CN202662000U

CN202662000U - Node virtualization deployment device of render farm

Info

Publication number: CN202662000U
Application number: CN 201220207416
Authority: CN
Inventors: 陈涛
Original assignee: MOTION MAGIC DIGITAL ENTERTAINMENT Inc
Current assignee: Shanghai Magic Digital Creative Technology Co ltd
Priority date: 2012-05-09
Filing date: 2012-05-09
Publication date: 2013-01-09
Anticipated expiration: 2022-05-09

Abstract

The utility model relates to a node virtualization deployment device of a render farm. The device comprises a host machine, a dynamic host configuration protocol (DHCP) server, a file server and a rendering computing cluster, wherein the host machine is connected with the DHCP server, the file server and the rendering computing cluster respectively. Compared with the prior art, the device has the advantages of being capable of quickly and dynamically converting a user workstation which is completely idle or partially idle into a system of a rendering machine inside the render farm and the like.

Description

A kind of virtual deployment devices of farm node of playing up

Technical field

The utility model relates to a kind of animation and plays up correlation technique, especially relates to a kind of virtual deployment devices of farm node of playing up.

Background technology

In the three-dimensional animation production field, playing up and simulating is the step that consumes maximum computational resources.In order in project, to finish smoothly this process, can use a large amount of computing machines to carry out cluster in the mode of grid computing (grid computing) and calculate.Although the price of computing equipment is comparatively cheap at present, corresponding to making upper ever-increasing effect demand, calculated amount also rises at double by leaps and bounds.The result's who causes result is exactly that a lot of manufacturing companies must consume a large amount of costs of manufacture at these render farms.Simultaneously, for most of projects, the calculated amount that remove exactly to control cluster is very difficult, thereby is easy to cause the idle of cluster, perhaps paroxysmal great demand.One causes waste, second might affect owing to the computational requirements that the cluster of buying is difficult to tackle burst the progress of project.At this, the employed workstation of the computational resource that for most of manufacturing companies, all has and use seldom fully---producer itself.These workstations generally have powerful computational resource, and these resources usually only were used to the time of 1/3-1/2 among one day, and for the wright of some flow process, computational resource is to be in for a long time idle state more.Go if these computational resources fast can be added to calculate in the cluster easily, so not only can subdue a part of hardware spending, also provide a desirable buffering source for the computation requirement that happens suddenly simultaneously.

The utility model content

The purpose of this utility model is exactly to provide a kind of virtual deployment devices of farm node of playing up for the defective that overcomes above-mentioned prior art existence.

The purpose of this utility model can be achieved through the following technical solutions:

A kind of virtual deployment devices of farm node of playing up is characterized in that, comprise host, Dynamic Host Configuration Protocol server, file server and play up the calculating cluster, described host respectively with Dynamic Host Configuration Protocol server, file server, play up and calculate cluster and be connected.

Described host is provided with virtual machine, this virtual machine respectively with Dynamic Host Configuration Protocol server, file server, play up and calculate cluster and be connected.

Compared with prior art, the utlity model has following advantage:

1) teller work station that can dynamically will leave unused fully fast or partly leave unused is converted to the system of playing up machine in the farm of playing up, by this system, the keeper can isolate CPU and memory source idle in the teller work station easily, directly add in the render farm and go, also dynamically enlarge fast arithmetic capability;

2) use the linux system as the operating system of virtual machine in native system, the hardware device owing to virtual machine is completely fixed again, so can to significantly cutting of core, only stay necessary driving and mixed-media network modules mixed-media.Simultaneity factor is also simplified as far as possible, does not finally comprise calculating the size that can only have tens of megabyte with the core system of software.Therefore, native system can directly be downloaded this system virtualization dish from public file server when customer requirements starts, so can guarantee the integrality of system and more difficult by user's accidental modification.Because extremely the relation of simplified system can be finished in 1 minute usually so start to the adding cluster from system.Close when service at needs, because system obtains mirror image again, thereby direct hard closing virtual machine and do not worry system damage.The means of the quickest withdrawal system resource are provided for the user thus.

3) in a lot of situations, playing up the employed computational tool in farm can need to carry out regular update, in the utility model, only has system itself and needs the buffering of write permission can be comprised on the virtual hard disk with catalogue.All business softwares and library file all will be stored on the public file server, and virtual machine is mounted to this locality with these contents after startup.This mode will solve the problem of software/config update well.

4) because the each startup in core of system all can obtain a new copy, and other softwares also can be read-only carries, all buffer mediums that can write all can create when starting again, so it all is on all four that the utility model can guarantee to enter each time when calculating cluster, effectively stops physics host user's maloperation.

