News and Technical Resources - CRU Solutions - Introduction to Virtualization Technology

Join Our Email Newsletter List

Read Previous

Email Newsletters

440.891.0330

sales@crusolutions.com

Resources > Technical Resources

Introduction to Virtualization Technology

By Jim Kerr, President, CRU Solutions

A Brief Overview
Virtualization is a technique for organizations with at least three servers to consolidate and get better use out of server resources. In simple terms, a well-designed virtualization solution combines the right servers (often blade servers), storage devices, and virtualization software to reduce the number of server boxes, mitigate possible hardware failures and significantly reduce the risk of downtime. A well-designed virtualization solution can save organizations thousands of dollars every year in operational costs.

Traditional Server Configuration before Virtualization
Let’s start with a straightforward example of how servers are configured today, including redundancy and processor utilization considerations. A mid-level server today has built-in redundancy—disks, power supplies, processors, and cooling. (Memory can also be configured in a fault tolerant method, but that’s out of the scope of this paper.) Processor utilization will be discussed later.

Let’s look at the disk system, including redundancy considerations. When hard drives are inside the server, it is called “direct attached storage” because the storage is directly attached to the server. In the diagram below, we see that “storage” is actually a three drive RAID-5 disk array (“RAID”, say it like the bug spray) with a hot spare (an empty back-up disk for immediate data replication). RAID-5 requires a minimum of three disks. Data and parity are stripped across the three drives such that if one drive fails, the data that was on that drive can be recreated from the data on the remaining two. During the disk failure the server is in diminished capacity, but running. The hot spare kicks in taking the place of the failed disk. The hot spare is automatically rebuilt, it replaces the failed drive, and the failed drive is deactivated.

Let’s see what that takes to accomplish. If each disk is 250GB and you need four disks—the minimum number for RAID-5 with a hot spare—you buy 1,000GB of disk space (4 x 250GB), yet have only 500GB usable. Half of your disk space goes to redundancy and fault tolerance!

Why do we use disk space for redundancy? For two main reasons:
• Compared to the cost of down-time, disk redundancy is cheap
• Your data is valuable and worth this level of protection.

Power supplies, processors, and cooling are also redundant. Each server system has all of these redundancies – two hot swap power supplies, two processors, and four redundant hot swap fans. All of these are critical to maintaining business critical up-time.

Look at the multiplication effect when you add an eMail server and an SQL server.

Server 2 is dedicated to eMail. This is likely a Microsoft Exchange server, although it could be another such as iMail. Depending on your Spam defense, the server could also be running software to help identify spam and keep it out of your inbox.

Server 3 is a database (SQL or structured query language) server. It houses more robust database applications such as sophisticated accounting applications, CRM applications (Microsoft CRM, SalesLogix, etc.) as well as custom applications.

Now, let’s revisit the redundancy across three servers. We have a great deal of unused disk space. In fact, if Server 2 and Server 3 were configured just as we discussed above, you would have three hot spare drives and three redundant drives in the RAID-5 array. You would purchase twelve hard drives and lose half of them to redundancy! Disk space is relatively inexpensive, but wouldn’t it be nice to have a shared storage facility that all the servers access? That would greatly reduce the number of drives dedicated to redundancy. We’ll come back to this.

Moving from disk redundancy, let’s examine how busy the processors are. The charts below show processor utilization over a two week period. Processor utilization data is essential when developing an appropriate virtualization solution.

These charts include actual data collected by CRU Solutions’ Server Monitoring service. The red line indicates the highest reading during the interval, green is the lowest reading during the interval, and blue is the median during the interval.

Server 1 processor utilization (above) shows average utilization at about 6% (blue line). The red line indicates peak utilization. The highest peak is 20%.

Server 2 has an average utilization (blue line) as high as 16%. That was one brief peak over the two week period. Excluding that spike, utilization is about 7%.

Server 3 has a much different look to it with red peaks as high as 40% and median activity at less than 5%.

Based on these utilization rates, we have a great deal of unused process capacity spread across the servers. If we look at peak utilization (red line), server 1 is 20%, Server 2 is 25%, and Server 3 is 40%. It is apparent that a great deal of processor remains unused. Wouldn’t it be nice if there were a way to collect the idle processor utilization and use it for something?

