CN106933654B - A cache-based virtual machine startup method - Google Patents

Abstract

When a physical node of the method provided by the invention receives a request for applying a virtual machine, firstly, whether a target mirror image is cached in the physical node or a brother node is searched, and if the target mirror image is cached and the physical node or the brother node can meet the resource requirement, the virtual machine is directly started in the physical node or the brother node. Therefore, the method provided by the invention can reduce the occurrence of mirror image transmission to the maximum extent. Then, steps S5 and S6 plan transmission paths from the individual source points and the selection of intermediate switches, which optimizes bandwidth utilization across the network. Therefore, the method provided by the invention can effectively improve the transmission speed. In addition, step S8 includes determining whether the target image needs to be cached after the virtual machine is started. This enables the target image to be cached in the physical node included in the cluster where the target is physically located. The occurrence of image transfers is reduced for subsequent virtual machine boots. Therefore, the method provided by the invention can effectively accelerate the starting speed of the virtual machine.

Description

A cache-based virtual machine startup method

technical field

The present invention relates to the field of cloud computing, and more particularly, to a cache-based virtual machine startup method.

Background technique

In recent years, with the development of cloud computing technology, many traditional companies and individual users have moved their computing tasks from traditional servers to cloud computing centers. Through the virtual technology of cloud computing, the utilization rate of many IT resources (including cpu, memory, hard disk, and network resources) is greatly improved, and the cost is reduced.

As an important technology in cloud computing, virtual machine technology has been widely used in various cloud computing platforms as a carrier of common computing resources. For example, well-known cloud computing platforms such as Azure, Amazon EC2, and RackSpace have adopted virtual machine technology to provide users with high-efficiency and highly scalable computing services, and users generally agree with this model.

Although virtual machine technology has been accepted by people, the technology itself still has some technical difficulties:

1) Concurrent use of users: Among cloud computing users, they are mainly divided into individual users and enterprise-level users. Among them, enterprise users usually apply for the activation of multiple virtual machines at one time due to large business volume, resulting in the same period of time. There are a large number of images in the data center that need to be transmitted, which instantly increases the load on the network bandwidth. During this period, the startup time of the virtual machine will be greatly improved, reducing the user experience.

2) Diversity of images: In cloud computing, users usually use various systems, such as: windows, linux, ioS, etc., and the same windows operating system is also divided into winXP, win7, win8, winServer, etc. version, and for open source operating systems such as linux, there are more types (Centos, Fedora, Ubuntu...). This requires the existence of these various images in the data center network.

3) Special architecture of the data center: In the data center, a special architecture represented by a fat tree is generally used instead of the traditional multi-fork tree structure. Corresponding to such a special structure, the traditional caching solution cannot make good use of the characteristics of the topology structure.

Centralized processing method: In traditional data centers, the method of processing data information is usually centralized, that is, there must be a central control node in the network, which masters global information (including request distribution, the situation of the network, the situation of the resources, etc.). When the data center receives a request, it will also be handed over to the centralized control node for processing to determine where the virtual machine for the request is instantiated and where to obtain the required resources. The advantage of a centralized processing scheme is that it can grasp the global information, so as to make an optimal decision. But the disadvantage is also obvious, because the data center needs to grasp the global information situation in the network in real time, so any scheduling and resource changes in the network must be requested and notified to the control node. This leads to the following problems:

1) These requests and notification information themselves additionally increase the bandwidth load in the network, and when there are multiple concurrent requests, the originally congested network is even more congested.

2) Now that the scale of the data center is getting larger and larger, when maintaining these network information, the requirements for the computing power of the control node itself will also increase. If there are too many requests, it is likely to occur, because this control The instantiation delay caused by the node processing is not timely.

3) Because all requests are processed through this single control node, "single point of failure" must be a problem that has to be considered.

During the instantiation process of a virtual machine, it is usually necessary to obtain the target virtual machine image from a specific place (resource pool), and then transmit the target image to the target node through the network for instantiation. The transmission of multiple images at the same time will definitely affect the bandwidth resources of the entire network, resulting in a longer time for virtual machine instantiation, resulting in a very unfriendly user experience.

SUMMARY OF THE INVENTION

In order to solve the defects of consuming network bandwidth resources and prolonging the instantiation time of the virtual machine caused by the need to transfer the virtual machine image between the data source and the target physical node in the virtual machine startup method provided by the above prior art, the present invention provides a system based on Cached virtual machine startup method.

