CN108388460A - Long-range real-time rendering platform construction method based on graphics cluster - Google Patents
Long-range real-time rendering platform construction method based on graphics cluster Download PDFInfo
- Publication number
- CN108388460A CN108388460A CN201810109865.XA CN201810109865A CN108388460A CN 108388460 A CN108388460 A CN 108388460A CN 201810109865 A CN201810109865 A CN 201810109865A CN 108388460 A CN108388460 A CN 108388460A
- Authority
- CN
- China
- Prior art keywords
- graphics
- cluster
- rendering
- graphics cluster
- node
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a kind of long-range real-time rendering platform construction method based on graphics cluster, the deployment of virtual desktop system is realized including the use of Docker frameworks inside graphics cluster;Embedded VNC server-sides establish Remote desk process after VNC clients access the VNC server-sides by VNC agreements in virtual desktop system;Graphics cluster carries out dynamic load leveling according to Remote desk process situation, establishes graphics cluster dynamic task scheduling and administrative mechanism;By graphics cluster dynamic task scheduling and administrative mechanism, realtime graphic transmission channel, long-range real-time rendering platform of the comprehensive integration based on graphics cluster are established between platform client and graphics cluster server-side.The present invention is directed to rendering application problem of a large amount of remote users based on different demands, according to the asynchronous distributed operating mode of facing multiple users, graphics cluster hardware environment is relied on, meets the needs of different user uses on demand using virtualization means, both management was facilitated to also ensure data safety.
Description
Technical field
The invention belongs to computer graphical Rendering field more particularly to a kind of long-range real-time wash with watercolours based on graphics cluster
Contaminate platform construction method, for providing a kind of graphics rendering contexts of long-range real-time high-efficiency to multi-user, contribute to Military Simulation,
The work of education and training, mobile office and the popularization of autonomous controllable system etc..
Background technology
Graphics-rendering system is the optimized integration of the technologies such as vision simulation, virtual reality and augmented reality.Figure renders solution
Certainly scheme develops to more from the initial dedicated graphics work station based on SGI (Silicon Graphics, Silicon Graphics of the U.S.)
The graphics cluster system of machine parallel connection, then arrive SLI (the Scaleable Link using NVIDIA (U.S. is tall and handsome to reach company)
Interface, scalable linkage interface) or AMD (Advanced Micro Devices, advanced micro devices company of the U.S.) CF
More video card graphics workstations of (CrossFire, video card are fought) technology, hardware technological development make figure render application model and be in
Reveal spiralling situation.In order to solve the problems, such as extensive, long-range, real-time rendering, and also to effectively integrate figure
Resource, improves resource utilization, and the render mode based on graphics cluster becomes the focus of industry research again.
Under normal circumstances, Distributed Rendering Environment environment or rendering are also referred to as with the matching used Cluster Rendering system of graphics cluster
Farm, it refers to being made of the computer of many operation Rendering softwares, under the unified allocation of resources of Cluster Rendering management software, is coordinated
Distributed rendering task is completed in work in a parallel fashion.It is broadly divided into two classes from the angle of real-time:One kind is non real-time
Rendering system, i.e. management end issue render instruction to graphics cluster, after graphics cluster completes rendering task within the regular hour
Render scenes are sent to management end or storage system, then is called by subscriber checking and renders achievement.This rendering system is chiefly used in
The fields such as video display, animation making;Another kind of is real-time rendering system, i.e. management end issues render instruction, figure to graphics cluster
By rendering result Real-time Feedback to user after cluster completion rendering task, while responding interaction and the control instruction of user.It is this
Rendering system is chiefly used in vision simulation and virtual reality system.
By the way that the relevant producer of current remote real-time rendering platform is investigated with technology, its core skill covered is found
Art includes mainly VDI (Virtual Desktop Infrastructure, virtual desktop framework), VMM (Virtual
Machine Monitor, virtual machine monitor), DCV (Desktop Cloud Visualization, desktop cloud visualization),
VGPU (Virtual Graphics Processing Unit, virtual pattern processor) etc., towards applicable cases it is main
It is embodied in following three aspects.
1, single machine accesses single station and obtains virtual desktop serve
User utilizes RDP in locally-installed client software, the graphics workstation node installation service software of distal end
(Remote Desktop Protocol, Remote Desktop Protocol) virtually arrives graphics workstation desktop locally, realizes soft to figure
The remote access and display of part.
In fact, the RDP agreements based on Windows can effectively support the file operation of remote desktop and simple text
Shelves processing etc., but it is not appropriate for the processing to video and dynamic image.On this basis, Hewlett-Packard proposes RGS
(Remote Graphics Software, remote image software) solution, the program have been done very on the basis of RDP agreements
More optimizations, including use to have from the image compression/decompression technology of dominant patent and (can reach 170:1 compression ratio), subtract significantly
Small network transport load can realize the function of remote multi-subscriber share desktop service additionally by the extension to RDP
(the Remote desk process same time of Windows only allows a user to access, and server-side becomes screen lock state after access;
The RGS of Hewlett-Packard allows multiple users while accessing the desktop of server-side, and the desktop of server-side can also normally be shown).It is this
Scheme supports Windows, Linux and Unix operating system, substantially disclosure satisfy that the demand of remote rendering task, but the clothes
Business end software requirement is necessarily mounted on the graphics workstation of Hewlett-Packard, and does not do it in the distribution of graphics rendering resources
His optimization processing work.
2, multimachine accesses single station and obtains virtual desktop serve
Above-mentioned project plan comparison is suitable for multi-person synergy and carries out same work, cannot meet more people by sharing figure
Resource carries out the demand of different operating, if to realize that each user can obtain independent virtual desktop and can share
Graphic resource, then need that be each user on graphical Work tiny node can hew out independent desktop system, i.e. VDI frameworks
Deployment and operation.Currently, the VDI technologies of mainstream include mainly Hyper-V+RGS, XenServer+HDX and NICE DCV tri-
Kind solution, comparative analysis are as follows:
(1)Hyper-V+RGS
Hyper-V (virtualization product based on Windows that MS releases) is built in Windows
In Server 2008 and later operating system product, relative to a layer operation system more than the past virtualization technology of Microsoft,
Good compatibility, but speed is slower.Hyper-V allow virtual machine can the relatively straightforward hardware resource using entity host, with improve
The efficiency of virtual system.Hyper-V borrows Microsoft platform advantage and is cut into server virtualization field, but for Linux systems
On system is supported, Hyper-V is also insufficient.The function of RGS softwares and deficiency are as previously described.
