VIRTUALIZATION

May 2007 
Communications News

Virtualized versus traditional blade platforms

Server consolidation has become a key strategy for reducing data center costs.

by Blaine Kohl

Maximizing the efficiency of data centers and providing high-availability computing services to organizations means increasing performance while minimizing costs and reducing power budgets. Various mechanisms can help IT departments accomplish these goals, but one that is rapidly increasing in popularity is virtualization. Virtualization lowers operating expenses by reducing IT labor costs through simpler configuration setups, decreases electrical expenses, and frees up costly floor space. In addition, virtualization can be employed via the operating system and/or hardware implementations.

Typical server deployments average $10,000 per server. Server network utilization averages between 10 percent and 15 percent, and CPU utilization rarely exceeds 10 percent. In addition, more servers frequently result in the need for a larger IT staff and increased network complexity.

If an IT professional’s pay averages $40 per hour and each server deployment requires up to 30 staff hours to complete, approximately $1,200 in labor costs alone can be reduced for each server that is eliminated via consolidation. This savings is in addition to the savings seen in reduced physical hardware expenses.

Many enterprises experience high hardware costs due to running one software application per server. For example, an entire server may be used for a print-server application or a file-sharing server application. Application multitasking on a single, traditional dedicated server, however, is risky, because if the print server crashes it could also cause the file-sharing server to crash, as well. On the other hand, dedicating one server per application contributes to poor CPU and network utilization rates.

A virtualized deployment can consolidate those dedicated servers into one physical server. Server virtualization is the masking of server resources, including the number and identity of individual physical servers, processors, and operating systems, from server users. The server administrator uses either a software application or dedicated hardware to divide one physical server into multiple isolated virtual environments. Virtual environments are sometimes referred to as virtual private servers, virtual machines (VM) or operating system domains (OSD).

A significant cost savings stems from a virtualized deployment requiring far fewer physical servers than a traditional enterprise deployment by moving physical application servers onto a virtual machine running on a host or virtualized blade. In real world cases, the consolidation ratio typically averages 10 virtual servers per physical machine.

The reduction in the number of physical servers saves real estate space, lowers electrical bills and saves on cooling expenses. Additionally, management tasks can be automated and controlled from a single console. Maintenance also becomes easier as virtual machines have the ability to be shut down, transferred across the network to a new physical server and brought up within seconds.

By reducing the number of servers, the number of hardware configurations is also reduced; therefore, software patches and upgrades become easier to manage. Server load balancing can be automated from a single point rather than by monitoring each server independently, allowing for better utilization of the corporate assets. Additionally, network bandwidth often improves in a virtualized environment as hardware can now fully utilize 10-Gigabit Ethernet (10GbE). While in a traditional deployment, each application server would not likely utilize 10GbE bandwidth, the aggregation of the virtualized servers can fully utilize the available network speed, leaving room for peak scenarios.

There are incremental costs associated with a consolidation of servers in the data center, including the virtualization machine software, network interface cards (NIC) and cabling. This is a fraction of traditional implementations, however, when compared to the potential savings of 10 servers for every one consolidated.

VM software inserts a thin layer of software between the operating system (OS) and the underlying server hardware. This layer, known as a virtual machine monitor (VMM) or hypervisor, is responsible for allowing multiple OS images and their running applications to share the resources of a single host server and allows multiple operating systems to run at the same time. Each OS appears to have the resources of the entire machine under its control, but beneath it, the hypervisor transparently ensures that resources are properly and securely shared between different OS images and their applications.

Virtualization allows each virtual machine to be isolated from the physical system by creating virtual processors, interrupts and devices. By doing so, no single virtual machine can cause a system fatal error; at worst, it can compromise only the individual virtual machine. Therefore, if the print server should crash, there will be no impact to the Web server or the file-sharing server running on the same physical hardware server. This allows mission-critical applications to be shared on a single physical server with no compromise to availability.

In addition to software virtualization, hardware virtualization is possible. One of the newer implementations gaining a lot of interest is the work being done by the PCI Express special interest group regarding IO virtualization and multiroot shared I/O.

The shared I/O approach maximizes I/O utilization in a blade chassis. In traditional blade

configurations, a network interface card is required for each blade unit to drive network traffic. When IT administrators consolidate server blades, they are reducing the number of touch points to the network, which also reduces the blade platforms’ overall network throughput.

An alternative approach allows a single shared 10GbE network controller to be shared by several blades simultaneously in a 10GbE environment, yet retains its simplicity as well as advantages such as independent operating systems on each blade. Incremental blades added to a platform can be significantly less expensive, as there is no need to purchase additional blade network adapters. Additionally, performance is able to reach near line-rate speeds, while the complete blade configuration can be up to $10,000 less expensive to deploy.

Blaine Kohl is vice president of marketing for Tehuti Networks, Austin, Texas.

For more information from Tehuti Networks:
www.rsleads.com/705cn-265