In order to achieve the above purpose of the invention, the technical scheme adopted is:

A cache-based virtual machine startup method, comprising the following steps:

S1. Physical node A receives a request to apply for a virtual machine. If physical node A caches a target image and the resources of physical node A meet the application requirements, the target image is directly used to start the virtual machine locally on physical node A; otherwise, go to step S1. S2;

S2. Find the parent node of physical node A, and then add the node of the son node of the parent node that is in the same cluster as physical node A and has the target image cached into the list PM_list;

S3.a) PM_list is not empty, and the resources of physical node B in PM_list meet the application requirements, use the target image cached by physical node B to start the virtual machine on physical node B;

b) If PM_list is empty, select the physical node C whose resources can meet the application requirements from the son nodes of the parent node of physical node A as the node where the virtual machine is placed, and then execute step S4;

c) If neither a) or b) can find a physical node whose resources can meet the application requirements, the request will be forwarded to the physical nodes of other clusters for processing;

S4. Starting from the physical node C, recursively search the mirror information cached by the son node of the parent node, and obtain all the clusters of physical nodes that have the target mirror cached;

S5. traverse the cluster obtained in step S4, select the corresponding physical node from the corresponding cluster as the data source according to the congestion degree of data transmission in the cluster;

S6. Loop through all layers in the fat tree, select the switch with the least workload among all the switches of equal status at each layer to construct the transmission path from the data source to the physical node C, and then transmit the target image from the data source through the transmission path to physical node C;

S7. Physical node C utilizes the target image to start the virtual machine locally; then judge whether it still has enough space to store the target image, if so, directly cache the target image, otherwise step S8 is performed;

S8. Calculate the appearance frequency of various images cached by the physical nodes included in the cluster where the physical node C is located. If the appearance frequency of the target image is higher than the appearance frequency of the other images, the target image will not be cached, otherwise, the image with the highest frequency will be executed. Delete, and then cache the target image in the physical nodes included in the cluster where physical node C resides.

In the above solution, when a physical node receives a request to apply for a virtual machine, it first checks whether there is a target image cached in itself or its sibling nodes. Start the virtual machine in the sibling node. Therefore, the method provided by the present invention can minimize the occurrence of mirror transmission. Then, steps S5 and S6 select from various source points and select intermediate switches to plan the transmission path, which optimizes the bandwidth utilization on the entire network. Therefore, the method provided by the present invention can effectively improve the transmission speed. Furthermore, after starting the virtual machine, step S8 further includes determining whether the content of the target image needs to be cached. This makes sure that the target image is cached in the physical nodes included in the cluster where the target is located. Reduces the occurrence of image transfers for subsequent virtual machine startups. Therefore, the method provided by the present invention can effectively speed up the startup speed of the virtual machine.

Preferably, the resources of the physical node can meet the requirements of the application, which means that the physical node can meet the following conditions: the amount of computing resources left by the physical node is greater than or equal to the amount of computing resources required to satisfy the target image.

Compared with the prior art, the beneficial effects of the present invention are:

When the physical node of the method provided by the present invention receives a request for applying for a virtual machine, it first checks whether the target image is cached in itself or its sibling nodes. Start a virtual machine on itself or on a sibling node. Therefore, the method provided by the present invention can minimize the occurrence of mirror transmission. Furthermore, after starting the virtual machine, step S8 further includes determining whether the content of the target image needs to be cached. This makes sure that the target image is cached in the physical nodes included in the cluster where the target is located. Reduces the occurrence of image transfers for subsequent virtual machine startups. Therefore, the method provided by the present invention can effectively speed up the startup speed of the virtual machine.

Description of drawings

FIG. 1 is a schematic diagram of steps S1 to S3.

FIG. 2 is a schematic diagram of steps S4 to S6.

FIG. 3 is a schematic diagram of steps S7-S8.

Detailed ways

The accompanying drawings are for illustrative purposes only, and should not be construed as limitations on this patent;

The present invention will be further elaborated below in conjunction with the accompanying drawings and embodiments.

Example 1

As shown in Figures 1, 2 and 3, the method provided by the present invention comprises the following steps:

S1. Physical node A receives a request to apply for a virtual machine. If physical node A caches a target image and the resources of physical node A meet the application requirements, the target image is directly used to start the virtual machine locally on physical node A; otherwise, go to step S1. S2;

S2. Find the parent node of physical node A, and then add the node of the son node of the parent node that is in the same cluster as physical node A and has the target image cached into the list PM_list;

S3.a) PM_list is not empty, and the resources of physical node B in PM_list meet the application requirements, use the target image cached by physical node B to start the virtual machine on physical node B;

b) If PM_list is empty, select the physical node C whose resources can meet the application requirements from the son nodes of the parent node of physical node A as the node where the virtual machine is placed, and then execute step S4;

c) If neither a) or b) can find a physical node whose resources can meet the application requirements, the request will be forwarded to the physical nodes of other clusters for processing;

