work as Compare
Self=0 and Compare
Left=0, expression input entity inserts before the upper level directory entry, and then the current directory item reads in the entity information of its record from the upper level directory entry, as the entity information of current directory item, simultaneously, original entity information is outputed on the data path that is connected with the next stage directory entry; Work as Compare
Self=0 and Compare
Left=1, the expression novel entities should insert in current location, and then the current directory item as the current information record, outputs to novel entities information on the data path that is connected with the next stage directory entry with original entity information simultaneously; Work as Compare
Self=1 o'clock, the expression novel entities should insert in the follow-up catalogue of current directory item, and therefore, the current directory item remains unchanged.Parallel work-flow between each directory entry.
3) go up layer queue and realize that by transmitting models of priority layer queue on each comprises lower floor's processing unit, between the last layer queue, according to priority, series arrangement:
Last layer queue is realized according to transmitting models of priority, between each level, connects by data path, control signal, and series arrangement according to the priority.Its inside is lower floor's processing unit;
4) go up entity transmission between the layer queue, when the directory entry that occurs in previous stage upper strata inner queue lower floor processing unit exhausts:
When novel entities need insert, if in lower floor's processing unit of novel entities insertion position, priority query's directory entry less than, then insert according to the lower floor processing unit novel entities insertion method of above mentioning, need not anyly move between the last layer queue.In lower floor's processing unit of novel entities insertion position, priority query's directory entry is full, then need transmit entity information between last layer queue, at this moment, novel entities is inserted the position that needs insertion, simultaneously, to overflow the directory entry of the lowest priority of formation, move on the next stage in the layer queue.The entity information of the entity that transmits between the last layer queue and control signal are exported by the current layer queue of going up, layer queue is received signal on the next stage, and its inner lower floor processing unit begins the insertion of novel entities, if the directory entry of lower floor's processing unit less than, then insert successfully, operation stops; If, then need to continue and the last layer queue generation entity transmission of next stage again because the directory entry of lower floor's processing unit is full, carry out successively, stop up to operation;
5) whole piece priority query, be made up of the formation of some upper stratas:
Whole piece priority query by some layer queues of going up, according to priority orders, is in series, and in the layer queue, comprises lower floor's processing unit on each, and lower floor's processing unit has comprised some directory entries.
6) the limit priority entity dequeue of constant time:
The entity information of limit priority, always leave in the highest highest directory entry of going up in the layer queue of lower floor's processing unit, when this entity dequeue, all follow-up directory entries are to higher level's translation, limit priority directory entry on the next stage in the layer queue will be moved in the last layer queue of upper level, and finishing the needed time of this process is constant; When the highest-priority queue dequeue is specifically implemented, only need five-star directory entry output in the layer queue on the superlative degree, then obtained the entity information of limit priority, simultaneously, control signal is issued to layer queue on all, makes lower floor's processing unit tell all directory entries, to directory entry of high priority direction translation, between the last layer queue, move to the high priority direction by low priority.
7) the priority catalogue of balanced load and performance is inserted:
When novel entities inserted, because its need are given on the first order all directory entries of lower floor's processing unit in the layer queue, so the situation that the whole piece formation is given in load more simultaneously reduced; Simultaneously, the many methods of local parallel have been adopted in the insertion of entity, transmit priority mode, have saved time and all more required temporary hardware resource of relatively transmitting, are improved on the performance.
Claims (1)
1. the hardware priority formation implementation method of balanced load and performance is characterized in that:
1) lower floor's processing unit is made up of some directory entries:
Each lower floor's processing unit, form by some identical directory entries, each directory entry is a hardware cell, comprise access hole, control signal and entity information I/O port are between the adjacency, by the data path identical with data bit width, and the compare result signal connection of a bit wide, each has write down entity information, comprises entity priority and other information;
2) lower floor's processing unit novel entities inserts, and according to parallel transposition lining up mode, adopts parallel mode relatively to realize:
When new entity arrives, the entity information of this entity will be sent to the entity input port of each directory entry simultaneously, simultaneously, control signal informs that each directory entry has novel entities to arrive, when directory entry finds to have novel entities to arrive, the entity priority of more own current record immediately and the priority of novel entities, if the priority of novel entities is higher than the entity of current record, and the comparative result of upper level directory entry output is higher than the current entity of its record for novel entities priority, then this directory entry reads in the current entity information recorded content of upper level directory entry, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage catalogue; If the priority of novel entities is higher than the current record entity, and the comparative result of upper level directory entry output is lower than the current entity of its record for novel entities priority, then this directory entry reads in novel entities information, as current entity information, simultaneously with original entity information, output on the data path that is connected with the next stage directory entry, and self comparative result is outputed on the compare result signal, as the input of next stage directory entry; If novel entities priority is lower than the current record entity, then keep original entity information constant, simultaneously self comparative result is outputed on the compare result signal, as the input of next stage directory entry, once insert operation, the periodicity that needs altogether is 3 bats, first count input novel entities, and second count produces comparative result, triple time, according to comparative result, whole priority query carries out the translation of above-mentioned rule, to reach the series arrangement of priority;
