CN108769102A - A method of improving QUIC agreement request dispatching efficiencies - Google Patents
A method of improving QUIC agreement request dispatching efficiencies Download PDFInfo
- Publication number
- CN108769102A CN108769102A CN201810306629.7A CN201810306629A CN108769102A CN 108769102 A CN108769102 A CN 108769102A CN 201810306629 A CN201810306629 A CN 201810306629A CN 108769102 A CN108769102 A CN 108769102A
- Authority
- CN
- China
- Prior art keywords
- annulus
- server
- connection
- point
- dynamic
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1025—Dynamic adaptation of the criteria on which the server selection is based
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
Abstract
The present invention discloses a kind of method improving QUIC agreement request dispatching efficiencies.This method realizes the dynamic select to Connection ID and the efficient scheduling to QUIC agreement requests by a kind of mapping mode based on Connection ID and consistent Hash annulus.Virtual server node by way of being mapped on consistent Hash annulus by the present invention, it has been dynamically determined effective control range of server, and the mapping relations based on effective control range Yu Connection ID, it is dynamically determined the selection section of Connection ID, by the way that QUIC requests are mapped to the efficient scheduling realized on the annulus to request.The present invention solves the problems, such as that QUIC connection smooth migrations fail under the characteristics of traditional software load balancing requests dispatching method can not adapt to QUIC agreement requests caused cluster environment, avoid the negative effect of the performance and scalability to server cluster.
Description
Technical field
The present invention relates to a kind of network request dispatching techniques, more particularly to a kind of improvement QUIC agreement request dispatching efficiencies
Method.
Background technology
Transmission Control Protocol is an extensive transport layer protocol of use, and reliable transmission, the company of TCP are provided for upper layer application
It connected the four-tuple that source IP address, source port number, purpose IP address, destination slogan are constituted and determines a connection, and worked as visitor
The four-tuple will necessarily change when network switching (such as being switched to 4G networks from Wi-Fi network) occurs for family end, this will lead
The reconstruction for causing TCP connection, affects the efficiency of communication.And QUIC agreements are then the application layer protocols based on UDP, compared to TCP
In such a way that four-tuple determines connection, QUIC agreements by 64-bit without symbol shaping value Connection ID come
Determine a connection.When client sends network switching, the Connection ID before network switching are kept by continuing,
Re-establishing for connection is can avoid, the smooth migration of connection is realized, improves the efficiency of communication.Mobile network appliance makes at present
With extensive, the case where network switching, happens occasionally, and the characteristic for connecting smooth migration is very meaningful.
But due to the server-side often presence in the form of server cluster in practical application, and server cluster needs to bear
Equalizing equipment is carried to be dispatched to client request on different servers, and the software address transformed load currently used
Equalization program is using the Hash result of IP address as the standard of scheduling so that client request before and after network switching most probably by
It is dispatched on different servers and is handled, since the session information of client and server-side is only remained in specific processing request
Server on, therefore the scheduling process result in QUIC agreements connection smooth migration characteristic failure, it means that the strategy
It is not particularly suited for the request scheduling of QUIC agreements.
A solution is global synchronization and distributes Connection ID unitedly, but since Connection ID are one
A needs frequently create the data destroyed, and the program will cause larger network I/O expense, influence the operation effect of server cluster
Rate.In addition, in order to ensure the uniqueness of Connection ID, Connection ID distribution servers also to cache it is all
Connection ID are distributed, and are searched before dispensing, when distribution is cached, and when destruction is deleted, due to entirely taking
Being engaged in device cluster, Connection ID quantity is more, and the resource which is consumed is also larger.
Another solution is to distribute a static Connection ID to each server in cluster to select model
It encloses, ensures that Connection ID are not clashed by ensureing the tatic range non-overlapping copies, but due to server-side cluster
Scale meeting dynamic change, thus the range of choice of static allocation Connection ID can cause the scalability of cluster by shadow
It rings.
