Content starts here
For all of the innovation of the last 10 years, the typical data center is still one of the past: racked, stacked and wired. A lot of companies have forayed into virtual machines to introduce efficiency and flexibility, often beginning with a few servers and applications. Compelling ROI leads to new ad-hoc projects, then more ad-hoc projects. Before long, thousands of virtual servers flood the environment.
Data centers likely will always consist of both physical elements and virtual elements. With so much interest in virtual environments, companies can forget that every virtual server must run on a physical system. When virtualization springs up without a master plan, businesses wind up with two sets of tools to manage two separate infrastructures, the physical and the virtual. Chaos ensues, and the savings of initial virtualization projects are swallowed up.
Today, you can apply the principles of virtualization throughout your infrastructure in ways that could only be imagined a few years ago. But before you get entrenched in ad-hoc projects, take a step back and examine how to build in virtualization at all levels of your infrastructure, from servers to storage to networking and management. Here are some points to consider as you create your plan.
Go for a balanced architecture. Quad-core servers have tremendous potential. But if they’re not matched with equally impressive memory and networking capacity, you’ll face constant bottlenecks that prevent you from realizing the potential of four cores. Don’t skimp on the other specs.
Avoid a multi model setup. Most applications have specific recipes for the amount of network, storage and server resources they need. That’s why vendors tend to offer so many different models. But each model has its own firmware, spare parts and procedures. Expecting administrators to know all of the nuances by heart is asking a lot. The alternative, a one-configuration-per-box model, can quickly become expensive and complicated.
By moving to a bladed infrastructure with shared storage, you can break free of multiple configurations. You have more flexibility to take one configuration and expand its memory, network capacity and storage. Because of this flexibility, you can standardize on fewer configurations for more applications. And you can make better use of storage by pooling and sharing it, instead of stranding capacity on multiple disk drives in every server.
Use one big pipe. Virtual machines need lots of network connections and that can get expensive. Rather than having six cables and six cards, why not have one fat pipe with one card and one cable connected to one switch port? What if you could carve up that fat pipe and use the network bandwidth where it's needed most? That would translate into more flexibility in how virtual machines use capacity for different workloads.
Virtualize network connections. In the physical world, network changes require physical changes. You must literally unplug cables and plug them in elsewhere, and most likely call your network administrator to help. That requires a lot of coordination, people and time. On the other hand, when you have a single high-bandwidth network connection with virtualized connections to servers and storage, you can change configurations within the pipe, logically, all without moving or unplugging anything. This way, the network simply follows the application, regardless of the location of physical or virtual server.
Invest in a SAN Optimized for virtual environments. Consolidating storage resources using a networked shared storage system, such as a SAN, offers cost, flexibility and availability benefits for both physical and virtual server environments. In addition, shared storage is required for advanced server virtualization features, such as high availability computing. But virtual server environments exert unique capacity utilization, performance, management, and availability requirements on a shared storage system. So be sure to invest in a SAN with the architecture and features to meet such requirements.
Leverage what you have and know. For small environments, new storage innovations are providing flexible approaches. For example, virtual SAN software can pool direct attached storage and present it as shared SAN storage to both physical and virtual machines. This is a great way to meet the unique demands of a virtual server environment if you don’t have expertise in SAN technology or require the scalability of a dedicated SAN.
For most virtual server deployments, a dedicated iSCSI SAN is becoming the preferred storage option. iSCSI offers the flexibility, scalability and ease-of-management necessary for virtual server environments, especially as it relates to enabling virtual machine (VM) mobility. And because iSCSI is a well understood, cost effective IP storage networking standard, iSCSI SANs are more affordable and easier to manage than traditional SANs.
For enterprise virtual server environments that need to store large amounts of information quickly, a Fibre Channel SAN is ideal. Fibre Channel is designed for the enterprise and many large organizations may have Fibre Channel technology and expertise already in place, thanks to existing Fibre Channel networking technology and storage disk arrays.
Treat the physical as virtual. Instead of having separate management consoles for physical resources and virtual resources, choose management software that views them the same way and lets you manage them together - at anytime from anywhere. With a holistic management tool, you can take full advantage of a virtual infrastructure’s inherent flexibility.
Better use the capacity you have. You probably have management consoles that spit out data about how much storage and network bandwidth you’re using at any given moment. But raw data is like raw flour: It takes lots of time, know-how and additional tools to turn it into something consumable.
What consoles need is built-in capacity planning, a tool to turn raw data into the answers to “what-if” questions. That way, you spend your time rolling out new applications, not crunching numbers in spreadsheets.
Those tools need to understand both virtual and physical resources. And given the rising cost of power, they need to take electricity usage into account, too.
Automate provisioning. Now that all of your connections are virtual, find a way to make them work on their own. Imagine that you need to add an application; if your management solution enables you to build and save automated workflows, you can design a process that will locate the correct application image on a server somewhere in the network and load it for you. In fact, you can design templates for many frequently repeated tasks for specific applications, such as Exchange or SAP, and remove the cost and time of manual processes. You can automate the entire life cycle of your infrastructure, from provisioning through retirement and cut weeks out of the provisioning effort in the process.
Virtualization is all about the flexibility to modify your infrastructure when conditions require it. But the luxury of changing things on the fly and at a moment’s notice is possible only after you’ve done some careful planning.
After all, if you’ve planned adequately, you’ll build enough flexibility into your infrastructure that you can change anything—even the very plan that guides you—anytime.