(2)XenServer+HDX
XenServer (the virtualization software product that Si Jie companies of the U.S. release) is service that is a kind of comprehensive and being easily managed
Device virtual platform can efficiently manage Windows and Linux virtual server, it is possible to provide the Server Consolidation of economical and efficient
And business continuance, XenServer have the function of operating system, can be mounted directly and be booted up and transport on the server
Row.The HDX (High Definition Experience, high definition usage experience) of Si Jie companies of the U.S. is then virtual for desktop
Change the long-range high definition Interactive Product released with application virtualization market, can realize and multimedia, voice, video and 3D figures are handed over
Mutually.But in engineering design field, which can only realize that the desktop virtualization based on windows platform interacts, and can not be real
Now GPU resource is shared.
(3)KVM+NICE DCV
NICE DCV (the DCV products that NICE companies under Amazon Company of the U.S. release) be it is a kind of can be in standard network
The Interactive Product of upper remote access 2D or 3D applications, makes 3D High-end graphics card, quick I/O of the user using distal end
Energy and a large amount of memory node.It is emphasized that by KVM (Kernel-based Virtual Machine, base
In the virtual machine of kernel) NICE DCV are disposed on virtual platform, the performance of the software can be given full play to.The program can be supported
(SuSE) Linux OS, and NICE DCV support various GPU vitualization technologies on hardware view, utilize NVIDIA companies
Dedicated graphics card coordinates vGPU to drive journey, distributes vGPU to each virtual machine, can realize being total to for GPU resource to a certain extent
It enjoys.
3, multimachine accesses graphics cluster and obtains virtual desktop serve
The virtual desktop serve mentioned in the two schemes of front, realization is empty on the basis of being all built upon separate unit graphics workstation
Quasi- desktop services, if number of users is excessive, the form that virtual machine is created on separate unit graphics workstation just cannot be satisfied
Application demand, therefore just need to replace separate unit graphics workstation in the form of more graphics workstations, here it is graphics clusters
Concept.The third scheme is mainly built upon on the basis of first two scheme, and emphasis solves the aggregation problem of graphics cluster, i.e.,
More current graphics workstations how are organized, enable them to the realization of the convenience and high-efficiency as a machine to multi-purpose
The graphical services at family.
If the polymerization of independent more graphics workstations is become unified graphics cluster system, it is necessary first to there is rendering
Task management software carries out resource allocation and scheduling, it is typical as VCM (Visualization Cluster Manager, visually
Change cluster manager dual system) and TechViz (the visual software product of French TechViz companies) etc..
(1)VCM
The VCM that Orad companies of Israel release is the cluster management software installed in graphics cluster management node, it is
Rendering task is distributed by planning the connection type of graphical nodes, special integrated interface is recycled to realize the integrated and same of image
Step.This cluster management mode preferably realizes the distribution and management of rendering task, but its theory is mainly aobvious for multichannel
Show that system provides service, is relatively suitble to the large-size screen monitors of Multi-path synchronous output to show, in the apparent underaction of application aspect and intelligence.
(2)TechViz
TechViz uses another thinking in terms of cluster management, is installed in the management node of cluster complete
Figure application software, after figure application software is run be originally to directly invoke local graphics card and carry out rendering to show, but
It is that by high-resolution figure rendering task, with OpenGL, (Open Graphics Library open figure to TechViz softwares
Library) form of instruction stream is assigned to by network on rendering node, and recycle the graphics card of rendering node to carry out rendering output, respectively
Output synchronization is carried out by sync card between rendering node.This cluster management scheme need not install figure on rendering node
Application software, and it is also flexibly many in the distribution of rendering task.
(3) other softwares
Other than above two software, there are some to be directed to the cluster management software that figure renders both at home and abroad, such as
Muster (the Cluster Rendering software of Virtual Vertex companies of the U.S.), Platform LSF (the cluster pipes of IBM Corporation of the U.S.
Manage dispatcher software) and EnFuzion (the Cluster Rendering software of Chinese TurboLinux Inc.) etc., they can real-time management and monitoring wash with watercolours
The working condition for contaminating node, avoids the free time of rendering node, user is made to obtain final rendering product in the shortest time.
Graphics cluster system provides good hardware foundation in terms of providing long-range real-time rendering service for multi-user, and
Cluster Rendering software systems research at this stage lays particular emphasis on aggregated structure, cluster management, load balancing and parallel rendering etc.
Aspect, though solution more rapidly, more efficient can be provided to the user under simultaneously and concurrently pattern, and it is unresolved a large amount of remote
Rendering application problem of the journey user based on different demands.
Different from traditional Cluster Rendering system, technology proposed by the present invention mainly uses the different of facing multiple users
Distributed work pattern is walked, graphics cluster hardware environment is relied on, is used on demand using virtualization means to meet different user
Demand.Its core concept is that user remotely accesses graphics cluster, and graphics cluster provides the running environment of real-time high-efficiency to the user, together
When response user in local interactive operation, so that user is obtained a kind of work close to graphics workstation and experience, data resource
It is all deployed in distal end with calculation procedure, both management was facilitated to also ensure data safety.
Invention content
In order to solve the above technical problem, the present invention provides a kind of long-range real-time rendering platform structure based on graphics cluster
Construction method is disposed by the virtual desktop system based on Docker frameworks, and VDI is disposed and run on graphics cluster, is user
Independent virtual desktop system is opened up, display driving is directly invoked under Docker running environment to realize GPU (Graphics
Processing Unit, graphics processor) accelerate to render, while Remote desk process is established using VNC agreements, it is built for user
Real-time image transmitting channel is stood, in conjunction with graphics cluster dynamic task scheduling and administrative mechanism, is provided for multi-user a kind of asynchronous
Distributed real-time rendering service.
The purpose of the present invention implements by the following technical programs:
Long-range real-time rendering platform construction method based on graphics cluster, including:
Step 1: the deployment of virtual desktop system is realized inside graphics cluster using Docker frameworks;
Step 2: embedded VNC server-sides, VNC clients access the VNC server-sides in the virtual desktop system
Afterwards, Remote desk process is established by VNC agreements;
Step 3: graphics cluster carries out dynamic load leveling according to Remote desk process situation, graphics cluster dynamic is established
Task scheduling and administrative mechanism;
Step 4: by graphics cluster dynamic task scheduling and administrative mechanism, in platform client and graphics cluster service
Realtime graphic transmission channel, long-range real-time rendering platform of the comprehensive integration based on graphics cluster are established between end.
Step 1 specifically includes:
Docker containers are created in graphics cluster rendering node operating system;
Virtual desktop system after encapsulation is made in Docker mirror images deposit shared memory systems;
Docker containers realize the deployment of virtual desktop system, together by loading Docker mirror images inside graphics cluster
When realize that GPU based on Docker frameworks is rendered in virtual desktop system and accelerate.