Invention content
It is an object of the invention to overcome deficiencies of the prior art, providing improves the scheduling of QUIC agreement requests
The method of efficiency.
The object of the invention is achieved through the following technical solutions：
A method of improving QUIC agreement request dispatching efficiencies, method specifically comprises the following steps：
It 1) will ranging from [0,232- 1] integer set constitutes a consistent Hash torus space, will be in server cluster
Server is mapped in the form of several virtual server nodes on the annulus, and the position of each virtual server node corresponds to
An integral point in the range represented by annulus；
2) determine that the dynamic of virtual server node effectively controls in the position relationship of annulus by virtual server node
Range；
3) determine that the dynamic of server is effective by the effective control range of dynamic of the virtual server corresponding to server
Control range；
4) Connection is determined by the mapping relations of the effective control range of the dynamic of server and Connection ID
The dynamic select range of ID；
5) for arbitrarily asking, determine request on annulus by the reverse Mapping relationship of Connection ID to annulus
Mapping position, find the server of specific processing request clockwise on annulus, will be in request scheduling to the server.
Further, above-mentioned server cluster is specifically：One or more server host is realized by network communication
Loose conglomerate, server cluster cooperation are completed evaluation work, provide service to client jointly, come with the visual angle of client
It sees, specific service effect is as only having separate unit host in offer service.
Further, above range is [0,232- 1] space that integer set constitutes an annulus shape is specifically：If
The space of a fixed annulus shape, will ranging from [0,232- 1] uniform increments divide integer value in the direction of the clock on this annulus
Cloth, be necessarily for any point X+1 on annulus point X in the clockwise direction closest to a point, and point X+2 and point X+1
Distance it is inevitable and point X+1 is equal at a distance from X, put 232- 1 point closest clockwise is point 0, and puts 232‐1
With point 0 at a distance from and annulus on any point X+1 it is equal at a distance from X.
Further, above by virtual server node virtual server node is determined in the position relationship of annulus
Dynamically effectively control range specific method is：[0,232- 1] integer set constitutes an annulus shape spatially, each
An integral point of a virtual server node in wherein occupying range represented by an annulus, for arbitrary virtual server
Nodes X, if its corresponding integer value is NX, using its position on annulus as starting point, counterclockwise moved along annulus, until meeting
Until other any virtual server node Y, if the corresponding integer value of its point is NY, then the dynamic of the virtual server nodes X
Effective control range is (NY,NX]。
Further, server is determined above by the effective control range of dynamic of the virtual server corresponding to server
The effective control range of dynamic be specifically：For arbitrary server S, if its corresponding virtual server node on annulus is
S1、S2、……、SX, wherein SNThe effective control range of dynamic of virtual server node is RSN, then the dynamic of server S effectively control
Make ranging from RS1∪RS2∪……∪RSX。
Further, true above by the effective control range of the dynamic of server and the mapping relations of Connection ID
The dynamic select range for determining Connection ID is specifically：The Connection ID of QUIC are 64-bit length without symbol
Number shaping value, value range are [0,264- 1], for any point N on annulus, the Connection ID that can map out
Value range be [N × 232,(N+1)×232- 1], for any two points A, B and B on annulus>A, if the dynamic of server S has
It imitates ranging from [A, B], then dynamic select ranging from [A × 2 of its corresponding Connection ID32,(B+1)×232‐1]。
Further, determine request on annulus above by the reverse Mapping relationship of Connection ID to annulus
Mapping position, finding the server specific method of specific processing request clockwise on annulus is：For arbitrary Connection
For ID is the connection of C, the mapping point on annulus is (C+1)/232Downward rounding obtain as a result, setting the result as N, then
It is searched clockwise along positions of the point N on annulus, the server corresponding to first encountered virtual server node is i.e.
Specifically to handle the server of request.
Above-mentioned virtual server node is specifically：Mapping point of the server on consistent Hash annulus, a server can
To be mapped as multiple virtual server nodes, each virtual server node has and only a server is corresponding to it,
For arbitrary server S, the virtual server node S that is mapped out on consistent Hash annulus1、S2、……、Sn?
