CN104065568B - Web server cluster routing method - Google Patents
Web server cluster routing method Download PDFInfo
- Publication number
- CN104065568B CN104065568B CN201410321120.1A CN201410321120A CN104065568B CN 104065568 B CN104065568 B CN 104065568B CN 201410321120 A CN201410321120 A CN 201410321120A CN 104065568 B CN104065568 B CN 104065568B
- Authority
- CN
- China
- Prior art keywords
- request
- web
- server
- record
- hotspot
- 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.)
- Active
Links
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a Web server cluster routing method. A cache server is arranged to generate and maintain a hot spot request record, and an analyzing server is utilized for generating and maintaining an ordinary request record and updating the hot spot request record according to the ordinary request record. When the request reaches, the cache server is used for carrying out matching in the hot spot request record, if a matching item is found out, the web request is forwarded to a recorded target response server, and if the matching item is not found out, the request is forwarded to the analyzing server to be analyzed and routed to the target response server. According to the Web server cluster routing method, the hot spot request and the ordinary request are processed in a distinguished mode, the requests are reasonably allocated, and the processing capacity and the responding speed for the hot spot request can be greatly improved. Meanwhile, the hot spot request record is updated, the validity is kept, and the method can adapt to the accessing environment with the variable load and the low determinacy. The method is suitable for routing management of the Web server cluster.
Description
Technical field
The present invention relates to network transmission technology field is and in particular to a kind of routing policy of web server cluster.
Background technology
Cloud computing refers to for computational load, application program to move to high in the clouds by extension, Intel Virtualization Technology, and composition is based on web
A large amount of distributed computer common pool of the Internet, user passes through ordinary terminal, and such as computer, notebook, mobile phone accesses meter
Calculation machine common pool carries out data calculating as required, data storage data obtains service.Cloud computing mode provides available
, network access easily, on demand, enter configurable computing resources shared pool, its resource may include network, server,
Storage, application software, service, these resources can quickly be provided, and only need to put into little management work, or and service provision
Business carries out little interaction.And web server Clustering is then a basic necessary supportive skill of cloud computing platform
Art.
Web server Clustering is on the basis of conventional web server technology, integrating parallel process, distributed
The technology such as calculating develop, and it is to pick up multiple conventional web server using parallel or distributed type colony technical battery
Come, constitute the web server cluster of collaborative work, obtain high-performance calculation ability, massive store, for ten million or even more than one hundred million levels
User provides high reliability, high security, the online game of heavy load, ecommerce and streaming media video program request web application clothes
Business.
Whole web server cluster system, has single virtual ip address, i.e. cluster address, user side passes through virtual ip
Address accesses web server cluster system.The request response of web server cluster is to be realized by web request route system.
Web request route system is generally made up of some servers of web request allotter and rear end, and web request allotter serves as group
Knit and coordinate the role of server work in cluster, its ip address is cluster address, web request allotter is as arrival request
Agency, is responsible for concentrating the http request receiving all arrival, and according to specific load distribution and balance policy by user side
The back-end server requesting assignment in cluster.The server of rear end is then responsible for place's reason web request allotter and is forwarded
Similar request, optimization invoking server resource carrys out respond request.
Therefore, load sharing policy namely routing policy directly affect the performance of web server group system, be one non-
The research field often enlivened.In load distribution, one needs the key factor considering to be load balancing, and load in a balanced way is assigned
Help balance the calculation resources playing every server and ability, thus obtaining satisfied response performance.
Traditional web services, most of web page is static content, and the load of web request is little, therefore load point
Join all using better simply algorithm, such as rr Round-Robin Algorithm and stochastic selection algorithm.These allocation strategies expense very little itself, and not
Need to keep the status information of client server, execution efficiency is very high, but these allocation strategy load balance ability are very
Weak, more primary system load distribution can only be realized.
But the appearance with web site dynamic content, load difference produced by request is increasing, traditional repeating query etc.
