RU158705U1 - VIRTUAL SERVER LOAD DETERMINATION SYSTEM - Google Patents

Abstract

1. A system for determining the degree of load of a virtual server, comprising: a central processor; a central processor control unit configured to allocate time for executing instructions from virtual servers, and each of the virtual servers contains an interface configured to transmit control commands to virtual servers; - server resource usage sensor; - sensors of commands received via virtual server interfaces; - command sorting unit, configured to to combine commands for each virtual server according to the corresponding types of commands; - a hypothesis generation block about the connection between the server’s virtual server resource consumption and the command groups for each server; - a correlator that determines the mismatch between the resources consumed by the virtual containers of resources, according to the hypotheses, with the server’s resources actually used and generating a discrepancy signal at a predetermined size of discrepancy, and the hypothesis generation unit is made to refine the hypothesis for, at least one virtual server nesootvetstviya.2 detection signal. The system according to claim 1, characterized in that if, when changing the hypothesis for one virtual server, the mismatch value does not change, the system returns to the previous hypothesis for this server and the hypothesis for the other server changes.

Description

The technical field to which the utility model relates.

The utility model relates to virtual servers, in particular, to determining the load created by a separate virtual server on the physical resources of a computer or other computing system that supports many virtual servers. This utility model can be used to optimize the operating conditions of a computer or other computing system by determining the load created by the virtual server and subsequent reasonable redistribution of resources between computing systems or virtual servers.

In this case, “load” refers to the percentage of use by a separate virtual server in the total consumption of hardware resources of a real server.

When using virtual servers in Internet hosting systems, there is a problem of determining the consumption of real resources by individual virtual servers. Means for determining the consumption of resources provided by the operating system, for example, the Windows 8 task manager, described on the web page: http://interface31.ru/tech_it/2011/11/dispetcher-zadach-windows-8.html, determine the overall load of the computing system, and the use of resources by individual processes.

Due to the fact that the same process can be used by different virtual servers, it is not always possible to determine for which server the RAM, disk space or time of the central processing unit is used.

Thus, at present, there are no reliable methods and systems for measuring or calculating the real load on the hardware of the servers created by individual virtual servers.

The technical result provided by the implementation of the utility model is to increase the accuracy and reliability of determining the load created by individual virtual servers, for example, containers and virtual machines.

To achieve a technical result, the system for determining the degree of virtual server load contains:

- central processor;

- the control unit of the central processor, configured to allocate time to perform certain actions on the set of virtual servers located on the server and executed with an interface through which control commands are sent to the virtual servers;

- Sensor of server resource use;

- sensors of commands coming through the interfaces of virtual servers;

- block sorting teams, made with the ability to combine for each virtual server commands for the respective types of commands;

- a hypothesis generation block about the connection between the server’s consumption of server resources and server groups for each server;

- a correlator that determines the mismatch of the amount consumed by the virtual containers of resources, according to the hypotheses, with the server resources actually used and generates a mismatch signal at a predetermined mismatch size, and the hypothesis generation unit is made to refine the hypothesis for at least one virtual server when a signal is detected inconsistencies.

In the particular case of the device’s implementation, when the hypothesis for one virtual server is changed, the mismatch value does not change, it returns to the previous hypothesis for this server and the hypothesis for the other server changes.

This utility model is intended to improve the working conditions of a computer or other computing system by determining the load created by the virtual server.

When using software that supports virtual servers on hardware servers in computer centers, a set of software (virtual server) serving one user or one specific user task requires the use of hardware server resources to execute instructions and also to service related tasks. For example, the CPU time is spent on servicing the program code; disk space is used to store data. To carry out calculations and store intermediate results, random access memory is used, network applications use network communication channels. At the time of the task, some of the currently available resources are allocated. Depending on the type of actions performed, a particular task may require more or less resources: 1% of processor time is enough to complete some tasks (because most of the time is spent, for example, on input / output or waiting for data), but some complicated calculations and data processing may require 70% of processor resources or more. Similarly, various tasks require more or less RAM, high or low disk activity. Nevertheless, it was experimentally established that within the same virtual server, the same tasks consume approximately the same amount of resources, while the consumption of resources by the same tasks related to different virtual servers can vary significantly. This is often due to the fact that separate virtual servers are designed to serve the same type of operations, for example, one server can register user requests for the purchase of consumer goods, which requires a large number of calculations, per request and frequent access to local databases and does not require a large number Internet of data, and other servers can be used for broadcast broadcasting of multimedia content, which is accompanied by a large amount of outgoing traffic and large storage volume of data on disks, while the computational load on the system is minimal.

