RU2543962C2 - Hardware-computer system for virtualisation and management of resources in cloud computing environment - Google Patents

Abstract

FIELD: physics, computer engineering.

SUBSTANCE: invention relates to cloud computing systems. A virtualisation module is connected to a second group of computers for storing program sessions and a control computer. A scheduling module is connected to a second group of computers for storing program sessions and a control computer; a graphics control means is connected to the second group of computers for storing program sessions and the control computer. The virtualisation module is configured create a virtual machine by emulating at least one hardware resource selected from: high-efficiency computer resources, partitioned file storages, memory and create one-to-one links between at least one selected hardware resource and emulation thereof. The scheduling module is configured to control processes of allocating/retrieving hardware resources for the virtual machine, assign a priority for each virtual machine at the launch thereof and assign weight and limit parameters thereto, where the weight parameter determines the value of the real available processor time, and the limit parameter determines the value of the maximum available processor time. The graphics control means is configured to allocate network resources between virtual machines based on information obtained from the control computer.

EFFECT: shorter time for switching a load between nodes.

Description

The claimed invention, in General, relates to the field of cloud computing systems designed for data processing.

The development and modernization of IT infrastructure (IT - information technologies) of public sector organizations, strategic industrial enterprises and companies, as well as in the commercial sector as a whole, is moving towards centralizing computing, building high-performance data centers (data centers), or renting computing resources and services. For large companies and organizations, this involves the construction / rental of distributed computing environments.

The key disadvantages of the current IT landscape of systems built on the basis of the classical architecture are: low resource utilization, heterogeneous load, excessive duplication, complexity and non-transparency of management, low resource mobility, which impedes its operational redistribution between different functions.

From the prior art ([1], p.44) it is known the construction of computer systems in the form of clusters (multi-machine systems), which are a computer complex consisting of several computers (nodes), as well as software and hardware communications computers that provide work all computers of the complex as a whole.

Clusters are used to increase the reliability and performance of a computing system. Reliability is increased due to the fact that when one of the cluster nodes fails, the computational load (or part of it) is transferred to another node.

For the exchange of service information, nodes use interprocessor communications. Using these links, each cluster node periodically checks the status of other nodes and the computational tasks they perform. If any node or one of its tasks (included in the set of fail-safe tasks) changed its state to non-working, then the procedure of moving (reconfiguring) the load to one of the working nodes begins. At the same time, the protected task must store its data on one of the shared disks, so that the new node can continue to use them after the failure of the main one.

However, such a solution has all of the above disadvantages, since there is no resource management mechanism in such a way as to ensure optimal load distribution in the system.

The complexity and opacity of cluster management is due to the fact that each of the computers included in its composition can operate under the control of its own, different from the others, operating system.

In addition, this solution has a long reconfiguration time, since it is associated with the activation of a new copy of the software process on another node.

In this case, it is also possible to lose some of the data in the RAM of the failed node.

The closest analogue of the present invention is a hardware-computer complex for providing access to software in the concept of cloud computing, including interconnected and connected to the Internet computer group, a computer group for storing program sessions, file storages and high-performance computing resources, while the computing complex is additionally equipped with a computer for resource accounting, through which a group of computers for storing program sessions is connected and is highly productive th computing resources ([2]).

The specified complex, as follows from the description, ensures the achievement of a technical result in the form of an extension of the class of software products that can be installed in the hub infrastructure (the platform on which the software product can be installed). At the same time, the resource accounting (management) system, which such a complex provides, provides only a restriction of access to resources to persons who have not passed the audit (see [2], p.3, p. 41-43).

Thus, the complex, adopted as the closest analogue of the present invention, also has all the above disadvantages, since the resource accounting system used in it does not provide their optimal distribution taking into account the computing power of the nodes.

In addition, in this complex there is no technical solution to the problem of a long time reconfiguration of nodes.

The above disadvantages are solved by the claimed invention, which allows for efficient load distribution in the system by resource management, as well as to reduce the time of reconfiguration of nodes.