Algorithm increasingly seems limitation.And under cloud computing environment, web server cluster adopts virtual technology, is more concerned with expansible
Property, motility, distributed high efficiency computation structure and pattern, it comprises magnanimity level dynamic content-data, and web request belongs to
Magnanimity rank and have multiformity limited nature, leads to the changing load of web server cluster, definitiveness is low, therefore, traditional
Web server group system routing policy, can not be adapted to current application and require, easily because the locality that user accesses is lived
Jump property, and it is excessive partial load.
Content of the invention
The technical problem to be solved is to provide a kind of method for routing of web server cluster, and it can adapt to
Access the low feature of changing load, definitiveness in current web application, can reasonably distribute web request, and equalization server is negative
Carry so that whole system execution efficiency is high, faster system response.
The present invention solves technical problem and be the technical scheme is that a kind of method for routing of web server cluster, described
Web server cluster includes web request route system and response server cluster, and described web request route system includes asking
Allotter and resolution server cluster, are provided with caching clothes between described web request allotter and parsing server cluster
Business device, is generated and safeguarded by caching server the hotspot request record for recording focus web request property value, and focus web please
Ask property value to include web request name, purpose response server address and request to count;By the parsing clothes in resolution server cluster
Business device generates and safeguards common request record, and described common request record is used for recording the property value of common web request, commonly
Web request property value includes web request name, purpose response server address and request and counts;Web request route system is according to general
Logical request record is updated to hotspot request record managing;
Its method for routing is as follows:
Step one, the web request unification from outside are distributed to caching server through request distributor;
Step 2, caching server carry out filtration treatment to the web request being received, and described filtration treatment is buffer service
Device carries out coupling in hotspot request record and filters to received web request, as follows according to matching result processing mode;
If mode 1 finds occurrence in hotspot request record, this web request is gone in hotspot request record
The purpose response server of record, and update the request counting of this focus web request;
If mode 2 does not find occurrence in hotspot request record, forward requests to resolution server cluster,
Resolution server passes through to search registered routing configuration table, the purpose response server that this request is routed to, and updates general
The data of logical request record.
Further, set a cache management cycle t, every the t time, once more new management carried out to hotspot request record,
Step is as follows:
The common request record of each resolution server in step a, collect statistics resolution server cluster, calculates each common
The request of web request counts;
Step b, caching server to each common web request request count and each focus web request request count into
Row calculates sequence, redefines focus web request and common web request;
Step c, the focus web request renewal focus web request record according to new determination.
Further, it is provided with least two caching clothes between described web request allotter and parsing server cluster
Business device simultaneously press trunking mode and is configured, functional role in web request route system for each caching server, process logic and data
All consistent;When hotspot request record is updated with management, before calculating sequence, first to each caching server hotspot request
The identical web request of record merges;After calculating sequence, please to each web in hotspot request record and common request record
The value of calculation asked resets;After completing the renewal of hotspot request record, the hotspot request record of each caching server is synchronized.
Specifically, in described step one, from outside web request unification by request distributor according to tcp/ip agreement wheel
Follow and be transmitted to each caching server.
Further, Hash process is carried out to the field in hotspot request record, in described step 2, caching server
Calculate the cryptographic Hash of received web request, and coupling is carried out in hotspot request record by cryptographic Hash and filter.