Description of drawings

Fig. 1 is structural representation of the present utility model;

Fig. 2 is applicating flow chart of the present utility model.

Embodiment

Below in conjunction with the drawings and specific embodiments the utility model is elaborated.

Embodiment

As shown in Figure 1, a kind of virtual deployment devices of farm node of playing up comprises host 1, Dynamic Host Configuration Protocol server 2, file server 4 and plays up and calculate cluster 3, described host 1 respectively with Dynamic Host Configuration Protocol server 2, file server 4, play up and calculate cluster 3 and be connected; Described host 1 is downloaded from file server 4 and is started virtual disk, and virtual machine 5 is started, and this virtual machine 5 joins the idle computational resource in the host to play up to calculate in the cluster plays up calculating.

Select virtualbox as virtual system, select Debian 6.02 x64 as virtual system, network connection is used the bridge joint mode, installs by bootstrap and simplifies system component most, recompilates core code, only keeps virtual hardware and drives.Samba client is installed, NFS client, virtualbox instrument, python2.6, the necessary assembly such as rsync.

Configuration dhcp server is for virtual machine distributes the assigned address section.Set up nfs at file server and derive point is placed calculating for virtual system necessary assembly and software.

The needed assembly of Cluster Rendering and client are installed.Write if necessary the automatic configuration software for the Cluster Rendering client, and in booting script, add and call.Following catalogue is moved on the position that nfs derives: bin, lib lib64, opt, usr.Empty all temporary files, for tmp, root, home, the original creation of var catalogue are built virtual disk, and carried out snapshot/backup.The extra etcMain catalogue that creates.Copy current etc content to this catalogue.Corresponding modify/etc/fstab.Note bin, lib lib64, opt, the read-only carry of usr, etcMain.

Configuration rsync writes the synchronize script from etcMain to etc.

Modification system booting script obtains the information from host behind the loading virtual system kit, generate necessary machine name according to managerial demand, revises current machine name; Behind the resource carry, the etc catalogue is upgraded, checked the rsync synchronizing content, if change, carry out mark, system is restarted, by restarting continuously in the short time avoided in the judgement of mark.

Use dd that root directory is filled 0, use virtual disk optimization tool reduced size after the deletion.And snapshot/backup, this backup is left on the file server, downloads for the host that needs to start virtual machine.

The utility model is distributed to the teller work station that each needs to start virtual machine;

Revise the nfs configuration sharing, read and write carry for keeper's (whole system is unique) carry, software maintenance is afterwards all finished by this special virtual machine, and the work of whole system deployment configuration is complete.

As shown in Figure 2, workflow of the present utility model is as follows:

1) resource of detection host judges whether its idling-resource satisfies minimum requirements, if yes, execution in step 2), if NO, return failure;

2) judge whether host can be connected with the access point in the file server, if yes, execution in step 3), if NO, return failure;

3) judge whether there is new startup virtual disk on the file server, if yes, copy new startup virtual disk to host, and execution in step 4), otherwise, execution in step 4);

4) host is cleared up left all interim virtual disks of using when starting last time, the original state after recovering to install;

5) write host information;

6) host starts virtual machine;

7) read host information, the machine name information of generating virtual machines;

8) Dynamic Host Configuration Protocol server is that virtual machine distributes the assigned ip address field;

9) judge whether virtual machine normally starts, if yes, execution in step 10), if NO, otherwise, execution in step 14);

10) attempt the carry remote resource;

11) judge whether successfully carry, if yes, execution in step 12), otherwise, execution in step 14);

12) virtual machine activation Cluster Rendering client, and judge whether to calculate cluster and successfully be connected with playing up, if yes, execution in step 13), otherwise, execution in step 14);

13) finish;

14) close virtual machine, clear up interim virtual disk, and return failure.

The resource of described host is by the resource of online allotment use, to comprise idle CPU, free memory and remanence disk space.The characteristic information of described virtual machine is automatically to give behind virtual machine activation, comprises machine name, Ip, login user information.

Claims

1. play up the virtual deployment devices of farm node for one kind, it is characterized in that, comprise host, Dynamic Host Configuration Protocol server, file server and play up the calculating cluster, described host respectively with Dynamic Host Configuration Protocol server, file server, play up and calculate cluster and be connected.

2. a kind of virtual deployment devices of farm node of playing up according to claim 1 is characterized in that described host is provided with virtual machine, this virtual machine respectively with Dynamic Host Configuration Protocol server, file server, play up and calculate cluster and be connected.