3) go up layer queue and realize that by transmitting models of priority layer queue on each comprises lower floor's processing unit, between the last layer queue, according to priority, series arrangement:
Last layer queue is realized according to transmitting models of priority, between each level, connects by data path, control signal, and series arrangement according to the priority, its inside is lower floor's processing unit;
4) go up entity transmission between the layer queue, when the directory entry that occurs in previous stage upper strata inner queue lower floor processing unit exhausts:
When novel entities need insert, if in lower floor's processing unit of novel entities insertion position, priority query's directory entry less than, then insert according to lower floor's processing unit novel entities insertion method of above mentioning, need not anyly move between the last layer queue, in lower floor's processing unit of novel entities insertion position, priority query's directory entry is full, then need transmit entity information between last layer queue, at this moment, novel entities is inserted the position that needs insertion, simultaneously, will overflow the directory entry of the lowest priority of formation, move on the next stage in the layer queue, the entity information of the entity that transmits between the last layer queue and control signal are exported by the current layer queue of going up, layer queue is received signal on the next stage, and its inner lower floor processing unit begins the insertion of novel entities, if the directory entry of lower floor's processing unit less than, then insert successfully, operation stops; If, then need to continue and the last layer queue generation entity transmission of next stage again because the directory entry of lower floor's processing unit is full, carry out successively, stop up to operation;
5) whole piece priority query, be made up of 1 or many upper strata formations:
Whole piece priority query by 1 or many last layer queues, according to priority orders, is in series, and in the layer queue, comprises lower floor's processing unit on each, and lower floor's processing unit has comprised a plurality of directory entries;
6) the limit priority entity dequeue of constant time:
The entity information of limit priority, always leave in the highest highest directory entry of going up in the layer queue of lower floor's processing unit, when this entity dequeue, all follow-up directory entries are to higher level's translation, limit priority directory entry on the next stage in the layer queue will be moved in the last layer queue of upper level, and finishing the needed time of this process is constant;
7) the priority catalogue of balanced load and performance is inserted:
When novel entities inserted, because its need are given on the first order all directory entries of lower floor's processing unit in the layer queue, so the situation that the whole piece formation is given in load more simultaneously reduced; Simultaneously, local parallel method has relatively been adopted in the insertion of entity, transmits priority mode, has saved time and all more required temporary hardware resource of relatively transmitting, is improved on the performance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101629034A CN101419562B (en) | 2008-12-04 | 2008-12-04 | Hardware PRI queue implementing method for balancing load and performance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101629034A CN101419562B (en) | 2008-12-04 | 2008-12-04 | Hardware PRI queue implementing method for balancing load and performance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101419562A CN101419562A (en) | 2009-04-29 |
| CN101419562B true CN101419562B (en) | 2010-06-02 |
Family
ID=40630358
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101629034A Expired - Fee Related CN101419562B (en) | 2008-12-04 | 2008-12-04 | Hardware PRI queue implementing method for balancing load and performance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101419562B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103246653B (en) * | 2012-02-03 | 2017-07-28 | 腾讯科技（深圳）有限公司 | Data processing method and device |
| CN106685854B (en) * | 2016-12-09 | 2020-02-14 | 浙江大华技术股份有限公司 | Data sending method and system |
| EP3535956B1 (en) | 2016-12-09 | 2021-02-17 | Zhejiang Dahua Technology Co., Ltd | Methods and systems for data transmission |
| CN107728953B (en) * | 2017-11-03 | 2021-03-02 | 记忆科技(深圳)有限公司 | Method for improving mixed read-write performance of solid state disk |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100417114C (en) * | 2005-03-01 | 2008-09-03 | 华为技术有限公司 | Method for realizing load balance between access devices in WLAN |
-
2008
- 2008-12-04 CN CN2008101629034A patent/CN101419562B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100417114C (en) * | 2005-03-01 | 2008-09-03 | 华为技术有限公司 | Method for realizing load balance between access devices in WLAN |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101419562A (en) | 2009-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100437522C (en) | Long-distance inner server and its implementing method | |
| US10412021B2 (en) | Optimizing placement of virtual machines | |
| CN101667144B (en) | Virtual machine communication method based on shared memory | |
| CN100346268C (en) | Dynamic power consumption management method in information safety SoC based on door control clock | |
| CN110262754B (en) | NVMe and RDMA-oriented distributed storage system and lightweight synchronous communication method | |
| CN101419562B (en) | Hardware PRI queue implementing method for balancing load and performance | |
| CN102118434A (en) | Data packet transmission method and device | |
| CN102882864A (en) | Virtualization system based on InfiniBand cloud computing network | |
| CN101051280A (en) | Intelligent card embedded operation system and its control method | |
| US20050094649A1 (en) | Logical ports in trunking | |
| CN101902390B (en) | Unicast and multicast integrated scheduling device, exchange system and method | |
| CN101467117A (en) | Power saving in circuit functions through multiple power buses | |
| WO2022037265A1 (en) | Edge computing center integrated server | |
| CN111611180B (en) | Dynamic shared buffer area supporting multiple protocols | |
| Sun et al. | Republic: Data multicast meets hybrid rack-level interconnections in data center | |
| CN104360982A (en) | Implementation method and system for host system directory structure based on reconfigurable chip technology | |
| US7443799B2 (en) | Load balancing in core-edge configurations | |
| Zhang et al. | A multi-VC dynamically shared buffer with prefetch for network on chip | |
| US11936758B1 (en) | Efficient parallelization and deployment method of multi-objective service function chain based on CPU + DPU platform | |
| JP2001142845A (en) | Computer system and data transfer control method | |
| CN101437033A (en) | Method and network appliance for supporting variable velocity | |
| CN101853185B (en) | Blade server and service dispatching method thereof | |
| Luo et al. | Traffic-aware VDC embedding in data center: A case study of fattree | |
| US20050105904A1 (en) | Frame traffic balancing across trunk groups | |
| CN113821317A (en) | Edge cloud collaborative micro-service scheduling method, device and equipment |
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: 20100602 Termination date: 20111204 |