Representative server S, namely the request being scheduled on certain virtual server node is by the clothes corresponding to the virtual server node
Business device is handled.
Compared with prior art, the invention has the advantages that and technique effect：
1) scheduling that the present invention is made requests on the basis of Connection ID, in server-side in the form of server cluster
In the presence of, it is ensured that QUIC agreements connection can client occur network switching when smooth migration.Meanwhile the present invention
By the mapping relations of Connection ID and consistent Hash annulus range, it is dynamically determined the selection model of Connection ID
It encloses, improves QUIC agreement request dispatching efficiencies.
2) compared to IP address-based cryptographic Hash carry out QUIC request schedulings strategy, the present invention due to the use of
Benchmark of the Connection ID as scheduling, can avoid client in the presence of server-side is in the form of server cluster
When holding network switching due to load balance scheduling the problem of caused connection smooth migration failure.
3) compared to by a certain or a few servers to the Connection ID on Servers-all in cluster into
Row global synchronization and the method for unified distribution, in the present invention, servers of the Connection ID still by specifically handling request
It is responsible for creating, cache and destroying, does not will produce the huge network I/O expense caused by global synchronization Connection ID, also not
It can be because focusing on the performance pressures caused by Connection ID all in entire server cluster.Typically, once
The expense that cluster internal Connection ID synchronize caused network I/O was about 500000 nanoseconds, and was locally generated
The expense of Connection ID was only about for 100 nanoseconds, for the opposite program, Connection ID selection schemes of the invention
The dispatching efficiency of request can be made to promote about 500 times.
4) compared to by distributing a static Connection ID range of choice simultaneously to each server in cluster
Server by specifically handling request carry out the establishment of Connection ID, caching and destruction method, in the present invention,
The range of choice of Connection ID can avoid static range of choice and make according to the state dynamic change of server cluster
At the constraint to server cluster scalability.
Description of the drawings
Fig. 1 is the schematic diagram of consistent Hash annulus of the present invention；
Fig. 2 is the schematic diagram present invention determine that Connection ID ranges of choice；
Fig. 3 is Connection ID ranges of choice in the present invention according to the schematic diagram of cluster scale dynamic change；
Fig. 4 is scheduling schematic diagram of the present invention to client request；
Fig. 5 is the flow chart of load balance scheduling client request in the present invention；
Fig. 6 is the flow chart for the server processing requests that actual treatment is asked in server cluster of the present invention.
Specific implementation mode
To more fully understand the present invention, the present invention is further illustrated with embodiment below in conjunction with the accompanying drawings, but this
The way of example of invention is unlimited so.
To solve the problems, such as that QUIC connection migrations fail under server cluster environment caused by conventional load equalizing equipment, together
When avoid performance issue caused by being distributed due to Connection ID global synchronizations and static state regulation Connection ID from selecting
Select the problem that the cluster retractility caused by range is limited, a kind of side improving QUIC agreement request dispatching efficiencies presented below
Method effectively controls model by the mapping and server of Connection ID to consistent Hash annulus on consistent Hash annulus
The Mapping implementation to Connection ID that encloses efficiently and is dynamically selected Connection ID, and is carried out as benchmark
The scheduling of request.
As shown in Figure 1, will ranging from [0,232- 1] a consistent Hash torus space is constituted in integer set, in set
All integers sequentially discharge along clockwise direction on annulus, and first place is connected, and is uniformly distributed.For arbitrary N ∈ [0,232- 1),
It must come across on some position of annulus, and N+1 must also come across on some position of annulus, and be with the position where N
It must be positions of the N+1 on annulus that point searches first point encountering along annulus clockwise, and the position where N and N+1
The distance between the position at place is equal with the distance between 0 position and 1 position.
