Server virtualization

Server virtualization is a technology for abstracting computing resources (CPU, RAM, disk, network) that allows running several isolated virtual machines (VMs) simultaneously on one physical server, each with its own operating system (Windows, Linux, Astra Linux and others), applications and settings. This technology is the basis of modern IT infrastructure, since it radically increases the efficiency of equipment use, reduces capital and operating costs, simplifies server fleet management and provides a high level of fault tolerance and flexibility. Without virtualization it is impossible to imagine modern data centers, cloud platforms (IaaS, private clouds) and corporate IT landscapes of any size. According to Gartner and IDC analysts, more than 90% of all server workloads in the world today run in virtual environments. Virtualization allows solving the key problem of traditional server infrastructure — low equipment utilization: physical servers are rarely loaded more than 10-15%, while virtualization allows achieving utilization of 70-80% and higher while maintaining a performance margin for peak loads.

How server virtualization works

The key component of virtualization is the hypervisor — special software (VMware ESXi, Microsoft Hyper-V, Proxmox VE, KVM/QEMU, domestic developments) that is installed directly on the server hardware (bare-metal, type 1) or on top of the operating system (hosted, type 2). The hypervisor creates an abstraction layer between the physical resources of the server (a processor with support for Intel VT-x/AMD-V hardware virtualization, RAM, disks, network interfaces) and virtual machines, dynamically distributing resources between them depending on the current load using a scheduler. Each virtual machine has its own virtual devices: a virtual processor (vCPU), virtual memory (vRAM), virtual disks (VMDK, VHDX, QCOW2, RAW formats) and virtual network adapters (VMXNET, E1000, VirtIO). The hypervisor intercepts privileged instructions and calls of the guest operating system to devices and translates them into calls to physical equipment (binary translation or Intel EPT/AMD NPT hardware virtualization). Thanks to the use of the virtualization technology, almost native performance of virtual machines is achieved (performance losses are usually less than 5% for CPU-bound workloads) with complete isolation from each other at the level of memory address spaces. This means that an operating system failure, memory leak or compromise of one VM does not affect the operation of neighboring virtual machines on the same host. The administrator can easily move running virtual machines between physical servers without stopping services (live migration, vMotion), combine the resources of several servers into clusters with high availability (HA) support and automatically restart VMs on healthy equipment in case of a node failure, as well as ensure continuous operation (Fault Tolerance, FT) for critical VMs with the creation of a synchronous copy on another host.

Advantages and benefits for business

The introduction of virtualization brings a number of measurable advantages to the enterprise, confirmed by numerous studies. First, a radical reduction of equipment costs (CAPEX): instead of dozens of underloaded physical servers, the company uses several powerful servers with 70-80% utilization, which reduces capital costs by 3-5 times with comparable total performance. Second, significant savings in operating expenses (OPEX): reduction of electricity consumption by up to 80% (fewer servers = fewer kilowatt-hours), a proportional reduction of data center cooling costs and the release of useful space in racks. Third, a radical simplification and acceleration of the deployment of new services: creating a new virtual machine from a template takes 2-5 minutes instead of the several hours, days or even weeks needed to purchase, deliver, unpack, install, connect and install the OS on a physical server. Fourth, increased fault tolerance and business continuity due to clustering, live migration, automatic recovery in case of node failures, as well as the ability to quickly create snapshots and backup VMs at the hypervisor level. Fifth, simplified testing and development: you can quickly deploy isolated test and pre-production environments, create snapshots for rollback on failed updates, clone production VMs to analyze problems without risk to production. The Fintech company offers a full cycle of virtualization implementation work — from architecture design and platform selection to installation and commissioning. The installation service includes hypervisor installation, configuration of high availability (HA) clusters and load balancing (DRS), integration with storage systems (SAN) and backup systems (SRK), as well as migration of existing physical workloads to a virtual environment without interrupting business operations (P2V — Physical to Virtual migration).

Virtualization and import substitution

Within the framework of the import substitution policy, Russian public sector organizations are actively switching to domestic virtualization platforms. Mature products are available on the market (zVirt from Orion Soft, PK SV Brest from Astra Group, SpaceVM, SimpleVM, HOSTVM), included in the register of Russian software and compatible with leading Russian operating systems — Astra Linux Special Edition, RED OS, Rosa Linux, ALT Linux, Elbrus Linux and others. These platforms provide all necessary enterprise-level functions: centralized management of a pool of physical hosts and virtual machines through a single console (Web GUI), management of virtual machines and containers, hot migration (live migration), automatic load balancing (DRS), integration with storage systems via NFS and iSCSI protocols, integration with backup, monitoring systems (Zabbix, Saymon). With a competent approach, migration from foreign hypervisors (VMware vSphere, Microsoft Hyper-V) to domestic solutions using V2V conversion is painless for business and allows fully removing the risks associated with restrictions on the use of foreign software in critical information infrastructure (CII, 187-FZ). The specialists of the Fintech company have deep competencies in this area and are ready to provide a full range of services for the transition to domestic virtualization platforms, including auditing the current infrastructure, planning migration (with an assessment of application compatibility), performing P2V/V2V migration, training administrators and subsequent technical support.

Containerization and virtualization

It is important to distinguish between classical virtualization (hardware, hypervisor) and containerization (OS-level virtualization). Virtualization runs full-fledged virtual machines with their own operating system, drivers and a separate kernel, which provides maximum isolation at the hardware level and compatibility with any applications and OS, but requires more resources (each guest OS takes up several gigabytes of disk space and up to 1-2 GB of RAM for service needs). Containerization (Docker, containerd, Kubernetes) uses the shared kernel of the host operating system, isolating applications at the process level through namespaces (PID, Network, Mount, User, IPC) and control groups (cgroups v2) to limit consumed resources. Containers start in milliseconds, consume significantly less memory (the image size can be units-tens of megabytes for Go/statically compiled applications), but require application compatibility with the host OS (Linux compatibility) and provide less strict isolation compared to VMs. Both approaches have their advantages and areas of application, and modern IT infrastructure often uses them in combination: virtual machines run on the hypervisor, inside which container clusters with Kubernetes (k8s) orchestration are deployed — the so-called "virtualization + containers" architecture. The choice of a specific approach and their combination depends on business tasks, security requirements (containers or VMs), the level of team competencies and architectural preferences.