Wherein, described to realize that the GPU based on Docker frameworks renders the step of accelerating in virtual desktop system, including:
The video card of graphics cluster rendering node operating system is driven into the collection cyclization that journey kernel file is mapped to Docker containers
In border, while the Window Scale plug-in unit of OpenGL (Open Graphics Library, open graphic library) is called, realized in void
GPU (Graphics Processing Unit, graphics processor) real-time rendering in quasi- desktop system accelerates.
Wherein, by the way of virtual external output, which includes for the display of the virtual desktop system:
For the virtual display output of each Docker container allocation, when user accesses the Docker containers, Docker
Figure in container is exported by virtual display, output to client so that GPU renders defeated independent of actual graphics card
Outgoing interface.
Step 2 specifically includes:
VNC server-sides screen changed image-region using the regional change detection algorithm of divided-fit surface,
It is sent to corresponding VNC clients after compression of images, the Remote desk process based on C/S frameworks is established by VNC agreements;
Wherein, the VNC server-sides utilize the regional change detection algorithm of divided-fit surface to changed image-region
It is screened, including:
Setting needs to change the unit area of detection;
Obtain the information that interception system screen redraws region;
If the system screen intercepted and captured, which redraws region, is less than unit area, the zone position information is sent;
If the system screen intercepted and captured, which redraws region, is more than unit area, interception system screen is redrawn into region and is split,
Subregional coordinate information will be torn open to store into chained list, to each tearing open in chained list after subregional coordinate information carries out traversal detection
Obtain the region of variation location information for needing to retransmit, it would be desirable to which the location information of the region of variation of retransmission is stored
And it sends.
In step 3, the graphics cluster dynamic task scheduling includes with administrative mechanism:
Under certain load balancing, service money is provided to the user by cluster management node reasonable arrangement rendering node
Source;
The graphics cluster dynamic task management is related to including but not limited to single user to single node, multi-user with scheduling
Single node, single user dynamically turn multinode and/or multi-user between one or more situations in four kinds of situations of multinode
It changes.
The load balancing includes:
In single user to single node, rendering node is distributed using polling schemas;
The case where multi-user is to multinode, resource utilization, network bandwidth and the I/O of rendering node are considered
The indexs such as (Input/Output, input/output) rate, using dynamic load leveling model solution rendering system load variance or
Standard deviation is the objective programming problem of minimum, the graphics cluster load balancing after being optimized.
Wherein, the long-range real-time rendering platform based on graphics cluster of the comprehensive integration includes platform client, figure
Four cluster management node, graphics cluster rendering node and shared memory systems parts;
Platform client is deployed in user's operation terminal, for initiating application request to graphics cluster server-side;
Graphics cluster management node and graphics cluster rendering node are deployed in graphics cluster server-side, for response platform visitor
Initiate application request in family end;
Shared memory systems are used for storage management user profile, virtual desktop system mirror image and/or rending model number
It is believed that breath.
Step 4 specifically includes:
VNC clients are embedded into platform client, Docker containers and VNC server-sides are embedded into graphics cluster and render section
Point, Docker mirrored storages are in shared memory systems, by graphics cluster management node according to dynamic task scheduling and supervisor
System carries out cluster load balance;
When receiving platform client initiation application request, graphics cluster management node, which is that platform client is specified, renders section
Point, rendering node start Docker containers and load Docker mirror images, while establishing long-range real-time rendering using VNC agreements and leading to
Road provides long-range real-time rendering service to the user.
Wherein, the application model of the long-range real-time rendering platform includes:
Graphics cluster server-side is after the application request for being connected to platform client initiation, by graphics cluster management node by wash with watercolours
Dye task is distributed to rendering node, and rendering node provides rendering service to the user by creating independent virtual desktop system.
The present invention is directed to rendering application problem of a large amount of remote users based on different demands, relies on graphics cluster hardware loop
Border using providing the running environment of real-time high-efficiency to the user based on the virtualization means of Docker frameworks, while responding user
In local interactive operation.This method is applicable not only to different hardware environment and operating system platform, and can realize more
Users to share GPU rendering resources make user obtain a kind of work experience close to graphics workstation.
Description of the drawings
Fig. 1 is the long-range real-time rendering platform construction method flow chart based on graphics cluster.
Fig. 2 is virtual machine and the technology mechanism comparison diagram of Docker.
Fig. 3 is the long-range real-time rendering platform overall framework figure based on graphics cluster.
Fig. 4 is the long-range real-time rendering platform feature module design figure based on graphics cluster.
Fig. 5 is the application flow exemplary plot that user uses long-range real-time rendering platform.
Fig. 6 a are host operating system rendering effect figures;Fig. 6 b are Docker container rendering effect figures.
Fig. 7 a are that single node supports two users' schematic diagram;Fig. 7 b are that single node supports four user's schematic diagrames.
Specific implementation mode
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The every other embodiment that member is obtained without making creative work should all belong to the model that the present invention protects
It encloses.
Referring to Fig.1, Fig. 1 shows a kind of long-range real-time rendering platform construction side based on graphics cluster provided by the invention
The flow chart of one embodiment of method, this method include:Step 1 is to step 4.
In step 1, the deployment of virtual desktop system is realized inside graphics cluster using Docker frameworks.
First, VDI is disposed and run on graphics cluster, and independent virtual desktop system is opened up for user, it can be preferable
Meet the application demand of different user.Restriction in view of virtual machine operational mode to GPU rendering capabilities, system are used and are based on
The VDI solutions of Docker (application container engine) directly invoke display driving under Docker running environment, greatly promote
GPU rendering efficiencies.
The virtual desktop system for being generally basede on virtual machine is all aobvious by calling the display driving of underlying Operating System to realize
Show function, itself display resource is limited and tupe belongs to work in series pattern more, is unable to give full play graphics cluster height
Hold the GPU rendering capabilities of video card.VDI solutions based on Docker penetrate underlying Operating System using GPU vitualization technology
Display driving, realize that the figures based on GPU accelerate in multiple independent virtual machine internals, it is more to realize to a certain extent
User shares GPU resource.With reference to figure 2, Fig. 2 is the technology mechanism contrast schematic diagram of virtual machine and Docker.
The step specifically includes:Docker containers are created in graphics cluster rendering node operating system;After encapsulation
Virtual desktop system is made in Docker mirror images deposit shared memory systems;Docker containers by load Docker mirror images,
The deployment of virtual desktop system is realized inside graphics cluster, while being realized based on Docker frameworks in virtual desktop system
GPU, which is rendered, to be accelerated.
Wherein, Docker is that the advanced container based on LXC (Linux Container, linux container) is increased income engine, main
To include three kinds of components, i.e. finger daemon (Docker daemon), mirror image (Image) and container (Container), wherein keep
Shield process receives the request from client as server-side and handles, read-only template when mirror image is container operation, container
Including the required the integration environment of user's application operation.