As illustrated in fig. 2, it is assumed that server cluster is by S1、S2、S3Three servers are constituted, and each server corresponds to respectively
Ranging from [0,232- 1] three virtual server nodes on the consistent Hash annulus of range, virtual server node, that is, server
Mapping point on consistent Hash annulus, a server can be mapped as multiple virtual server nodes, each is virtual
Server node has and only a server is corresponding to it, and for arbitrary server S, is shown in consistent Hash annulus
The virtual server node S of injection1、S2、……、SnCan representative server S, namely be scheduled for certain virtual server node
On request handled by the server corresponding to the virtual server node.With server S2For, it is corresponding virtual
Server is respectively S2_0、S2_1、S2_2, and the mapping position on annulus of these three virtual server nodes is respectively N2_0、
N2_1、N2_2, the position of corresponding previous virtual server is respectively N1_0、N1_1、N1_2, then, with N2_1Position it is virtual
For server, effective control range on annulus is [N1_1+1,N2_1], that is to say, that it is mapped within the scope of this
Request can be given S2Processing.Clearly for server S2For, effective control range (overstriking in figure on annulus
Circle ring area) be：
RangeS2=[N1_0+1,232–1]∪[0,N2_0]∪[N1_1+1,N2_1]∪[N1_2+1,N2_2]
It is the partial enlarged view of annulus on the right side of Fig. 2, since each mapping point on entire annulus is that value range is
[0,232- 1] integer, and Connection ID itself are that 64-bit is without symbol shaping value, that is, value range
[0,264- 1], therefore, for any point N on annulusx, it can be mapped as [a Nx×232,(Nx+1)×232–1]
Connection ID value ranges.For virtual server S2_1For, effective control range on annulus maps out
Connection ID ranges of choice be [(N1_1+1)×232,(N2_1+1)×232–1].For server S 2,
The Connection ID ranges of choice that its effective control range on annulus maps out are：
CIDRangeS2=[(N1_0+1)×232,264–1]∪[0,(N2_0+1)×232–1]∪
[(N1_1+1)×232,(N2_1+1)×232–1]∪[(N1_2+1)×232,(N2_2+1)×232–1]
Since effective control range of each server is non-overlapping copies, it is clear that this selection algorithm can ensure to select
The uniqueness for the Connection ID for selecting out.And when a new server is added to cluster, which will be with void
The form of quasi- server is mapped on annulus, and the mapping will pass through effective control of the virtual server on dynamic adjustment annulus
Range processed achievees the purpose that dynamic adjusts the range of choice of Connection ID.
As shown in figure 3, working as virtual server node Sx_1When being mapped on annulus, clockwise direction encountered
One virtual server node (S2_1) effective control range be affected, from (N1_1,N2_1] it is reduced to (Nx_1,N2_1], accordingly
Ground, S2_1Connection ID ranges of choice also from [(N1_1+1)×232,(N2_1+1)×232- 1] it is reduced to [(Nx_1+1)×
232,(N2_1+1)×232- 1], S2_1Corresponding server S2On belong to [(N1_1+1)×232,(Nx_1+1)×232- 1] range
Connection ID and its related data are migrated to Sx_1Corresponding server SxOn.At this point, S2_1New effective control range
All Connection ID on Connection ID selections section mapped out can be correctly dispatched to corresponding with service
Device S2On without with Sx_1Effective control range conflict.Influence caused by increasing with virtual server node is corresponding, works as void
Quasi- server node SxWhen being deleted from annulus, S2_1The virtual server node on annulus will be searched counterclockwise, when finding the
One virtual server node (S1_1) when, Connection ID ranges of choice will be extended to [(N again1_1+1)×232,
(N2_1+1)×232–1]。
As shown in figure 4, when load-balancing device receives the request from client, load-balancing device asks extraction
Request is simultaneously mapped on consistent Hash annulus by the Connection ID that ask.After request is mapped on annulus, along annulus up time
Needle direction is moved, and the server corresponding to first virtual server node encountered is the service of specific processing request
Device, load-balancing device will be in the request schedulings to the server.
