In Part 1 of this four-part series, we introduced the concept of virtualization and elastic provisioning in a private cloud environment and the driving forces behind the need for such systems within the telecommunications industry. While the amount of available hardware resources seems limitless, the industry is pushing OEM providers to minimize the usage of these resources and develop flexible and scalable solutions. Advances in hardware design and software efficiencies make this possible, and in Part 2, we’ll look at these aspects and their effect on the use of virtual machines in solutions.
Earlier, we described a private cloud as ‘a pool of resources a telecom provider has designed and developed for a specific set of purposes.’ Recent hardware advances make deployment of private clouds possible within the existing framework of a data center. Single-board computers (SBCs), dense memory configurations, and multi-core processors are now common in the industry, providing a tremendous amount of raw computing capacity. SBCs supporting up to 16 cores and 128 GB of RAM are now a reality. In addition, the AdvancedTCA architecture is moving rapidly to support 40G Ethernet fabrics. When these two aspects are combined, data plane and control plane performance is dramatically increased, as is the ability to perform Deep Packet Inspection and advanced cryptography.
Virtual machines deployed in a processor-rich private cloud environment allow for accelerated functioning without the performance penalties previously seen with emulated virtual devices. Direct assignment of I/O devices to virtual machines is now possible with technologies such as Intel’s VT-x. Advances in hardware and software cryptography acceleration are bringing to fruition the desire for enhanced security within the telecom industry while decreasing CPU usage in the process. Intel’s AES-NI instructions for its latest generation of Xeon processor, for instance, can double or triple AES encryption processing on standard SBCs, and Intel’s Cave Creek and Cavium’s Nitrox III are moving toward 20 to 40 Gbps encryption speeds.
Without the use of virtualization, control, and data plane applications tend to plateau between 2 and 6 CPU cores, preventing the complete utilization of available ATCA SBC capabilities. To counteract this in a private cloud, applications that are I/O-intensive can be mixed with applications that are CPU-intensive on the same computing node through the use of different virtual machines, and hardware usage can be maximized while keeping resources separate. Several instances of the same application can be run on a single SBC, each with its own virtual memory and CPU, all tied to a single physical processor to maintain a constant level of performance. This avoids the performance penalties seen when stand-alone applications are run that must access memory within an SBC through an extra interconnect.
Programmers can take advantage of virtual machines running in a private cloud because they create a uniform development environment, and by using a range of tested hardware platforms, can be deployed with little or no additional development needed. This makes the adoption of new hardware faster and makes it easier for multiple vendors to deploy applications using the same source computing elements.
In Part 3, we’ll investigate the design flexibility provided by deploying virtual machines in a private cloud, and how they can help reduce energy consumption at the same time.
Contact NEI online *or by calling (877) 792-9099 to learn more about carrier-grade deployment solutions and service capabilities. Working with NEI to build your solution allows you to focus on your core competency and reduce costs.*