The drawings show:

in FIG. 1 - schematically shows the structure of a utility model with the designation of the interaction of all components of the system to determine the load on the system and optimize the system, according to one possible implementation;

in FIG. 2 is a graphical representation of the sequence of operations carried out by a utility model to identify the load on the system;

In order for the load on the virtual servers to be distributed optimally, it is necessary to correctly determine the load created by individual servers, and, if necessary, either redistribute the load or change the amount of resources available to individual virtual servers.

Client-server and Web-technologies are intended for the effective implementation of the functionality of applied software systems, the creation of conditions for the effective implementation of the company's business process system. In them, information processing processes are distributed between the database server, client and server parts. The application server receives data from the database server, serves N applications. Each application serves N clients, and one client can work with N applications.

The principal advantages of these technologies are: reduced requirements for workstation capacities; logic segmentation due to the movement of part of the code to secondary servers, distribution of processor load, increased reliability and availability; simplification of the problem of scaling; server optimization for specific tasks, reducing overhead. In addition, Web-based technologies provide a high degree of independence of the main components of software systems, although, on the other hand, they are less secure compared to client-server technologies.

Virtualization technologies are designed to abstract from the features of individual resource groups and combine them into hardware and software systems of the desired configuration - to simplify their management. They distinguish platform virtualization, which relates to the process of creating virtual machines, and resource virtualization, which transfers approaches to creating virtual machines to all types of resources (equipment, namespaces, networks, etc.). Virtualization reduces operating costs to support a server fleet, increases the availability of IT services and other benefits.

Cluster hosting technologies are aimed at solving the problems of typical distributed IT infrastructures, virtualizing the resources that physical servers provide. Joint operation of several systems improves the price-to-quality ratio, achieving high availability and performance, and balancing the load.

Each node runs a separate copy of the specified server applications (for example, applications for the web, FTP, and Telnet servers). Network load balancing distributes incoming client requests between nodes in the cluster. The load that each node must handle can be customized to meet your requirements. If you need to handle the additional load, nodes can be added to the cluster dynamically. In addition, network load balancing can direct the entire data stream to one dedicated host, called the default host.

Network load balancing allows you to access all the programs in the cluster. For applications with load balancing, when a node crashes or is disconnected, the load is automatically redistributed between working nodes of the virtual server. If the computer crashes or unexpectedly disconnects from the network, all active connections to such a server will be lost. In any case, the computer outside the network, when it is ready for operation, can again join the cluster in the operating mode and take on part of the load, which will reduce the amount of data attributable to the virtual servers of the computer.

In FIG. Figure 1 shows an example of a structural diagram of a utility model with a designation of the interaction of all components of the system to determine the load on the system and optimize the system. The example system shown in FIG. 1, comprises a monitor 10 or a display for displaying a result for viewing by a user; general-purpose computing device in the form of a central processor 11; random access memory 12; hard drive 13; network card 14; sensors 20, providing the determination of the number of requests or other commands generated by separate virtual servers or received on the virtual server from the outside and the system bus 16, which combines the above system components. The system also contains sensors 15 that determine the use of real resources of the computing system at any given time. The system bus is designed for the coordinated transfer of data and control commands between system elements. Sensors 15 determines the degree and total volume of computer resources consumption - processor loading at different points in time, network usage, RAM usage. The system bus 16 may be any type of bus providing data, addresses and control commands between system elements, including, it may contain hardware controllers and other active elements that provide data flow control functions and can be constructed using any bus architecture corresponding to computer system requirements. The correlator 17, with the support of the central processor 11, and using the system bus 16 directly interacts with the sensors 15 and 20. The sensors 20, each of which can correspond to either a separate virtual server or a separate group of virtual server commands, determine the current activity of virtual servers 18, intercepting commands passing through the interfaces 19 of the virtual servers. In the event that the sensors are not configured for specific commands, in the particular case of the implementation of the utility model, information from the sensors 20 is sent to the sorting unit 21.