The step of accelerating is rendered to specifically include:The video card of graphics cluster rendering node operating system is driven journey kernel file to reflect
It is mapped in the integration environment of Docker containers, while calling OpenGL's (Open Graphics Library, open graphic library)
Window Scale plug-in unit realizes that the GPU (Graphics Processing Unit, graphics processor) in virtual desktop system is real
When render accelerate.
Due to Docker containers and host operating system sharing operation system kernel, user's application can directly invoke host behaviour
Make the hardware resource in system.But in actual application, user runs application in the integration environment that container directly provides
Effective GPU is cannot achieve to render.By a large amount of test experiments, the video card of host operating system is driven into the mapping of journey kernel file
Into the integration environment of container, while the X Window expansion plugin of OpenGL is called, can preferably realize that the rendering of GPU accelerates.
Further, by the way of virtual external output, which includes for the display of the virtual desktop system:
For the virtual display output of each Docker container allocation, when user accesses the Docker containers, Docker
Figure in container is exported by virtual display, output to client so that GPU renders defeated independent of actual graphics card
Outgoing interface.
Under normal conditions, GPU rendering pipelines carry out the online rendering of screen, and the asynchronous access of multi-user can not be real
Existing Screen sharing.It is that the virtual display of each container allocation is defeated by the way of virtual external output for this problem
Go out, when user accesses the container, figure exports display to client so that GPU renders defeated independent of actual graphics card
Outgoing interface, the problem of preferably resolving multiuser asynchronous application.
In step 2, embedded VNC (Virtual Network Console, virtual net in the virtual desktop system
Network console) server-side establishes Remote desk process after VNC clients access the VNC server-sides by VNC agreements.
VNC server-sides screen changed image-region using the regional change detection algorithm of divided-fit surface,
It is sent to corresponding VNC clients after compression of images, the Remote desk process based on C/S frameworks is established by VNC agreements.
Wherein, the VNC server-sides utilize the regional change detection algorithm of divided-fit surface to changed image-region
It is screened, including:
Setting needs to change the unit area of detection;
Obtain the information that interception system screen redraws region;
If the system screen intercepted and captured, which redraws region, is less than unit area, the zone position information is sent;
If the system screen intercepted and captured, which redraws region, is more than unit area, interception system screen is redrawn into region and is split,
Subregional coordinate information will be torn open to store into chained list, to each tearing open in chained list after subregional coordinate information carries out traversal detection
Obtain the region of variation location information for needing to retransmit, it would be desirable to which the location information of the region of variation of retransmission is stored
And it sends.
Wherein, which is specially:VNC mechanism redraws region letter by using message hook mechanism interception system screen
Breath.Setting needs to change the unit area of detection, such as remembers that unit area rgn sizes are 32 × 32, if the system screen weight intercepted and captured
It paints region and is less than unit area, then directly transmit the zone position information;If interception system screen, which redraws region, is more than unit area
Domain will then cut a system screen and redraw and obtain region and split, and obtain a rectangle chained list (traversal chained list) and split region to store
Coordinate information is denoted as array_list, defines a region of variation and is denoted as vnc_rgn, special to preserve the area for needing to retransmit
Location information domain calls this detection algorithm to be changed region detection in each member's rectangles of array_list.
One preferred embodiment:
(1) assume that one of member's rectangle is arri, if the length and width of arri are both less than 32 × 32, directly as need
The region of variation of retransmission, is saved in vnc_rgn, otherwise carries out analysis detection to arri.
(2) a new rectangle arrj (arri.left, 0, arri.right, 0) is defined first, and hair is found in arri
The row for changing records the y-coordinate of the row, remembers arrj.top=y, while y+=16 incrementally finds the row that do not change, temporarily
Remember arrj.bottom=y;Then from the bottoms arrj, y-=1 is compared, and the changed row of content is found, by this after finding
Capable y-coordinate is arranged to arrj.bottom=y.So far final top, bottom coordinate of variation rectangle arrj is determined.Together
When the value of arri is set as arri (arri.left, arrj.bottom, arri.right, arri.bottom).To arrj points
Analysis finishes followed by the analysis of arri traversals until terminating.
(3) after determining rectangle arrj, rectangle arrj is analyzed.A rectangle arrn (0,0,0,0) is defined first,
The row to change at first are found in arrj, and left, top coordinate in changed region, arrn.left=are recorded after finding
X, arrn.top=y are analyzed to x, y-axis direction searching loop respectively and (are denoted as x+=32, y+=32), if this row screen
Pixel data changes, then x+=32 continues to compare;Otherwise arrn.right=x will be recorded, y-axis direction will similarly be compared
Compared with, y+=32, and arrn.bottom=y is recorded, finally obtain the region of variation rectangular coordinates for needing to send, and by this rectangle
Be saved in the rectangle chained list finally sent, at the same record arrk (arrn.right, arrn.top, arrj.right,
Arrn.bottom), arrm (arrn.left, arrn.bottom, arrj.right, arrj.bottom), respectively by arrk,
Arrm is assigned to arrj and carries out similar analysis detection, successively the entire arrj of searching loop, until terminating.
By this screen change Region detection algorithms, after being traversed to member's rectangle of entire array_list chained lists
The region of variation coordinate information retransmitted to needs.
In addition, the real-time Transmission for image controls, system uses C/S (Client/Server, client/server)
Framework establishes image transmitting channel between client and server-side, is sent to client after server-side acquisition, compression image, visitor
Family end receives, is shown to user after decompressed image.Image Real-time Transmission uses VNC agreements, server-side to utilize the area of divided-fit surface
Domain change detection algorithm screens changed image-region, and client is sent to after compression of images, greatly reduces
Network transmission data amount.Client image refreshing speed is by user's autonomous control, and server-side is in the refreshing for receiving client
Image transmitting is carried out after request, when client is larger compared with busy or network delay, image refreshing speed can reduce accordingly.
The method for enhancing image transmitting real-time includes mainly two aspects, first, the effect of lifting region change detection algorithm
Rate, second is that using compression ratio higher and scramble time shorter image compression algorithm.To the regional change detection algorithm of VNC into
It has gone improvement, image change region has been finely divided by interlacing scan, further obviate redundancy does not change image data.