The scheduling flow is specifically as shown in Figure 5：
Step 501：Load-balancing device obtains the request data of QUIC connections；
Step 502：Load-balancing device parses request by Connection ID in the packet header QUIC
Connection ID；
Step 503：Load-balancing device judge the request whether be in QUIC agreements CHLO (Client Hello, visitor
Shake hands at family end) message, in QUIC agreements, Connection ID are selected before client is shaken hands with server-side by client at random
It selects, when shaking hands, server-side can select to be replaced, in order to ensure Connection ID are in the uniqueness of server-side, service
End must generate unique Connection ID in a server-side.CHLO messages are determined whether here, it can be determined that should
Whether the Connection ID of request are that server-side generates；
Step 504：If request is CHLO messages, illustrate that the Connection ID of the request are generated by client, in order to keep away
The mistake of the malicious requests or client of exempting from client, which is realized, causes request by centralized dispatching to part host, will ask herein
Ask any point on Random Maps to annulus；
Step 505：If request is not CHLO messages, illustrate that the Connection ID are generated by server-side, according to
On the value reverse Mapping to annulus of Connection ID.The reverse Mapping be specially：If the Connection ID of request are
C, then its corresponding points on annulus isNamely to (C+1)/232The downward rounding of result, such knot
Fruit may insure to find corresponding point on annulus.Request in server S is scheduled for for any one, since it is in circle
Mapping position on ring necessarily belongs to effective control range of some virtual server node corresponding to server S, and services
The range of device S selection Connection ID is also in the range, it is ensured that after the Connection ID updates of this connection,
The request of the client still can be correctly scheduled in server S；
Step 506：After load-balancing device determines position of the request on annulus, Virtual Service is searched along clockwise direction
Device node selects first virtual server node encountered, and will be in request scheduling to the server.
The flow for the server processing requests that actual treatment is asked in server cluster is as shown in Figure 6：
Step 601：Server judges whether request is CHLO, whether is CHLO according to request, server may determine that one
Whether the Connection ID of a request are generated by server；
Step 602：If request is CHLO, illustrate that Connection ID are generated by client, to ensure Connection
ID needs to generate one in the unique Connection ID of server, according to server by server in the uniqueness of server
The union of effective control range of the corresponding virtual server node on annulus determines effective control range of server, according to
Effective control range of server maps out the range of choice of Connection ID, and server randomly chooses newly in the range
Connection ID；
Step 603：To ensure that Connection ID in the uniqueness of server, need to compare buffered
Connection ID need to regenerate if Connection ID have existed；
Step 604：For ensure Connection ID server uniqueness, what local cache had just generated
Connection ID, for searching later；
Step 605：The Connection ID new to client loopback, in this connection, the subsequent request of client must
The new Connection ID must be used；
Step 606：If request is not CHLO, illustrate that Connection ID are generated by server, requires to look up in caching
Connection ID, and determine corresponding session information, with the legitimacy of the certification Connection ID and after extended meeting
Words illustrate that the Connection ID are no longer valid if not having the Connection ID in caching, and refusal is needed to ask；
Step 607：After conversation end, need to be removed from the cache the Connection ID, it is ensured that Connection ID
It can be multiplexed by request later.