A hardware-computer complex for virtualization and resource management in a cloud computing environment is proposed.

The preferred implementation of the proposed hardware and computer complex for virtualization and resource management in a cloud computing environment includes the first group of working computers and the second group of computers connected to each other and connected via a network for storing program sessions, as well as the control computer through which the second group of computers for storage of software sessions, high-performance computing resources and shared file storage.

Moreover, the complex is different in that it additionally includes a virtualization module, a scheduling module and a schedule management tool, while a virtualization module is connected to a second computer group for storing program sessions and a control computer, a scheduling module is connected to a second computer group for storing program sessions and with the control computer, the schedule management tool is connected with a second group of computers for storing program sessions and with a control computer, the virtualization module being configured to create a virtual oh machine (VM) by emulating at least one hardware resource selected from: high performance computing resources, shared file storage, memory and creating an unambiguous relation between the at least one hardware resource selected and its emulation; the scheduling module is configured to control the allocation / return of hardware resources for the VM and assign weight and limit parameters for each VM, where the weight parameter determines the amount of real available processor time, and the limit parameter determines the maximum available processor time; the graph management tool is configured to distribute network resources between the VMs based on information received from the management computer.

The technical result achieved by the claimed invention is to significantly reduce the transfer time of the load between nodes due to the presented virtualization technology, since it does not depend on the type of installed operating system on the user's computer (working computer) and does not require activation of a new copy of the software process. In addition, the introduction of a dispatching module with declared functions allows for efficient load distribution in the system, taking into account requests from users' computers and the available computing power of real devices.

The above and other advantages of the claimed invention will become clear to the specialist from the following detailed description of the essence of the claimed invention.

It should be noted that the information below should not be taken as limiting the scope of legal protection of the claimed invention defined by the formula, since they are given only as an example of the invention. One skilled in the art will recognize that there may be other embodiments of the present invention consistent with the spirit and scope of the present invention.

In this case, a network can be understood as the Internet, as well as other types of networks (for example, LAN, WLAN, etc.).

To overcome the above disadvantages existing in the prior art, a system of virtualization of hardware resources is additionally introduced into the proposed hardware and computer complex.

Virtualization technology in this case is represented by hypervisor virtualization of hardware resources. Hypervisor virtualization is based on the hypervisor (virtual machine monitor). It acts as an intermediary between the physical devices of the server and their presentation in the guest operating system. It features flexibility and support for almost any operating system as a guest operating system.

The virtualization and resource management system allows for the following functions.

Providing virtualization and separation between the following virtual machines of the following types of hardware (physical) host resources: physical processors, RAM, storage resources, external hardware interfaces (USB, IEEE 1394, RS-232, FC, SAS, SCSI, IDE, LPT , COM). The exclusive provision of virtual machine access to a specific interface is a special case of virtualization. Virtualization of these interfaces is carried out both selectively, in accordance with the needs for these interfaces of the created virtual machines, and simultaneously.

Providing the ability to start and execute (in resource sharing mode) within the virtual machines of the following types of guest operating systems (hereinafter OS): a) MS Windows family of OS certified for information security requirements with the possibility of use in automated systems with a security class of at least 1G ; b) OS of the Linux family, certified according to information security requirements, with the possibility of use in automated systems with a security class of at least 1G.

Providing the possibility of discrete changes in the number of resources allocated to virtual machines (processor time quotas, disk storage resource sizes, RAM sizes).

Providing scheduling the execution of virtual machines on physical servers (the second group of computers), taking into account the priorities assigned to individual virtual machines at the time of their launch.

Providing management of running virtual machines, including: changing execution priority, changing quotas for allocated resources, moving virtual machines between hosts.

Providing monitoring of the state of virtual machines (loaded / running / stopped), the number of assigned resources.

Maintaining at least 64 virtual processors on a guest OS, with the possibility of a subsequent increase to 128.

Maintaining at least 64 GB of virtual RAM on the guest OS, with the possibility of increasing to 128 GB.