System is tested the algorithm, and by taking 1280 × 800 24 screen pictures as an example, client screen refresh interval is
40ms can averagely be reduced in the upper data traffic volume of the application such as word processing, web page browsing in 200kb/s or so using the algorithm
About 10% data volume;Using upper data traffic volume it is more than 4Mb/s in high-dynamics image, data volume is influenced very using the algorithm
It is small, and CPU usage is obviously improved, so the real-time influence of method a pair of high-dynamics image transmission is weaker.Consider
Compression ratio and coding the time it takes, system is frequently with a kind of more efficient MJPEG (image compression algorithm) algorithms to figure
As carrying out compressed encoding, the experimental results showed that, Zlib (one can be down to using MJPEG algorithm datas transmitted traffic under the same terms
The image compression algorithm of kind of VNC agreements acquiescence) algorithm 50% hereinafter, can avoid to a certain extent since data volume is excessive
Network congestion problem caused by and.
In step 3, graphics cluster carries out dynamic load leveling according to Remote desk process situation, establishes graphics cluster
Dynamic task scheduling and administrative mechanism.
The graphics cluster dynamic task scheduling includes with administrative mechanism:
Under certain load balancing, service money is provided to the user by cluster management node reasonable arrangement rendering node
Source;The graphics cluster dynamic task management and scheduling be related to including but not limited to single user to single node, multi-user to list
Node, the single user dynamic translation between one or more situations in four kinds of situations of multinode to multinode and/or multi-user.
The load balancing includes:
In single user to single node, rendering node is distributed using polling schemas;In multi-user to the feelings of multinode
Condition considers resource utilization, network bandwidth and I/O (Input/Output, input/output) rate of rendering node etc.
Index loads variance or standard deviation using dynamic load leveling model solution rendering system and is asked for the objective programming of minimum
Topic, the graphics cluster load balancing after being optimized.
The task scheduling of long-range real-time rendering system is under certain load balancing, by cluster pipe with administrative mechanism
Reason node reasonable arrangement rendering node provides Service Source to the user.Under normal circumstances, cluster management is related to single user to list
Node, multi-user to single node, single user to multinode and multi-user to four kinds of situations of multinode, and in system operation
The dynamic translation problem of a variety of situations may be there is also.
In single user to single node, polling schemas distribution rendering node can be used, and it is more in practical applications
Be the case where multi-user is to multinode.In the case of number of users and certain rendering node number, rendering node is considered
The indexs such as resource utilization, network bandwidth and I/O rates, load balancing can be summarized as doing in the selection of rendering node
Decision solves rendering system load variance or standard deviation as the objective programming problem of minimum;If number of users renders section
Points change, then solution procedure may develop as complicated dynamic programming problems, but current within a period
Moment still can treat as objective programming problem.
If EiFor the load value of i-th rendering node, then:
Ei=w1Ci+w2Mi+w3Gi+w4Ni+w5IOi (1)
In formula, Ci、Mi、GiThe respectively occupancy of i-th rendering node CPU, memory and GPU, NiFor network utilization,
IOiFor I/O rates, wj(j=1,2 ..., 5) is the weighted value of corresponding above-mentioned 5 indexs, each index item and load value EiIt is both needed to do
Normalized.
The total load mean value of rendering system is:
Rendering system load variance will be solved and be reduced to solve the minimum of each rendering node load and total load mean value difference
Value, i.e.,:
After load balancing, the load value of each rendering node can generally be stablized above and below system total load mean value.To keep away
Exempt from due to the rendering experience that is excessive and influencing active user that Adds User, management node is monitoring rendering node active user's
When rendering frame frequency less than setting value (being defaulted as 24 frames/second), preferentially distribution can be asked to be rendered to other user according to rule and saved
The access request of point or refusal user.
In step 4, by graphics cluster dynamic task scheduling and administrative mechanism, in platform client and graphics cluster
Realtime graphic transmission channel, long-range real-time rendering platform of the comprehensive integration based on graphics cluster are established between server-side.
As shown in figure 3, the step specifically includes:
When receiving platform client initiation application request, VNC clients are embedded into platform client (in Fig. 3 referred to as
Client n, n >=1), Docker containers (abbreviation container n, n >=1 in Fig. 3) and VNC server-sides are embedded into graphics cluster and render section
Point (abbreviation node in Fig. 3), Docker mirror images (abbreviation mirror image in Fig. 3) are stored in shared memory systems, by graphics cluster pipe
It manages node and carries out cluster load balance according to dynamic task scheduling and administrative mechanism;
Graphics cluster management node is that platform client specifies rendering node, rendering node to start Docker containers and load
Docker mirror images, while long-range real-time rendering channel is established using VNC agreements, comprehensive integration is long-range real-time based on graphics cluster
Rendering platform.
The long-range real-time rendering platform based on graphics cluster of the comprehensive integration includes platform client, graphics cluster pipe
Manage four node, graphics cluster rendering node and shared memory systems parts;Platform client is deployed in user's operation terminal, uses
In to graphics cluster server-side initiate application request;Graphics cluster management node and graphics cluster rendering node are deployed in graphical-set
Group's server-side initiates application request for response platform client;Shared memory systems, for storage management user configuration text
Part, virtual desktop system mirror image and/or rending model data information.
The application model of the long-range real-time rendering platform includes:Graphics cluster server-side is being connected to platform client initiation
Application request after, rendering task is distributed to rendering node by graphics cluster management node, rendering node pass through create it is independent
Virtual desktop system provide to the user rendering service.
One embodiment, long-range real-time rendering system are realized that virtual desktop system is disposed using Docker frameworks, are built using VNC
Vertical Remote desk process.Remote user sends out the request of Docker Container Managements to cluster service end, utilizes VNC clients and VNC
Server-side establishes Remote desk process.In graphics cluster server-side, cluster management node is by Docker finger daemons to container
It is managed scheduling, Docker registration tablies are for preserving user-defined image file.It is sent out when server-side receives client
Container Management request after, the mirror image searched in registration table by Docker finger daemons simultaneously creates user's container, while in container
Middle startup VNC server-sides.
Graphics cluster is distributed rendering task to rendering by management node after the application request for being connected to user terminal initiation
Node, rendering node provide rendering service to the user by creating independent virtual desktop system.Based on the long-range of graphics cluster
Real-time rendering platform includes mainly platform client, graphics cluster management node, graphics cluster rendering node and shared storage system
It unites four parts, wherein platform client is deployed in user's operation terminal, and graphics cluster management node and rendering node are deployed in figure
Shape cluster service end, shared memory systems are mainly user profile, virtual desktop system mirror image and rending model data
Deng the convenient and reliable storage management of offer.Platform feature module is as shown in Figure 4.
Platform client sends virtual desktop request by user's registration/login mode to graphics cluster management node, waits for
Management node is responsible for establishing Remote desk process with the rendering node for providing the service of rendering after specifying rendering node.Platform client
Shared memory systems can be directly accessed, specific configuration file is uploaded, downloaded or changed within the scope of user right.