Further, when selecting resolution server in resolution server cluster, using concordance hash algorithm.Further
, it is respectively directed to each resolution server and introduce dummy node, and the quantity inverse ratio of dummy node corresponding with each resolution server
Loading level in this resolution server.Specifically, it is provided with j platform resolution server, by siRepresent i-th resolution server, lead to
Cross following steps to determine and resolution server siThe quantity of corresponding dummy node:
Step 1, calculating resolution server siPerformance indications p (si) and load value l (si), formula is as follows:
p(si)=k1pcpu(si)+k2pmem(si)+k3pdsk(si)+k4pbd(si)
Wherein, pcpuRepresent nominal cpu frequency, pmemRepresent nominal memory size, pdskThe disk representing nominal holds
Amount, pbdRepresent the nominal network bandwidth, knRepresent each index nominal value weight coefficient and
l(si)=c1lcpu(si)+c2lmem(si)+c3lio(si)+c4lbd(si)+c5lrsp(si)
lcpuRepresent cpu utilization rate, lmemRepresent memory usage, lioRepresent disk i/o occupancy, lbdRepresent Netowrk tape
Wide occupancy, lrspRepresent response time, cmRepresent each index utilization rate weight coefficient and
Step 2, the l (s being obtained according to step 1i) and p (si) calculate resolution server siWeight w (si)=l (si)/p
(si);
Step 3, between the cryptographic Hash ring region of definition on, total dummy node quantity is n, corresponding resolution server siVoid
Intending number of nodes is
The invention has the beneficial effects as follows: in substantial amounts of web request, being constantly present a lot of web request is identical or class
As, that is, the localized phenomena that accesses, therefore, the setting by caching server for the present invention, by focus web request and common
Web request is made a distinction and is processed by way of different.Caching server, when receiving request, was carried out to request first
Filter, focus web request is directly routed to purpose according to the purpose response server address of record in its hotspot request record and rings
Answer server,
Disposal ability and the response speed to focus web request can greatly be strengthened, reasonably can distribute web request, hold
Line efficiency is high, faster system response;Meanwhile, web request route system is carried out more to hotspot request record according to common request record
New management, keeps the effectiveness of hotspot request record so as to can adapt to access changing load, definitiveness in current web application
Low feature.It is particularly suited for the routing management of web server cluster under cloud computing environment.
Brief description
Fig. 1 is the deployment diagram of the router engine of the present invention.
Fig. 2 is the process chart to web request for the routing engine of the present invention.
Specific embodiment
With reference to Figure of description and embodiment, the present invention is further described.
A kind of method for routing of web server cluster, as shown in figure 1, institute includes web request using web server cluster
Route system and response server cluster, described web request route system includes request distributor and resolution server cluster,
It is provided with caching server between described web request allotter and parsing server cluster, generated and tieed up by caching server
Protect the hotspot request record for recording focus web request property value, focus web request property value includes web request name, purpose
Response server address and request count;Generated and safeguarded by resolution server common request record, described common request record
For recording the property value of common web request, common web request property value includes web request name, purpose response server address
Count with request;Web request route system relatively conventional request record is updated to hotspot request record managing.
Above-mentioned focus web request is the higher web request of frequency, and common web request is the relatively low request of frequency, focus
The judgement of web request and common web request can judge can also be judged according to actual access situation according to a setting value,
In the present embodiment, as described later, judged according to the current intelligence of actual access.
Specifically, its routing policy is as follows, as shown in Figure 2:
Step one, the web request unification from outside are distributed to caching server through request distributor;
Step 2, caching server carry out filtration treatment to the web request being received, and described filtration treatment is buffer service
Device carries out coupling in hotspot request record and filters to received web request, as follows according to matching result processing mode;
If mode 1 finds occurrence in hotspot request record, this web request is gone in hotspot request record
The purpose response server of record, and update the request counting of this focus web request;
If mode 2 does not find occurrence in hotspot request record, forward requests to resolution server cluster,
Resolution server to the web request being received, by searching registered routing configuration table, ring by the purpose that this request is routed to
Answer server, and update the data of common request record.
In order to realize quick locating function, Hash process is carried out to the field in hotspot request record, in described step 2
In, caching server calculates the cryptographic Hash of received web request, and carries out mating in hotspot request record by cryptographic Hash
Filter.Certainly also can directly mate or be mated using other algorithms.
Further, when selecting resolution server in resolution server cluster, using concordance hash algorithm.Using one
A certain component requests can be stably routed to particular server by cause property hash algorithm, when the distribution situation change of request,
Concordance hash algorithm can guarantee that these requests and the pair relationhip change minimum between corresponding processing server originally.