Therefore the present invention can realize the request quilt of the QUIC protocol clients with same Connection ID
Be dispatched on same server, it is ensured that QUIC agreements connection can client occur network switching when smooth migration.Together
When, the present invention has been dynamically determined Connection by the mapping relations of Connection ID and consistent Hash annulus range
The range of choice of ID improves QUIC agreement request dispatching efficiencies.Compared to by a certain or a few servers to cluster
Connection ID on middle Servers-all carry out global synchronization and the method for unified distribution, in the present invention,
Connection ID still are responsible for creating, cache and destroying by the server for specifically handling request, not will produce global synchronization
Huge network I/O expense caused by Connection ID, will not be all in entire server cluster because of focusing on
Performance pressures caused by Connection ID.Typically, a cluster internal Connection ID synchronizes caused net
The expense of network IO was about 500000 nanoseconds, and the local expense for generating Connection ID was only about for 100 nanoseconds, the opposite party
For case, Connection ID selection schemes of the invention can make the dispatching efficiency of request promote about 500 times.Compared to passing through
A static Connection ID range of choice and the service by specifically handling request are distributed to each server in cluster
The establishment of device progress Connection ID, the method for caching and destruction, in the present invention, the range of choice of Connection ID
It can be avoided caused by static range of choice to server cluster scalability according to the state dynamic change of server cluster
Constraint, when the scale enlargement of server cluster, virtual server number of nodes on annulus increases, and dynamically reduces annulus
Effective control range of upper existing virtual server node, thus the range of choice of the Connection ID of its corresponding server
Also dynamic reduces, conversely, when the scale down of server cluster, the virtual server number of nodes on annulus is reduced, dynamic
Increase effective control range of remaining virtual server node on annulus, thus the Connection ID of its corresponding server
Range of choice also dynamically increase.
Claims (7)
1. a kind of method improving QUIC agreement request dispatching efficiencies, which is characterized in that include the following steps：
It 1) will ranging from [0,232- 1] integer set constitutes a consistent Hash torus space, by the service in server cluster
Device is mapped in the form of several virtual server nodes on the annulus, and the position of each virtual server node corresponds to circle
An integral point in range represented by ring；
2) determine that the dynamic of virtual server node effectively controls model in the position relationship of annulus by virtual server node
It encloses；
3) determine that the dynamic of server effectively controls by the effective control range of dynamic of the virtual server corresponding to server
Range；
4) Connection ID are determined by the mapping relations of the effective control range of the dynamic of server and Connection ID
Dynamic select range；
5) for arbitrarily asking, request reflecting on annulus is determined by the reverse Mapping relationship of Connection ID to annulus
Position is penetrated, finds the server of specific processing request clockwise on annulus, it will be in request scheduling to the server.
2. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that the clothes
Business device cluster is the loose conglomerate that one or more server host is realized by network communication, and server cluster cooperation is completed
Evaluation work provides service to client jointly, and as viewed from the perspective of client, specific service effect is as only having separate unit master
Machine is in the service of offer.
3. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that the model
Enclose is [0,232- 1] space that integer set constitutes an annulus shape is specifically：The space of an annulus shape is set, it will
Ranging from [0,232- 1] uniform increments are distributed integer value in the direction of the clock on this annulus, for any point X+1 on annulus
Necessarily be point X in the clockwise direction closest to a point, and point X+2 at a distance from point X+1 certainty and point X+1 and X away from
From equal, point 232- 1 point closest clockwise is point 0, and puts 232- 1 is any one at a distance from point 0 and on annulus
Point X+1 is equal at a distance from X.
4. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that described logical