Maintaining at least 2 virtual network interfaces to a virtual machine, with the possibility of a subsequent increase in number during operation to 8.

Enabling virtualization of physical network interfaces between virtual systems. The exclusive provision of network resources to a dedicated virtual server is a special case of virtualization.

Providing scheduled snapshots of virtual machines

Providing the ability to archive virtual machine snapshots.

Providing an increase in disk space for guest operating systems without rebooting / shutting down to 10 TB.

The proposed hardware and computer complex works as follows.

The hypervisor introduced into the complex includes a virtualization module, which is designed to virtualize hardware resources. Virtualization means masking hardware resources because it is impossible to share physical resources between different computers.

The virtualization process consists of emulating a device (hardware resource), creating connections between the device and its emulation, assigning (if necessary) access policies of virtual machines (VMs) to the virtualized resource.

A set of VMs can be implemented through a second group of computers for storing program sessions.

The virtualization module is responsible for creating virtual devices, and the dispatching module is responsible for the distribution of virtual resources.

Memory virtualization is defined as translating the address space of physical memory into a virtual address space.

Network virtualization is defined as the process of combining hardware and software network resources into a single virtual network. Network virtualization is divided into external, that is, connecting several hardware networks into one virtual, and internal, creating a virtual network between VMs running on the same host.

Virtualization of disk devices (file storages) is defined as a transparent representation of the storage system at the block level, when the logical address of the block is in no way tied to its real, physical address. Virtualization of shared disk devices enables you to combine physical storage devices into virtual disk pools. Separate virtual disks can be allocated from pools, which are connected to the VM if necessary.

The information used in the implementation of the VM creation algorithm is: unique identifiers of all hardware resources; information about the availability of hardware resources (RAM, processors, storage); information about the created VM (RAM, processors, virtual disk size, network parameters).

The result of the implementation of the VM creation algorithm is the creation of a unique VM identifier and registration of the VM parameters (in the control computer or in the virtualization module).

The information used in the implementation of the VM removal algorithm is: a unique identifier for the deleted VM; information about the state of the deleted VM (running / not running).

The result of implementing the VM removal algorithm is to stop the VM, if it is running, and to delete the VM metadata and the hard disk image. At the same time, a VM status request should occur - if the VM is running, it must be correctly completed and resources released before uninstalling.

VM migration is carried out in case of need for maintenance of the hardware node, when compacting lightly loaded VMs in order to save resources. The information used in the implementation of the VM migration algorithm is: a unique identifier for the transferred VM; portable VM metadata; Unique identifiers of all hardware resources information about the availability of hardware resources (RAM, processors, storage); unique identifier of the node to which the migration will be made;

The result of the implementation of the VM migration algorithm is the transfer of the VM to another hardware node.

During cold migration, the VM stops and starts it on another node. According to the requirements for fault tolerance, all pool resources are equivalent and have information about all VMs and access to a single repository, therefore, transferring information is not required, you just need to give a command to start the VM to the control module of another hardware node.

During hot migration, the image of the running memory of the running VM is transferred to the other hardware node in the background; thus, the VM does not stop. The term "live migration" means the process of switching the execution context of a VM from one pool host to another.

The virtualization module consists of several components: a component for providing hardware resources, designed to allocate hardware resources, virtualize them and provide them for use by virtual machines; a management component for managing a component for providing hardware resources and providing an external application interface (API) for computer management; event registration component designed to record events during virtualization and management.

These components as part of the module can be made in the form of agents located on a computer-readable medium or in memory, during which the computer provides the implementation of the described functions. In addition, such components can be made in the form of devices located on a common data bus, operating under the control of appropriate software.

The dispatching module (as part of the hypervisor) is designed to control the processes of allocation / return of hardware resources for VMs.

In particular, the distribution of processor resources occurs according to the following algorithm.

