CN105607948A - Virtual machine migration prediction method based on SLA - Google Patents

Abstract

The invention relates to the technical field of cloud computing, in particular to a virtual machine migration prediction method based on SLA. Firstly, monitoring the resource use condition of each virtual machine at intervals; then, performing curve fitting by a mathematical method based on the monitoring data to obtain an equation of the curve, and predicting the quantity of resources used by the virtual machine at the next time interval; then comparing the resource usage amount of the next time interval with the size of a threshold value specified by the SLA, if the resource usage amount exceeds the threshold value, the virtual machine is likely to be overloaded, otherwise, waiting for the next monitoring time interval to arrive; and finally, if the load is overlarge, the virtual machine is migrated to a more idle host. The invention provides a virtual machine migration prediction algorithm based on SLA, which realizes an active virtual machine migration strategy by actively predicting the resource use trend of a virtual machine, and solves the problem that the SLA cannot be met after the migration of the traditional migration strategy; can be used for migration of virtual machines.

Description

A SLA-based Virtual Machine Migration Prediction Method

technical field

The invention relates to the technical field of cloud computing, in particular to an SLA-based virtual machine migration prediction method.

Background technique

In a cloud computing environment, in order to make full use of resources, it may happen that virtual machines assigned by multiple applications according to the SLA are located on the same server. Moreover, it may also happen that the sum of virtual machine resources exceeds the upper limit of physical resources of the server. When the load of an application increases, the resources of other applications are preempted. At this time, the cloud service provider cannot meet the resource commitment made to the user in the SLA. At this time, virtual machine migration is often used to alleviate the problem.

The traditional method is to isolate the performance of virtual machines based on alarms. This method first monitors the resources of each host. Taking memory as an example, when the memory usage of a certain host is higher than a threshold, a warning is issued and automatically Virtual machine migration. The advantage of this method is that the migration operation must be correct, because resource preemption does occur. However, the disadvantage of this method is that when migration is required, the resources of the host are very scarce, and the virtual machine that cannot meet the SLA will increase the load because it cannot obtain more resources, and the resource load of all virtual machines on this physical machine will be reduced. Each tends to a certain constant. At this time, it is impossible to know which virtual machine needs to be migrated. And if the migrated virtual machine is not a virtual machine that cannot meet the SLA, the virtual machine with a sharp increase in load may still remain on this host, and it may still continue to seize resources, and the situation that cannot meet the SLA may continue to occur .

Contents of the invention

The technical problem solved by the present invention is an SLA-based virtual machine migration prediction method, which solves the problems in the aforementioned prior art.

The technical scheme that the present invention solves the problems of the technologies described above is:

Described method comprises the following steps:

Step 1: Monitor the resource usage of each virtual machine at regular intervals;

Step 2: After curve fitting by mathematical method based on the monitoring data, the equation of this curve is obtained, and then the amount of resources used by the virtual machine in the next time interval is predicted;

Step 3: Compare the resource usage in the next time interval with the threshold specified by the SLA. If the threshold is exceeded, the virtual machine is likely to be overloaded soon. Then perform step 4. Otherwise, wait for a while and monitor it next time When the time interval comes, execute step 1;

Step 4: Migrate this virtual machine to a relatively idle host.

The curve fitting refers to saving the monitoring data as historical data, then selecting a suitable curve for curve fitting, and then selecting an equation that conforms to the characteristics of the curve. Commonly used curves that can be selected include logarithmic function, exponential function, and quadratic function , polynomial functions above quadratic, trigonometric functions, etc.

The resource usage refers to performance indicators of the virtual machine, such as CPU usage, memory usage, etc. The specific indicators can be selected according to business requirements.

Method of the present invention can produce following beneficial effect:

1. The method of the present invention is an active virtual machine migration strategy, which predicts which virtual machine has a sharp increase in load and performs migration before the host resource shortage occurs. This approach ensures that virtual machine resource insufficiency will never occur.

2. The method of the present invention is a cost- and performance-balanced migration strategy, which can ensure that idle physical machine resources are used first, and migration is performed only when the load is too high to meet the SLA.

Description of drawings

Below in conjunction with accompanying drawing, the present invention is further described:

Fig. 1 is a flowchart of the present invention;

detailed description

The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

Algorithms can differentiate the importance of virtual machines according to their demand for resources, which can be compared to treasures with many different values. By monitoring the data, you can know the number of idle resources of each host. Larger free resources can accommodate more virtual machines, which is similar to having several empty backpacks. Therefore, the problem of virtual machine migration strategy becomes a 0-1 knapsack problem.

The input parameters of the algorithm are the virtual machine's demand for resources, the total number of virtual machines, and the idle resources of the host. SS is a virtual machine migration policy stack. After calculating the idle resources of m hosts to form a BagSet and sorting them, the resource usage of each virtual machine is predicted by PolyFitForecast.

Put the virtual machine that needs to be migrated into the treasure set TreasureSet. Solve the 0-1 knapsack problem for each knapsack in the knapsack set to get the set of virtual machines that need to be migrated to the current host and put them into the policy stack.

The algorithm for solving the 0-1 knapsack problem for each knapsack in the knapsack set (BagSet) is as follows:

To solve the 0-1 knapsack problem, you can use the dynamic programming algorithm. Since the method of solving the 0-1 knapsack problem is a classic algorithm, the solution obtained must be the optimal solution, so you can get the best virtual machine migration strategy .

.

Claims (3)

1. the virtual machine (vm) migration Forecasting Methodology based on SLA, is characterized in that, described method comprise withLower step:

Step 1: the resource service condition to every virtual machine is once monitored at set intervals;

Step 2: after carrying out curve fitting by the method for mathematics based on monitor data, obtain the side of this curveJourney, then dopes the resource quantity that next time interval virtual machine uses;

Step 3: the size of the threshold value that relatively next time interval resource use amount and SLA specify, if exceededThreshold value, this virtual machine is probably will load excessive, performs step so 4, otherwise waits for one sectionTime, while arrival at upper once monitoring period interval, execution step 1;

Step 4: this virtual machine (vm) migration is gone to comparatively idle main frame.

2. method according to claim 1, is characterized in that, described curve refers to monitoringData save as historical data, then choose suitable curve and carry out curve fitting, and then select to meet curve spyThe equation of levying, the conventional curve that can select has logarithmic function, exponential function, quadratic function, more than secondaryPolynomial function, trigonometric function etc.

3. method according to claim 1 and 2, is characterized in that, described resource service condition refers toBe the performance indications of virtual machine, as CPU usage, memory usage etc., concrete index can be according to businessDemand is selected.