Virtual server node is crossed in the position relationship of annulus, determines the effective control range specific method of the dynamic of virtual server node
It is：[0,232- 1] integer set constitutes an annulus shape spatially, each virtual server node accounts for wherein
According to an integral point in range represented by an annulus, for arbitrary virtual server nodes X, if its corresponding integer value is
NX, using its position on annulus as starting point, counterclockwise moved along annulus, until encountering other any virtual server nodes
Until Y, if the corresponding integer value of its point is NY, then the effective control range of the dynamic of the virtual server nodes X is (NY,NX]。
5. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that described logical
The effective control range of dynamic for crossing the virtual server corresponding to server determines that the effective control range of the dynamic of server is specific
It is：For arbitrary server S, if its corresponding virtual server node on annulus is S1、S2、……、SX, wherein SNVirtually
The effective control range of dynamic of server node is RSN, then the effective control range of the dynamic of server S is RS1∪RS2∪……∪
RSX。
6. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that described logical
The effective control range of dynamic for crossing server and the mapping relations of Connection ID determine that the dynamic of Connection ID is selected
Selecting range is specifically：The Connection ID of QUIC are 64-bit length without symbol shaping value, value range be [0,
264- 1], for any point N on annulus, the value range for the Connection ID that can be mapped out is [N × 232,(N
+1)×232- 1], for any two points A, B and B on annulus>A, if the dynamic effective range of server S is [A, B], then it is corresponded to
Connection ID dynamic select ranging from [A × 232,(B+1)×232‐1]。
7. a kind of method improving QUIC agreement request dispatching efficiencies according to claim 1, which is characterized in that described logical
The reverse Mapping relationship for crossing Connection ID to annulus determines mapping position of the request on annulus, on annulus clockwise
Finding the server specific method of specific processing request is：For arbitrary Connection ID are the connection of C, in circle
Mapping point on ring is (C+1)/232It is that downward rounding obtains as a result, setting the result as N, then positions of the edge point N on annulus is suitable
Hour hands are searched, and the server corresponding to first encountered virtual server node is the service of specific processing request
Device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810306629.7A CN108769102A (en) | 2018-04-08 | 2018-04-08 | A method of improving QUIC agreement request dispatching efficiencies |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810306629.7A CN108769102A (en) | 2018-04-08 | 2018-04-08 | A method of improving QUIC agreement request dispatching efficiencies |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108769102A true CN108769102A (en) | 2018-11-06 |
Family
ID=63981158
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810306629.7A Pending CN108769102A (en) | 2018-04-08 | 2018-04-08 | A method of improving QUIC agreement request dispatching efficiencies |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108769102A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110177082A (en) * | 2019-04-25 | 2019-08-27 | 阿里巴巴集团控股有限公司 | A kind of data processing method, equipment, medium and device |
| CN111835682A (en) * | 2019-04-19 | 2020-10-27 | 上海哔哩哔哩科技有限公司 | Connection control method, system, device and computer readable storage medium |
| CN114390043A (en) * | 2021-12-29 | 2022-04-22 | 华南理工大学 | Method for achieving multi-path QUIC transmission optimization based on enterprise-level cloud storage system |
| US11381506B1 (en) * | 2020-03-27 | 2022-07-05 | Amazon Tehonlogies, Inc. | Adaptive load balancing for distributed systems |
| CN115225644A (en) * | 2022-07-27 | 2022-10-21 | 天翼云科技有限公司 | Load balancing system and data packet distribution method |
| US11570100B2 (en) | 2019-04-25 | 2023-01-31 | Advanced New Technologies Co., Ltd. | Data processing method, apparatus, medium and device |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101589581A (en) * | 2006-11-27 | 2009-11-25 | 艾利森电话股份有限公司 | A method and system for providing arouting architecture for overlay networks |
| US8712982B2 (en) * | 2011-03-08 | 2014-04-29 | Rackspace Us, Inc. | Virtual multi-cluster clouds |
| CN104050270A (en) * | 2014-06-23 | 2014-09-17 | 成都康赛信息技术有限公司 | Distributed storage method based on consistent Hash algorithm |
| CN104298541A (en) * | 2014-10-22 | 2015-01-21 | 浪潮(北京)电子信息产业有限公司 | Data distribution algorithm and data distribution device for cloud storage system |