Graphics cluster management node includes mainly 3 initialization, task management and node administration modules.System is first after starting
First complete the function of initializing such as cluster internal connection and configuration file load;Task management module mainly complete task storage,
The functions such as scheduling and inquiry, after the virtual desktop request for being connected to platform client, to the load feelings of each rendering node in cluster
Condition is inquired, while being that platform client specifies rendering node using Dynamic Load-balancing Algorithm, if new user's registration mould
Then rendering node newly opens up virtual desktop system to formula, and if user's login mode, then rendering node is loaded from shared memory systems
Virtual desktop system file;Node administration mainly completes the functions such as additions and deletions, configuration and status inquiry to rendering node.
Graphics cluster rendering node is established after starting virtual desktop system by remote desktop management and platform client
Independent data connection channel, provides remote rendering service to the user.Rendering node mainly include virtual desktop system management with
Remote desktop system manages two parts, the former is responsible for creating and manages virtual desktop system, provides real-time rendering service to the user,
The latter is mainly used for remotely display and interactive controlling.
Virtual desktop system management module mainly creates Docker containers on (SuSE) Linux OS, after encapsulation
Virtual desktop system is made in Docker mirror images deposit shared memory systems, passes through load needing the when of creating virtual desktop system
Docker mirror images are solved using the GPU Renderings based on Docker in virtual desktop system to realize VDI deployment
Real-time rendering problem.
Remote desktop management module mainly uses VNC solutions, since VNC is to use message hook mechanism interception system
Screen redraws area information, thus remote desktop management module need for each user virtual desktop system create display it is defeated
Outgoing interface maps, that is, creates virtual " screen ".VNC server-sides are installed in virtual desktop system, platform client passes through calling
VNC clients establish Remote desk process with virtual desktop system, using the Image Real-time Transmission control method based on VNC to passing
Processing is optimized in defeated real-time.User is using the application flow example of long-range real-time rendering platform as shown in figure 5, including:
Platform client initiates the connection request, and graphics cluster management node receives connection request, graphics cluster management node
Rendering node is selected for user, new user's connection is determined whether, is then to start application container engine, creates virtual desktop system
System;It is no, then start application container engine, loads virtual desktop system mirror image;The response user's connection of graphics cluster rendering node is asked
It asks, graphics cluster rendering node creates VNC real-time renderings channel, and platform client completes Remote desk process, and user starts system
System application.
To test system application effect, two graphics workstations of selection form graphics clusters, and two are rendering node,
In one be used as management node, platform client be mounted on common PC on, network environment is gigabit LAN.Wherein graphics cluster
Node hardware is configured to Intel to strong four core processor, 16G memories, Nvidia Q7000 professional graphic cards (video memory 6G), behaviour
It is 64 Ubuntu Kylin 16.04 (Wu Bantu kylins operating system) to make system.Platform client log in after with graphics cluster
Rendering node establishes Remote desk process, carries out GPU rendering efficiencies respectively in 1280 × 800 resolution ratio and multiuser asynchronous answers
With experiment.
GPU rendering efficiencies are tested:Identical pattern is separately operable in host operating system and Docker container environment to survey
Sample example runs frame frequency due to being limited by screen refresh frequency, in operational effect host operating system as shown in Figure 6 a and does not surpass
60 frames/second is spent, for the operation frame frequency under Docker container environment up to 85 frames/second, operational effect is as shown in Figure 6 b.Test result is aobvious
Show, the GPU rendering efficiencies based on Docker are substantially close to the service efficiency in host operating system.
Multiuser asynchronous application experiment:After multi-platform client starts simultaneously, frame is rendered in system stable operation
Frequency gradually declines with the increase of platform client quantity, and the rendering frame frequency difference between each platform client is little, platform client
End display is without apparent Caton phenomenon.When single rendering node supports 1 user, average frame frequency supports 2 use up to 140 frames/second
About 120 frames/second when family, platform client display effect is as shown in Figure 7a, then or so near 75 frames/second when supporting 4 users,
Platform client display effect is as shown in Figure 7b.
To sum up, the embodiment of the present invention be directed to rendering application problem of a large amount of remote users based on different demands, according to towards
The asynchronous distributed operating mode of multi-user relies on graphics cluster hardware environment, meets different user using virtualization means
The demand used on demand.Its core concept is that user remotely accesses graphics cluster, and graphics cluster provides real-time high-efficiency to the user
Running environment, while user is responded in local interactive operation, so that user is obtained a kind of work experience close to graphics workstation,
Its data resource and calculation procedure are all deployed in distal end, both management are facilitated to also ensure data safety.
Claims (10)
1. the long-range real-time rendering platform construction method based on graphics cluster, including:
Step 1: the deployment of virtual desktop system is realized inside graphics cluster using the Docker frameworks;
Step 2: the embedded VNC server-sides in the virtual desktop system, after VNC clients access the VNC server-sides, lead to
It crosses VNC agreements and establishes Remote desk process;
Step 3: graphics cluster carries out dynamic load leveling according to Remote desk process situation, graphics cluster dynamic task is established
Scheduling and administrative mechanism;
Step 4: by graphics cluster dynamic task scheduling and administrative mechanism, between platform client and graphics cluster server-side
Establish realtime graphic transmission channel, long-range real-time rendering platform of the comprehensive integration based on graphics cluster.
2. the long-range real-time rendering platform construction method based on graphics cluster as described in claim 1, which is characterized in that step
One, it specifically includes:
Docker containers are created in graphics cluster rendering node operating system;
Virtual desktop system after encapsulation is made in Docker mirror images deposit shared memory systems;
Docker containers are by loading Docker mirror images, the deployment of realization virtual desktop system inside graphics cluster, while
It realizes that the GPU based on Docker frameworks is rendered in virtual desktop system to accelerate.
3. the long-range real-time rendering platform construction method based on graphics cluster as claimed in claim 1 or 2, which is characterized in that
It is described to realize that the GPU based on Docker frameworks renders the step of accelerating in virtual desktop system, including:
The video card of graphics cluster rendering node operating system is driven journey kernel file to be mapped in the integration environment of Docker containers,
The Window Scale plug-in unit of OpenGL is called simultaneously, realizes that the GPU real-time renderings in virtual desktop system accelerate.
4. the long-range real-time rendering platform construction method based on graphics cluster as described in one of claims 1 to 3, feature exist
In by the way of virtual external output, which includes for the display of the virtual desktop system:
For the virtual display output of each Docker container allocation, when user accesses the Docker containers, Docker containers
In figure pass through virtual display output, output to client so that GPU is rendered to be connect independent of the output of actual graphics card
Mouthful.