In order to avoid certain time period request integrated distribution is on a certain resolution server, cause this server super negative
Carry, further, be respectively directed to each resolution server and introduce dummy node, and dummy node corresponding with each resolution server
Quantity is inversely proportional to the loading level of this resolution server.Above-mentioned dummy node is the copy in hash space for the actual node, one
Individual actual node has corresponded to several dummy nodes, and this corresponding number also referred to as replicates number, and dummy node is in hash space
In with cryptographic Hash arrangement, can make cryptographic Hash between ring region on distributing equilibrium, improve hash algorithm balance.Each parsing clothes
The quantity of the business corresponding dummy node of device is inversely proportional to the loading level of this resolution server, namely the loading level of resolution server
More big then dummy node quantity is less, and the probability that corresponding resolution server is hit is less.
Specifically, it is provided with j platform resolution server, by siRepresent i-th resolution server, determine as follows and solution
Analysis server siThe quantity of corresponding dummy node:
Step 1, calculating resolution server siPerformance indications p (si) and load value l (si), formula is as follows:
p(si)=k1pcpu(si)+k2pmem(si)+k3pdsk(si)+k4pbd(si)
Wherein, pcpuRepresent nominal cpu frequency, pmemRepresent nominal memory size, pdskThe disk representing nominal holds
Amount, pbdRepresent the nominal network bandwidth, knRepresent each index nominal value weight coefficient and
l(si)=c1lcpu(si)+c2lmem(si)+c3lio(si)+c4lbd(si)+c5lrsp(si)
lcpuRepresent cpu utilization rate, lmemRepresent memory usage, lioRepresent disk i/o occupancy, lbdRepresent Netowrk tape
Wide occupancy, lrspRepresent response time, cmRepresent each index utilization rate weight coefficient and
Step 2, the l (s being obtained according to step 1i) and p (si) calculate resolution server siWeight w (si)=l (si)/p
(si);
Step 3, between the cryptographic Hash ring region of definition on, total dummy node quantity is n, corresponding resolution server siVoid
Intending number of nodes is
Between the ring region of above-mentioned concordance hash algorithm, value is [0,232- 1), this value is also common suggestion value model
Enclose.And the value of above-mentioned n is determined according to practical situation, but maximum should be able to should be protected less than the size between ring region, minimum
The minimum w of cardiMore than or equal to 1.
After determining dummy node quantity, then determine the cryptographic Hash of each dummy node according to hash algorithm, hash algorithm can
So that using arbitrary hash algorithm, in the present embodiment, specifically, using djb hash algorithm, formula is expressed as: it is defeated for setting str
Enter character array, length is n, the cryptographic Hash of str can represent, iterative process is: hash [0]=5381 with hash [n],
Hash [i+1]=hash [i] * 33+str [i+1], i=0,1 ..., n-1.Djb hash algorithm be preferable hash algorithm it
One, its advantage includes: 1) output valve after the process of this hash function for the input data being capable of relatively evenly dispersed arrangement;2) breathe out
Uncommon conflict situations are few;3) minor variations of input value can cause the very big change of output valve, that is, so-called " avalanche effect ".
The above-mentioned management for hotspot request record, can carry out Real-time and Dynamic management it is also possible to enter according to a certain threshold values
Line pipe is managed.But the load for balanced routing efficiency with to hotspot request record management, best, set a cache management cycle
T, t=1 hour in the present embodiment, every the t time, hotspot request record is carried out with once more new management, step is as follows:
Step a, resolution server cluster set one principal solution analysis server cluster, by principal solution analysis server cluster collect system
The common request record of each resolution server in meter resolution server cluster, the request calculating each common web request counts;
Step b, caching server to each common web request request count and each focus web request request count into
Row request calculates sequence, redefines focus web request and common web request;
Step c, the focus web request renewal focus web request record according to new determination.
In order to further improve disposal ability, arrange between described web request allotter and parsing server cluster
There are at least two caching servers and by trunking mode configuration, in the present embodiment, be provided with three buffer servers altogether, each slow
Deposit functional role in web request route system for the server, process logic and data all consistent;Hotspot request record is being entered
During row more new management, before calculating sequence, first the identical web request of each caching server hotspot request record is closed
And;After calculating sequence, the value of calculation of each web request in hotspot request record and common request record is reset;Complete heat
After point request record updates, the hotspot request record of each caching server is synchronized.