All virtual processors (vCPUs) of virtual machines are lined up in the dispatch module (for example, in its RAM). The scheduling module processes the vCPU queue, distributing the vCPU among the available physical processors in accordance with the policies. Distribution policies are described by VM parameters - weight and limit parameters. The weight value determines how much real processor time the VM will receive. For example, a VM with a weight of 512 will receive twice as much processor time of a physical processor on the host as a VM with a weight value of 256. The value of the parameter weight can vary from 1 to 65535, the default is 256.

The limit value can be used to indicate the maximum amount of processor time that a VM can receive, even if the host system is idle. The value is expressed as a percentage: 100 is 1 physical processor, 50 is half the processor, 400 is 4 processors. Moreover, the limit value cannot be greater than VCPU * 100, i.e. more than one real processor cannot correspond to one virtual processor.

Processor resources can be selected from the composition of high-performance computing resources (which can be, for example, a processor pool).

The distribution of memory occurs according to the following algorithm.

The dispatch module allocates memory to the virtual machine for exclusive use. Allocating VM memory more than physical memory is not possible. The memory available for the VM is defined as the amount of physical memory of the host that is part of the pool, minus the amount of memory needed to run the service VMs. The maximum amount of memory allocated by the VM is determined by the VM parameters when the VM starts for execution. If the amount of available virtual memory is not enough to start the VM with the specified amount of memory, starting the VM is not possible.

The resources of physical memory can be selected from the second group of computers for storing program sessions.

Distribution of network resources between virtual machines is performed by schedule management tools.

VM calls to the disk subsystem can be prioritized using the Linux CFQ (Completely Fair Queuing) scheduler mechanism.

The algorithm for allocating disk space between VMs is similar to the memory allocation algorithm.

The scheduling module consists of several components: a component of allocation / return of hardware resources, designed to control the processes of allocation / return of hardware resources for virtual machines and control structures that control the distribution of RAM between the processes of the complex and virtual machines; a management component for managing a software component for allocating / returning hardware resources and providing an external API for computer management; component of event registration, designed to register events in the process of scheduling and management.

These components as part of the module can be made in the form of agents located on a computer-readable medium or in memory, during which the computer provides the implementation of the described functions. In addition, such components can be made in the form of devices located on a common data bus, operating under the control of appropriate software.

The control computer provides monitoring and accounting of resources allocated and consumed by individual virtual machines and exchanges the information received and the corresponding commands with the virtualization module, the scheduling module, and the schedule management tool.

This allows you to provide flexibility in the distribution of load on the network, and also allows you to timely ensure the creation of the required number of VMs for processing incoming requests from working computers.

Virtualization, scheduling and schedule management tools can be made in the form of a hardware-software complex based on existing computers under the control of the corresponding software.

References

[1] V.G. Olifer, N.A. Olifer. Computer networks, textbook for universities, St. Petersburg, 2005.

[2] RU 122505 U1, 11/27/2012, bull. No. 33.

Claims (1)

The hardware-computer complex of virtualization and resource management in the cloud computing environment, which includes the first group of working computers connected to each other and connected via a network and the second group of computers for storing program sessions, as well as the control computer through which the second group of computers for storing program sessions is connected sessions, high-performance computing resources and shared file storages, characterized in that it additionally includes a virtualization module, a dispatch module, and a control tool the schedule, while the virtualization module is connected with the second group of computers for storing program sessions and with the control computer, the dispatch module is connected with the second group of computers for storing the program sessions and with the control computer, the schedule management tool is connected with the second group of computers for storing the program sessions and with a control computer, the virtualization module being configured to create a virtual machine (VM) by emulating at least one hardware resource selected from: high-performance computing Tel'nykh resources shared file storage, memory and creating an unambiguous relation between the at least one hardware resource selected and its emulation; the scheduling module is capable of controlling the processes of allocating / returning hardware resources for VMs, assigning priority for each VM at the time of its launch and assigning it weight and limit parameters, where the weight parameter determines the amount of real available processor time, and the limit parameter determines the maximum available processor time time; the graph management tool is configured to distribute network resources between the VMs based on information received from the management computer.