5. the long-range real-time rendering platform construction method based on graphics cluster as described in claim 1, which is characterized in that
Step 2 specifically includes:
VNC server-sides screen changed image-region using the regional change detection algorithm of divided-fit surface, through figure
As being sent to corresponding VNC clients after compression, the Remote desk process based on C/S frameworks is established by VNC agreements.
6. the long-range real-time rendering platform construction method based on graphics cluster as claimed in claim 5, which is characterized in that described
VNC server-sides screen changed image-region using the regional change detection algorithm of divided-fit surface, including:
Setting needs to change the unit area of detection;
Obtain the information that interception system screen redraws region;
If the system screen intercepted and captured, which redraws region, is less than unit area, the zone position information is sent;
If the system screen intercepted and captured, which redraws region, is more than unit area, interception system screen is redrawn into region and is split, will be torn open
Subregional coordinate information is stored into chained list, is obtained to each being torn open in chained list after subregional coordinate information carries out traversal detection
Need the region of variation location information retransmitted, it would be desirable to which the location information of the region of variation of retransmission store concurrent
It send.
7. the long-range real-time rendering platform construction method based on graphics cluster as described in claim 1, which is characterized in that
In step 3, the graphics cluster dynamic task scheduling includes with administrative mechanism:
Under certain load balancing, Service Source is provided to the user by cluster management node reasonable arrangement rendering node;
The graphics cluster dynamic task management and scheduling be related to including but not limited to single user to single node, multi-user to list
Node, the single user dynamic translation between one or more situations in four kinds of situations of multinode to multinode and/or multi-user.
The load balancing includes:
In single user to single node, rendering node is distributed using polling schemas;
The case where multi-user is to multinode, resource utilization, network bandwidth and I/O rates of rendering node etc. are considered
Index loads variance or standard deviation using dynamic load leveling model solution rendering system and is asked for the objective programming of minimum
Topic, the graphics cluster load balancing after being optimized.
8. the long-range real-time rendering platform construction method based on graphics cluster as described in claim 1, which is characterized in that
The long-range real-time rendering platform based on graphics cluster of the comprehensive integration includes platform client, graphics cluster management section
Four point, graphics cluster rendering node and shared memory systems parts;
Platform client is deployed in user's operation terminal, for initiating application request to graphics cluster server-side;
Graphics cluster management node and graphics cluster rendering node are deployed in graphics cluster server-side, are used for response platform client
Initiate application request;
Shared memory systems, for storage management user profile, virtual desktop system mirror image and/or rending model data letter
Breath.
9. the long-range real-time rendering platform construction method based on graphics cluster as claimed in claim 8, which is characterized in that
Step 4 specifically includes:
VNC clients are embedded into platform client, Docker containers and VNC server-sides are embedded into graphics cluster rendering node,
Docker mirrored storages in shared memory systems, by graphics cluster management node according to dynamic task scheduling and administrative mechanism into
Row cluster load balance;
When receiving platform client initiation application request, graphics cluster management node is that platform client specifies rendering node,
Rendering node starts Docker containers and loads Docker mirror images, while establishing long-range real-time rendering channel using VNC agreements, is
User provides long-range real-time rendering service.
10. the long-range real-time rendering platform construction method based on graphics cluster as described in one of claim 1-9, feature exist
In,
The application model of the long-range real-time rendering platform includes:
Graphics cluster server-side will be rendered by graphics cluster management node and be appointed after the application request for being connected to platform client initiation
To rendering node, rendering node provides rendering service to the user by creating independent virtual desktop system for business distribution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810109865.XA CN108388460B (en) | 2018-02-05 | 2018-02-05 | Remote real-time rendering platform construction method based on graphic cluster |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810109865.XA CN108388460B (en) | 2018-02-05 | 2018-02-05 | Remote real-time rendering platform construction method based on graphic cluster |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108388460A true CN108388460A (en) | 2018-08-10 |
CN108388460B CN108388460B (en) | 2021-05-18 |
Family
ID=63074556
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810109865.XA Active CN108388460B (en) | 2018-02-05 | 2018-02-05 | Remote real-time rendering platform construction method based on graphic cluster |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108388460B (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109213561A (en) * | 2018-09-14 | 2019-01-15 | 珠海国芯云科技有限公司 | The equipment scheduling method and device of virtual desktop based on container |
CN109725986A (en) * | 2019-01-29 | 2019-05-07 | 中山大学 | A kind of graphical and containerization virtual network environment building and batch dispositions method |
CN109949201A (en) * | 2019-01-29 | 2019-06-28 | 福建多多云科技有限公司 | A kind of image rendering accelerated method of linux container |
CN110032413A (en) * | 2019-03-05 | 2019-07-19 | 华为技术有限公司 | A kind of desktop virtualization method, relevant device and computer storage medium |
CN110442389A (en) * | 2019-08-07 | 2019-11-12 | 北京技德系统技术有限公司 | A kind of shared method using GPU of more desktop environments |
WO2020093843A1 (en) * | 2018-11-11 | 2020-05-14 | 长沙摩智云计算机科技有限公司 | Distributed multi-terminal and multi-network supporting system for android online game |
CN111163345A (en) * | 2018-11-07 | 2020-05-15 | 杭州海康威视系统技术有限公司 | Image rendering method and device |
CN111343219A (en) * | 2018-12-18 | 2020-06-26 | 同方威视技术股份有限公司 | Computing service cloud platform |
CN111984352A (en) * | 2020-08-10 | 2020-11-24 | 南京江北新区生物医药公共服务平台有限公司 | Desktop cloud service platform system based on spice protocol optimization |
CN112085988A (en) * | 2020-09-11 | 2020-12-15 | 北京信息科技大学 | Virtual simulation experiment system and method for penetration attack |