Due to functional role in web request route system for the above-mentioned each caching server, process logic and data is homogeneous
Cause, therefore, in order to simplify allocation strategy, in described step one, from outside web request unification by request distributor according to
Tcp/ip agreement repeating query is transmitted to each caching server.
Claims (7)
1. a kind of method for routing of web server cluster, institute includes web request route system and sound using web server cluster
Answer server cluster, described web request route system include request distributor and resolution server cluster it is characterised in that:
It is provided with caching server between described web request allotter and parsing server cluster, generated by caching server
And safeguard hotspot request record for recording focus web request property value, focus web request property value include web request name,
Purpose response server address and request count;Generated and safeguarded by resolution server common request record, described common request
Record the property value for recording common web request, common web request property value includes web request name, purpose response server
Address and request count;Web request route system is updated to hotspot request record managing according to common request record;
Its method for routing is as follows:
Step one, the web request unification from outside are distributed to caching server through request distributor;
Step 2, caching server carry out filtration treatment to the web request being received, and described filtration treatment is caching server pair
Received web request carries out coupling in hotspot request record and filters, as follows according to matching result processing mode;
If mode 1 finds occurrence in hotspot request record, this web request is gone to record in hotspot request record
Purpose response server, and update this focus web request request count;
If mode 2 does not find occurrence in hotspot request record, forward requests to resolution server cluster, parsing
Server passes through to search registered routing configuration table, the purpose response server that this request is routed to, and updates common asking
Seek the data of record;
Described hotspot request record is updated manage method be: set a cache management cycle t, every the t time to heat
Point request record carries out once more new management, and step is as follows:
The common request record of each resolution server in step a, collect statistics resolution server cluster, calculating each common web please
The request asked counts;
The request of each common web request is counted for step b, caching server and the request of each focus web request is counted
Calculate sequence, redefine focus web request and common web request;
Step c, the focus web request renewal focus web request record according to new determination.
2. as claimed in claim 1 a kind of method for routing of web server cluster it is characterised in that: divide in described web request
It is provided with least two caching servers and by trunking mode configuration, each caching server between orchestration and parsing server cluster
Functional role in web request route system, process logic and data are all consistent;
When hotspot request record is updated with management, before calculating sequence, first each caching server hotspot request is remembered
The identical web request of record merges;After calculating sequence, to each web request in hotspot request record and common request record
Value of calculation reset;After completing the renewal of hotspot request record, the hotspot request record of each caching server is synchronized.
3. as claimed in claim 2 a kind of method for routing of web server cluster it is characterised in that: in described step one,
From outside web request unification, each caching server is transmitted to according to tcp/ip agreement repeating query by request distributor.
4. as claimed in claim 1 a kind of method for routing of web server cluster it is characterised in that:
Hash process is carried out to the field in hotspot request record, in described step 2, caching server calculates and received
The cryptographic Hash of web request, and coupling filtration is carried out in hotspot request record by cryptographic Hash.
5. a kind of web server cluster as described in claim 1 or 4 method for routing it is characterised in that: in resolution server
When selecting resolution server in cluster, using concordance hash algorithm.
6. as claimed in claim 5 a kind of method for routing of web server cluster it is characterised in that: be respectively directed to each parsing
Server introduces dummy node, and the quantity of dummy node corresponding with each resolution server is inversely proportional to the negative of this resolution server
Load degree.
