OpenVZ technology scales as well as standard Linux kernel — up to thousands of CPUs and terabytes of RAM. Besides, a single container could be scaled up from taking a little fraction of available resources up to all resources available dynamically — you do not even have to restart the container. For example, containers can natively use up to all available CPUs which is different from hypervisor technology which requires special tricks like co-scheduling and even the best hypervisors are inefficient with more then 4-8 vCPUs. ===== How does OpenVZ improve efficiency of services? ===== For existing hardware, OpenVZ allows to utilize its processing power better by improving average load from 3-5% to at least 30-50%, while still providing ability to handle peak loads. To decrease complexity, OpenVZ provides standardized and centralized server management, logically decoupled from actual hardware. And when its time to buy new servers, you can now use few Read more powerful servers instead of many little ones — with added benefits of better reliability, better peak performance and typically longer lifespan. ===== How does OpenVZ improve flexibility of services? ===== By providing unified scalable platform with such unique features as rapid application and updates provisioning. Each container is hardware independent and can be moved to another about [[WP/Containers_density|OpenVZ-based system in seconds over the network. This allows for ease of hardware maintenance (move out all containers and do whatever you need with the box) and improved availability (keep a synchronized copy of your container elsewhere and start it up when primary service failed). If your old box is not able to cope with peak load anymore, just move your containers to a new one. ===== What is the performance overhead? ===== Near zero. There is no emulation layer, only security isolation, density]] and all checking is done on the kernel level without context switching. ===== What are [[performance expectations? ===== Peak performance is achieved when only one container has active tasks. In this case, it could use 100% of available resources: all CPUs, all physical memory, all disk and network bandwidth. OpenVZ is not limiting you to a single-CPU virtual machine]].