S4. Starting from the physical node C, recursively search the mirror information cached by the son node of the parent node, and obtain all the clusters of physical nodes that have the target mirror cached;

S5. traverse the cluster obtained in step S4, select the corresponding physical node from the corresponding cluster as the data source according to the congestion degree of data transmission in the cluster;

S6. Loop through all layers in the fat tree, select the switch with the least workload among all the switches of equal status at each layer to construct the transmission path from the data source to the physical node C, and then transmit the target image from the data source through the transmission path to physical node C;

S7. Physical node C utilizes the target image to start the virtual machine locally; then judge whether it still has enough space to store the target image, if so, directly cache the target image, otherwise step S8 is performed;

S8. Calculate the appearance frequency of various images cached by the physical nodes included in the cluster where the physical node C is located. If the appearance frequency of the target image is higher than the appearance frequency of the other images, the target image will not be cached, otherwise, the image with the highest frequency will be executed. Delete, and then cache the target image in the physical nodes included in the cluster where physical node C resides.

In the above solution, when a physical node receives a request to apply for a virtual machine, it first checks whether there is a target image cached in itself or its sibling nodes. Start the virtual machine in the sibling node. Therefore, the method provided by the present invention can minimize the occurrence of mirror transmission. Then, steps S5 and S6 select from various source points and select intermediate switches to plan the transmission path, which optimizes the bandwidth utilization on the entire network. Therefore, the method provided by the present invention can effectively improve the transmission speed. Furthermore, after starting the virtual machine, step S8 further includes determining whether the content of the target image needs to be cached. This makes sure that the target image is cached in the physical nodes included in the cluster where the target is located. Reduces the occurrence of image transfers for subsequent virtual machine startups. Therefore, the method provided by the present invention can effectively speed up the startup speed of the virtual machine.

In the specific implementation process, the resources of the physical node can meet the requirements of the application, which means that the physical node can meet the following conditions: the amount of the remaining computing resources of the physical node is greater than or equal to the amount of computing resources required to satisfy the target image .

Obviously, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the embodiments of the present invention. For those of ordinary skill in the art, changes or modifications in other different forms can also be made on the basis of the above description. There is no need and cannot be exhaustive of all implementations here. Any modifications, equivalent replacements and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (2)

1. A virtual machine starting method based on cache is characterized in that: the method comprises the following steps:

s1, a physical node A receives a request for applying a virtual machine, and if a target mirror image is cached in the physical node A and the resource of the physical node A meets the requirement of application, the target mirror image is directly utilized to start the virtual machine at the local part of the physical node A; otherwise, executing step S2;

s2, searching a father node of the physical node A, and adding a node which is in the same cluster with the physical node A and is cached with a target mirror image in son nodes of the father node into a list PM _ list;

s3.a) if the PM _ list is not empty and the resource of the physical node B in the PM _ list meets the application requirement, starting a virtual machine at the physical node B by using the target mirror image cached by the physical node B;

b) if the PM _ list is empty, selecting a physical node C, the resource of which can meet the application requirement, from child nodes of the parent node of the physical node a as a node where the virtual machine is placed, and then executing step S4;

c) if the physical node of which the resource can meet the application requirement cannot be found in the two conditions of a) and b), forwarding the request to the physical nodes of other clusters for processing;

s4, starting from the physical node C, recursively searching mirror image information cached by the son node of the father node of the physical node C to obtain all clusters of the physical nodes cached with target mirror images;

s5, traversing the clusters obtained in the step S4, and selecting corresponding physical nodes from the corresponding clusters as data sources according to the congestion degree of data transmission in the clusters;

s6, circulating all layers in the fat tree, selecting a switch with the minimum workload from all switches with the same position in each layer to construct a transmission path from a data source to a physical node C, and then transmitting the target mirror image from the data source to the physical node C through the transmission path;

s7, the physical node C locally starts a virtual machine by using the target mirror image; then judging whether the target mirror image has enough space to store the target mirror image, if so, directly caching the target mirror image, otherwise, executing the step S8;

s8, calculating the occurrence frequency of various mirror images cached in the physical node in the cluster where the physical node C is located, if the occurrence frequency of the target mirror images is higher than the occurrence frequency of the other mirror images, not caching the target mirror images, otherwise deleting the mirror images with the highest occurrence frequency, and caching the target mirror images into the physical node in the cluster where the physical node C is located.

2. The cache-based virtual machine startup method according to claim 1, characterized in that: the resource of the physical node can meet the requirement of application, which means that the physical node can meet the following conditions: the amount of computing resources remaining for the physical node is greater than or equal to the amount of computing resources needed to satisfy the target image.

Images