| CN104391863A (en) * | 2014-10-23 | 2015-03-04 | 中国建设银行股份有限公司 | Data storage method and device |
| CN107197035A (en) * | 2017-06-21 | 2017-09-22 | 中国民航大学 | A kind of compatibility dynamic load balancing method based on uniformity hash algorithm |
-
2018
- 2018-04-08 CN CN201810306629.7A patent/CN108769102A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101589581A (en) * | 2006-11-27 | 2009-11-25 | 艾利森电话股份有限公司 | A method and system for providing arouting architecture for overlay networks |
| US8712982B2 (en) * | 2011-03-08 | 2014-04-29 | Rackspace Us, Inc. | Virtual multi-cluster clouds |
| CN104050270A (en) * | 2014-06-23 | 2014-09-17 | 成都康赛信息技术有限公司 | Distributed storage method based on consistent Hash algorithm |
| CN104298541A (en) * | 2014-10-22 | 2015-01-21 | 浪潮(北京)电子信息产业有限公司 | Data distribution algorithm and data distribution device for cloud storage system |
| CN104391863A (en) * | 2014-10-23 | 2015-03-04 | 中国建设银行股份有限公司 | Data storage method and device |
| CN107197035A (en) * | 2017-06-21 | 2017-09-22 | 中国民航大学 | A kind of compatibility dynamic load balancing method based on uniformity hash algorithm |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111835682A (en) * | 2019-04-19 | 2020-10-27 | 上海哔哩哔哩科技有限公司 | Connection control method, system, device and computer readable storage medium |
| CN111835682B (en) * | 2019-04-19 | 2021-05-11 | 上海哔哩哔哩科技有限公司 | Connection control method, system, device and computer readable storage medium |
| US11316930B2 (en) | 2019-04-19 | 2022-04-26 | Shanghai Bilibili Technology Co., Ltd. | Connection control method and device |
| CN110177082A (en) * | 2019-04-25 | 2019-08-27 | 阿里巴巴集团控股有限公司 | A kind of data processing method, equipment, medium and device |
| US11570100B2 (en) | 2019-04-25 | 2023-01-31 | Advanced New Technologies Co., Ltd. | Data processing method, apparatus, medium and device |
| US11381506B1 (en) * | 2020-03-27 | 2022-07-05 | Amazon Tehonlogies, Inc. | Adaptive load balancing for distributed systems |
| CN114390043A (en) * | 2021-12-29 | 2022-04-22 | 华南理工大学 | Method for achieving multi-path QUIC transmission optimization based on enterprise-level cloud storage system |
| CN115225644A (en) * | 2022-07-27 | 2022-10-21 | 天翼云科技有限公司 | Load balancing system and data packet distribution method |
| CN115225644B (en) * | 2022-07-27 | 2023-11-03 | 天翼云科技有限公司 | Load balancing system and data packet distribution method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108769102A (en) | A method of improving QUIC agreement request dispatching efficiencies | |
| US10887380B2 (en) | Multi-cluster ingress | |
| US10284383B2 (en) | Aggregation protocol | |
| KR101754408B1 (en) | Method and system for load balancing anycast data traffic | |
| CN113826363A (en) | Consistent routing advertisements between redundant controllers in a global network access point | |
| US20150146539A1 (en) | Flow distribution table for packet flow load balancing | |
| US7877515B2 (en) | Identity assignment for software components | |
| EP3542268A1 (en) | Live migration of load balanced virtual machines via traffic bypass | |
| US20190116123A1 (en) | Load Balancing of Connectionless Traffic | |
| WO2013090699A1 (en) | Content delivery network | |
| US11159625B1 (en) | Efficiently distributing connections to service instances that stream multi-tenant data | |
| US8930547B2 (en) | Techniques for achieving storage and network isolation in a cloud storage environment | |
| US20210051573A1 (en) | Inclusion of a message proxy in a service based architecture | |
| WO2018045992A1 (en) | Address management method and apparatus | |
| CN108345643A (en) | A kind of data processing method and device | |
| CN114070822A (en) | Kubernetes Overlay IP address management method | |
| US7965630B1 (en) | Load balancing port proxy for dynamically controlling routing of query requests | |
| US9760370B2 (en) | Load balancing using predictable state partitioning | |
| CN109120556A (en) | A kind of method and system of cloud host access object storage server | |
| US20150127799A1 (en) | Hierarchical distribution of control information in a massively scalable network server | |
| Xu et al. | An energy-aware method for multi-domain service function chaining | |
| CN115242882A (en) | Method and device for accessing k8s container environment based on transport layer route | |
| Baker et al. | Scalable web server design for distributed data management | |
| US10148766B2 (en) | Methods, systems, and computer readable media for subscriber binding repository reconfiguration | |
| US20220019485A1 (en) | Preserving eventually consistent distributed state of multi-layer applications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181106 |
|
| WD01 | Invention patent application deemed withdrawn after publication |