7. as claimed in claim 6 a kind of method for routing of web server cluster it is characterised in that: be provided with j platform analysis service
Device, by siRepresent i-th resolution server, determine and resolution server s as followsiThe quantity of corresponding dummy node:
Step 1, calculating resolution server siPerformance indications p (si) and load value l (si), formula is as follows:
p(si)=k1pcpu(si)+k2pmem(si)+k3pdsk(si)+k4pbd(si)
Wherein, pcpuRepresent nominal cpu frequency, pmemRepresent nominal memory size, pdskRepresent nominal disk size, pbd
Represent the nominal network bandwidth, knRepresent each index nominal value weight coefficient and
l(si)=c1lcpu(si)+c2lmem(si)+c3lio(si)+c4lbd(si)+c5lrsp(si)
lcpuRepresent cpu utilization rate, lmemRepresent memory usage, lioRepresent disk i/o occupancy, lbdRepresent that the network bandwidth takies
Rate, lrspRepresent response time, cmRepresent each index utilization rate weight coefficient and
Step 2, the l (s being obtained according to step 1i) and p (si) calculate resolution server siWeight w (si)=l (si)/p(si);
Step 3, between the cryptographic Hash ring region of definition on, total dummy node quantity is n, corresponding resolution server siVirtual section
Putting quantity is
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410321120.1A CN104065568B (en) | 2014-07-07 | 2014-07-07 | Web server cluster routing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410321120.1A CN104065568B (en) | 2014-07-07 | 2014-07-07 | Web server cluster routing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104065568A CN104065568A (en) | 2014-09-24 |
CN104065568B true CN104065568B (en) | 2017-01-18 |
Family
ID=51553107
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410321120.1A Active CN104065568B (en) | 2014-07-07 | 2014-07-07 | Web server cluster routing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104065568B (en) |
Families Citing this family (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110096517B (en) * | 2014-11-04 | 2023-07-14 | 创新先进技术有限公司 | Method, device and system for monitoring cache data based on distributed system |
US11010341B2 (en) | 2015-04-30 | 2021-05-18 | Netflix, Inc. | Tiered cache filling |
CN105024938A (en) * | 2015-06-11 | 2015-11-04 | 北京思源置地科技有限公司 | Mobile load balancing method and system |
CN104994156A (en) * | 2015-07-01 | 2015-10-21 | 北京京东尚科信息技术有限公司 | Load balancing method and system for cluster |
CN106940660B (en) * | 2016-01-05 | 2020-08-14 | 阿里巴巴集团控股有限公司 | Method and device for realizing cache |
CN107346307B (en) * | 2016-05-04 | 2021-02-26 | 北京京东尚科信息技术有限公司 | Distributed cache system and method |
CN106060164B (en) * | 2016-07-12 | 2021-03-23 | Tcl科技集团股份有限公司 | Telescopic cloud server system and communication method thereof |
CN108243150B (en) * | 2016-12-23 | 2022-03-18 | 中科星图股份有限公司 | Dual-function processor |
CN108427497A (en) * | 2017-02-15 | 2018-08-21 | 联想企业解决方案(新加坡)有限公司 | Method and system for calculating power loss exponent |
CN107508758A (en) * | 2017-08-16 | 2017-12-22 | 北京云端智度科技有限公司 | A kind of method that focus file spreads automatically |
CN109102273B (en) * | 2018-08-22 | 2021-08-24 | 四川新网银行股份有限公司 | Method and system for realizing distributed intelligent payment routing |
CN109634746B (en) * | 2018-12-05 | 2022-03-01 | 四川长虹电器股份有限公司 | Web cluster cache utilization system and optimization method |
CN109617986B (en) * | 2018-12-27 | 2020-08-07 | 华为技术有限公司 | Load balancing method and network equipment |
CN110855708B (en) | 2019-11-26 | 2021-06-11 | 上海莉莉丝科技股份有限公司 | Game server architecture |
CN111246397B (en) * | 2020-01-19 | 2022-05-06 | 阿里巴巴集团控股有限公司 | Cluster system, service access method, device and server |
CN111309260B (en) * | 2020-02-16 | 2021-04-09 | 西安奥卡云数据科技有限公司 | Data storage node selection method |
CN111917851A (en) * | 2020-07-22 | 2020-11-10 | 电信科学技术第五研究所有限公司 | Load balancing scheduling method for realizing weighted load based on consistent hash |
CN112003945A (en) * | 2020-08-26 | 2020-11-27 | 杭州迪普科技股份有限公司 | Service request response method and device |