Thus, on the one hand, the correlator receives data on the actual consumption of resources of the computing system, and on the other hand, indirect data on the activity and types of activity of individual virtual servers. The exact relationship between these parameters can be established using the hypothesis formation block 22.

In the particular case of the implementation of the utility model, the functionality of the hypothesis generation unit allows you to assign a weight coefficient for a specific group of commands of a specific virtual server, for example, for a group of commands matching a request using an encryption protocol or a group of commands leading to the start page or page 404 - “ data not found". For example, with a simple start page of a site supported by a virtual server, a request from an external user specifying a start page does not lead to a significant consumption of resources, and a request to the same site using a secure protocol with the need for active content generation leads to high processor load. At the same time, it is not always possible to recognize the need for generating content, for example, tables or graphs. Also, if the start page of a site served by another virtual server contains active content and uses external resources, such a request can form a relatively high load on the processor and increase the amount of incoming traffic. In the particular case of implementation, with this approach, weights can be found by solving overdetermined systems of equations or by determining the real load created when active, mainly one virtual server.

Thus, the hypothesis generation unit generates data that can be used as functional parameters to bring individual requests into correspondence with the load on the processor, RAM, disk space, or other hardware elements of the computing system caused by these requests.

The correlator uses this kind of functional parameters to convert together with the signals of the sensors 20 to identify the possible correspondence of the calculated resource consumption and the actual resource consumption determined using the sensors 15. In the event that it is possible to identify a correspondence, for example, by varying the time intervals between commands virtual servers and the estimated change in load consumption, the hypothesis is recognized as confirmed and, in the future, can be used as a source of data x on the consumption of resources to a particular virtual server. In the event that it is not possible to identify the correlation dependence, the hypothesis generation unit 22 refines or modifies the hypothesis, for example, by using another method of solving systems of equations.

In FIG. 2 is a graphical representation of the sequence of operations performed by a utility model to identify system load.

This scheme of the algorithm of action of the utility model begins with step 210 - the start of the systems.

At step 220, the initial hypothesis is formed by the hypothesis generation unit.

At step 230, a hypothesis test is performed. Hypothesis testing takes place every time a new one appears. Hypothesis testing is carried out by comparing the resource consumption calculated on the basis of the hypothesis with the actual resource consumption. If it is possible to test a hypothesis for only one server, then it is checked and, for those servers for which successful hypotheses have been generated, further hypotheses are not changed. Hypothesis testing for one server can be carried out during periods of relatively high activity of such a server, aging with other servers.

At step 240, a transition to a further change in the hypothesis occurs, if the hypothesis is incorrect or correct, or the search for hypotheses is stopped, if the hypotheses are formed successfully, at step 260. In the particular case of implementation, the hypothesis is checked periodically, for example, 4 times a day to exclude the possibility of hidden Errors when changing the working conditions of virtual servers.

Confirmed hypotheses are used to determine the load created by virtual servers and form solutions based on this data. Possible solutions include migrating the virtual server to another system, proactively allocating hardware resources to the virtual server, and stopping the server, which creates an unacceptably high load under the influence of malicious software.

Claims (2)

1. A system for determining the degree of load of a virtual server, comprising: - central processor; - the control unit of the Central processor, configured to allocate time to execute commands from virtual servers, and each of the virtual servers contains an interface configured to transmit control commands to virtual servers; - Sensor of server resource use; - sensors of commands coming through the interfaces of virtual servers; - block sorting teams, made with the ability to combine for each virtual server commands for the respective types of commands; - a hypothesis generation block about the connection between the server’s consumption of server resources and server groups for each server; - a correlator that determines the mismatch of the amount consumed by the virtual containers of resources, according to the hypotheses, with the server resources actually used and generates a mismatch signal at a predetermined mismatch size, and the hypothesis generation unit is made to refine the hypothesis for at least one virtual server when a signal is detected inconsistencies. 2. The system according to claim 1, characterized in that if, when the hypothesis is changed for one virtual server, the mismatch value does not change, the previous hypothesis for the server is returned and the hypothesis for the other server is changed.

Images