In summary, virtualization is the key to maintaining necessary disk redundancy while increasing processor utilization.

Why Virtualize?
Virtualization has different benefits to different organizations. The prominent reasons already discussed are maintaining necessary redundancy while increasing processor utilization. For these objectives, as well as server consolidation, virtualizing results in the following positive outcomes:

• Improved Efficiency
• Reduced Costs
• Higher Utilization
• Ease of Management
• Availability/Security
• Load Balancing
• IT Flexibility
• Improved Business Agility

Below is an example from the IT@Intel Study describing the advantages of virtualization over a traditional multi-server configuration. There are several processors on the market, but for the purposes of this paper Intel products will be used. Though this example shows 8 server boxes combining into one virtualized server, individual business needs may dictate consolidating fewer boxes into more than one virtualized server.

Remember, reducing the number of physical servers magnifies the importance of server hardware reliability, future growth, and compatibility. Review your business needs carefully and make sure your requirements are adequately addressed by the virtualization solution. Be careful not to put all your eggs in one basket.

Virtualization Architecture
A virtualized server environment creates three unique networks: the Management Network, the Storage Network, and the User Network. This example shows two virtualized servers, each replacing several traditional servers, using intelligent™ hardware produced by CRU Solutions.

Components
There are several distinct component areas to a virtualized server project as illustrated in the diagram above:

1. Virtual Manager Server Console
2. Servers that will be the hosts for the several virtual servers
3. Storage repository (Storage area network, SAN) devices
4. Enclosure (rack cabinet) that will hold the servers
5. SAN Network
6. Management Network
Each one of these areas typically requires in-depth analysis to properly size the solution.

Virtualized Environment
At the core of the virtualized environment is virtualization technology (VT) that is part of the CPU itself. This hardware-level technology provides the hardware “hooks” for the virtualization manager software. This close interaction and communication between the CPU and the virtualization manager creates a stable platform for virtual servers.

The Virtual Machine (VM) Manager application controls the member servers. The VM Manager has the ability to move virtual servers to different physical servers in real time, to load balance, and to perform many other functions.

The servers that house the virtual instances are called member servers. Member servers communicate with each other and the Management server over the Management Network. The Management Network is a standard IP gigabit Ethernet network with a managed switch. This is a management-dedicated network and uses a dedicated network interface port in each server.

The data store, or disk storage, comes in the way of a SAN (Storage Area Network) device. SAN Devices vary in capacity, expansion, and cost. Medium-sized SAN devices hold up to twelve hot swap hard drives. This capacity is shared among the virtual servers. There are a number of ways to approach a SAN implementation. If you are unsure how to approach this, CRU Solutions offers a Technology Health Analysis, described on our website.

Optionally, and depending on Disaster Recovery objectives, a second SAN device may be installed in a different location. This could be at a branch office or even in a different state! The devices can replicate data across a company WAN or across the Internet.

Referring to the diagram, the SAN devices are part of a dedicated iSCSI storage network. The iSCSI network includes server and storage devices only. We continue to take full advantage of the knowledge base in IP Ethernet networking. The iSCSI network is a dedicated, gigabit Ethernet IP network. (For very high bandwidth applications, you may consider a Fiber Channel storage network instead of the IP Gigabit network.)

These devices are all rack-mounted. Industry best practices tell us that a lockable cabinet is the best solution to house the equipment. This cabinet can be rolled, fully encloses the data equipment, and has lockable doors to prevent unauthorized access.

Conclusion
Virtualization is a technique for organizations to maximize the use of their server resources. A well-designed virtualization solution will mitigate possible hardware failures and have the proper storage solution in place to significantly reduce the risk of downtime. With these details planned properly, organizations can save thousands every year in operational costs.

Jim Kerr is President and Founder of CRU Solutions, which is celebrating its 25th year in business. CRU is a comprehensive IT systems solution provider of exclusive, industry-leading hardware, software, connectivity, expertise and a unique approach to customer service. CRU’s intelligentTM brand of desktops, laptops and servers are custom built in Northeast Ohio exclusively with matched Intel components. You can reach Jim Kerr directly at jim.kerr@crusolutions.com. Additional information is available at www.crusolutions.com.

printer friendly