CN112076464B (en) * | 2020-09-04 | 2022-06-21 | 腾讯科技(深圳)有限公司 | Data request processing method and device, computer equipment and storage medium |
CN113079110B (en) * | 2021-04-23 | 2022-08-30 | 盛立安元科技(杭州)股份有限公司 | Message processing method, device, equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101610211A (en) * | 2009-07-15 | 2009-12-23 | 浪潮电子信息产业股份有限公司 | A kind of load balancing of cache method that realizes WRR |
CN102214192A (en) * | 2010-04-12 | 2011-10-12 | 腾讯科技(深圳)有限公司 | Method for realizing data curve chart display on Web page and server |
CN102420847A (en) * | 2010-10-20 | 2012-04-18 | 微软公司 | Routing traffic in an online service with high availability |
CN103338249A (en) * | 2013-06-26 | 2013-10-02 | 优视科技有限公司 | Cache method and device |
-
2014
- 2014-07-07 CN CN201410321120.1A patent/CN104065568B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101610211A (en) * | 2009-07-15 | 2009-12-23 | 浪潮电子信息产业股份有限公司 | A kind of load balancing of cache method that realizes WRR |
CN102214192A (en) * | 2010-04-12 | 2011-10-12 | 腾讯科技(深圳)有限公司 | Method for realizing data curve chart display on Web page and server |
CN102420847A (en) * | 2010-10-20 | 2012-04-18 | 微软公司 | Routing traffic in an online service with high availability |
CN103338249A (en) * | 2013-06-26 | 2013-10-02 | 优视科技有限公司 | Cache method and device |
Also Published As
Publication number | Publication date |
---|---|
CN104065568A (en) | 2014-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104065568B (en) | Web server cluster routing method | |
Zhang et al. | Proactive workload management in hybrid cloud computing | |
US20130332608A1 (en) | Load balancing for distributed key-value store | |
WO2017218473A1 (en) | Dynamic acceleration in content delivery network | |
US11061930B1 (en) | Dynamic management of storage object partitioning | |
Ma et al. | An improved web cache replacement algorithm based on weighting and cost | |
US20210044653A1 (en) | Method, apparatus, client terminal, and server for data processing | |
Caneill et al. | Locality-aware routing in stateful streaming applications | |
US20190207970A1 (en) | Clustering network addresses | |
Zhang et al. | Enhanced adaptive cloudlet placement approach for mobile application on spark | |
Xiong et al. | kBF: Towards approximate and bloom filter based key-value storage for cloud computing systems | |
Chou et al. | Bc-store: A scalable design for blockchain storage | |
Matri et al. | Keeping up with storage: Decentralized, write-enabled dynamic geo-replication | |
Xu et al. | Adaptive and scalable load balancing for metadata server cluster in cloud-scale file systems | |
Khanafer et al. | To rent or to buy in the presence of statistical information: The constrained ski-rental problem | |
Shabeera et al. | Bandwidth-aware data placement scheme for Hadoop | |
CN108418871A (en) | A kind of cloud storage performance optimization method and system | |
Lai et al. | A scalable multi-attribute hybrid overlay for range queries on the cloud | |
Al-Sakran et al. | A proposed performance evaluation of NoSQL databases in the field of IoT | |
Yadgar et al. | Cooperative caching with return on investment | |
Cremonezi et al. | Improving the attribute retrieval on ABAC using opportunistic caches for fog-based IoT networks | |
Huang et al. | Optimizing data partition for scaling out NoSQL cluster | |
Karolewicz et al. | On efficient data storage service for IoT | |
US11442632B2 (en) | Rebalancing of user accounts among partitions of a storage service | |
Zhang et al. | Speeding up vm startup by cooperative vm image caching |
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 | ||
CB03 | Change of inventor or designer information | ||
CB03 | Change of inventor or designer information |
Inventor after: Luo Guangchun Inventor after: Yin Guangqiang Inventor after: Tian Ling Inventor after: Qin Ke Inventor after: Chen Aiguo Inventor after: Duan Guiduo Inventor before: Luo Guangchun Inventor before: Tian Ling Inventor before: Duan Guiduo Inventor before: Ding Linbo |