CN112200712A (en) * | 2020-09-08 | 2021-01-08 | 成都安易迅科技有限公司 | GLES image rendering method and device, storage medium and computer equipment |
CN112398914A (en) * | 2020-10-28 | 2021-02-23 | 武汉大势智慧科技有限公司 | Cloud rendering platform based on Kubernetes container cluster |
CN112470194A (en) * | 2019-04-12 | 2021-03-09 | 艾司科软件有限公司 | Method and system for generating and viewing 3D visualizations of objects with printed features |
CN115185667A (en) * | 2022-09-13 | 2022-10-14 | 天津市天河计算机技术有限公司 | Visual application acceleration method and device, electronic equipment and storage medium |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631634A (en) * | 2012-08-24 | 2014-03-12 | 中国电信股份有限公司 | Graphics processor virtualization achieving method and device |
-
2018
- 2018-02-05 CN CN201810109865.XA patent/CN108388460B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103631634A (en) * | 2012-08-24 | 2014-03-12 | 中国电信股份有限公司 | Graphics processor virtualization achieving method and device |
Non-Patent Citations (2)
Title |
---|
DJ0379: "在低速网络下利用VNC实现应用程序共享的解决方案", 《HTTPS://BLOG.CSDN.NET/DJ0379/ARTICLE/DETAILS/52397788?BIZ_ID=102&UTM_TERM=JINGLE%20VNC&UTM_MEDIUM=DISTRIBUTE.PC_SEARCH_RESULT.NONE-TASK-BLOG-2~ALL~SOBAIDUWEB~DEFAULT-0-52397788&SPM=1018.2118.3001.4187》 * |
PREETH E N等: "Evaluation of Docker Containers Based on Hardware Utilization", 《2015 INTERNATIONAL CONFERENCE ON CONTROL, COMMUNICATION & COMPUTING INDIA (ICCC)》 * |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109213561A (en) * | 2018-09-14 | 2019-01-15 | 珠海国芯云科技有限公司 | The equipment scheduling method and device of virtual desktop based on container |
CN111163345B (en) * | 2018-11-07 | 2021-11-26 | 杭州海康威视系统技术有限公司 | Image rendering method and device |
CN111163345A (en) * | 2018-11-07 | 2020-05-15 | 杭州海康威视系统技术有限公司 | Image rendering method and device |
US11872482B2 (en) | 2018-11-11 | 2024-01-16 | Hunan Duoxingyun Cloud Technology Co., Ltd. | Distributed multi-terminal and multi-network supporting system for android online game |
WO2020093843A1 (en) * | 2018-11-11 | 2020-05-14 | 长沙摩智云计算机科技有限公司 | Distributed multi-terminal and multi-network supporting system for android online game |
CN111343219A (en) * | 2018-12-18 | 2020-06-26 | 同方威视技术股份有限公司 | Computing service cloud platform |
CN111343219B (en) * | 2018-12-18 | 2022-08-02 | 同方威视技术股份有限公司 | Computing service cloud platform |
CN109725986B (en) * | 2019-01-29 | 2023-01-20 | 中山大学 | Graphical and containerized virtual network environment construction and batch deployment method |
CN109725986A (en) * | 2019-01-29 | 2019-05-07 | 中山大学 | A kind of graphical and containerization virtual network environment building and batch dispositions method |
CN109949201B (en) * | 2019-01-29 | 2023-04-07 | 福建多多云科技有限公司 | Image rendering acceleration method for Linux container |
CN109949201A (en) * | 2019-01-29 | 2019-06-28 | 福建多多云科技有限公司 | A kind of image rendering accelerated method of linux container |
CN110032413B (en) * | 2019-03-05 | 2021-08-31 | 华为技术有限公司 | Desktop virtualization method, related equipment and computer storage medium |
CN110032413A (en) * | 2019-03-05 | 2019-07-19 | 华为技术有限公司 | A kind of desktop virtualization method, relevant device and computer storage medium |
CN112470194A (en) * | 2019-04-12 | 2021-03-09 | 艾司科软件有限公司 | Method and system for generating and viewing 3D visualizations of objects with printed features |
CN110442389A (en) * | 2019-08-07 | 2019-11-12 | 北京技德系统技术有限公司 | A kind of shared method using GPU of more desktop environments |
CN110442389B (en) * | 2019-08-07 | 2024-01-09 | 北京技德系统技术有限公司 | Method for sharing GPU (graphics processing Unit) in multi-desktop environment |
CN111984352A (en) * | 2020-08-10 | 2020-11-24 | 南京江北新区生物医药公共服务平台有限公司 | Desktop cloud service platform system based on spice protocol optimization |
CN112200712A (en) * | 2020-09-08 | 2021-01-08 | 成都安易迅科技有限公司 | GLES image rendering method and device, storage medium and computer equipment |
CN112200712B (en) * | 2020-09-08 | 2023-10-27 | 成都安易迅科技有限公司 | GLES image rendering method and device, storage medium and computer equipment |
CN112085988A (en) * | 2020-09-11 | 2020-12-15 | 北京信息科技大学 | Virtual simulation experiment system and method for penetration attack |
CN112398914A (en) * | 2020-10-28 | 2021-02-23 | 武汉大势智慧科技有限公司 | Cloud rendering platform based on Kubernetes container cluster |
CN112398914B (en) * | 2020-10-28 | 2023-03-24 | 武汉大势智慧科技有限公司 | Cloud rendering platform based on Kubernetes container cluster |
CN115185667A (en) * | 2022-09-13 | 2022-10-14 | 天津市天河计算机技术有限公司 | Visual application acceleration method and device, electronic equipment and storage medium |
CN115185667B (en) * | 2022-09-13 | 2022-12-20 | 天津市天河计算机技术有限公司 | Visual application acceleration method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN108388460B (en) | 2021-05-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108388460A (en) | Long-range real-time rendering platform construction method based on graphics cluster | |
CN109582425B (en) | GPU service redirection system and method based on cloud and terminal GPU fusion | |
CN101553795B (en) | Multi-user display proxy server | |
AU2018301289B2 (en) | Graphical rendering using multiple graphics processors | |
KR101956411B1 (en) | Delivering a single end user experience to a client from multiple servers | |
US20180095710A1 (en) | Systems and methods for providing adapted multi-monitor topology support in a virtualization environment | |
US7899864B2 (en) | Multi-user terminal services accelerator | |
CN101430649B (en) | Virtual computation environmental system based on virtual machine | |
US6710753B2 (en) | Multi-screen session mobility between terminal groups | |
US9507618B2 (en) | Virtual machine system supporting a large number of displays | |
USRE46386E1 (en) | Updating a user session in a mach-derived computer system environment | |
CN106020929A (en) | System and method for supporting 3D application in virtual environment | |
US20150356773A1 (en) | Systems and methods for providing virtual monitors | |
CN102918490A (en) | Interacting with remote applications displayed within a virtual desktop of a tablet computing device | |
CN108762934B (en) | Remote graphic transmission system and method and cloud server | |
CN103685441B (en) | A kind of remote desktop control system based on Loongson terminal | |
CN1441363A (en) | Distributive resource structure and system | |
CN113902866B (en) | Double-engine driven digital twin system | |
CN109697120A (en) | Method, electronic equipment for application migration | |
US8860740B2 (en) | Method and apparatus for processing a display driver in virture desktop infrastructure | |
CN105653347B (en) | A kind of server, method for managing resource and virtual machine manager | |
CN102968807B (en) | image automatic generation method and system | |
Carroll et al. | 3D rendering in the cloud | |
CN115827236A (en) | Method and system for optimizing load performance of live-action three-dimensional cloud release process | |
CN109144453A (en) | Massive information high-resolution collaborative work cloud platform based on parallel computation |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |