Our weakening economy means the enormous savings resulting from a virtualized data center make its implementation still more compelling. And yet, fearing their IT staffs lack the skill sets required to effectively administer the new environment, management is often slow to roll out virtual infrastructure as their organization's production data center platform. The reality is that a VMware virtualized data center is very intuitive, and it is simpler, less demanding and far more resilient than a physical architecture. Rather than being encumbered with the burden of learning a new technology, IT staff are empowered by being able to accomplish more in significantly less time. Virtualization takes the best tools from an administrator's tool belt and makes them better.
Virtualized data centers give administrators a degree of flexibility and control not possible in the physical world. This can be seen, perhaps, most prominently by the challenge of implementing effective remote access. Physical environments rely upon either a hardware or software based form of remote access. A hardware solution frequently utilizes a fairly expensive IP based KVM, a tool that does not always provide a good user experience. Alternatively, a software solution such as Microsoft RDP is not able to provide administrators with access outside the operating system which must be running in order to obtain access. And neither of these solutions provides an administrator with access to the power functions required to turn on a machine that was mistakenly shutdown.
Virtualized data centers enable an administrator to open a console connection to a virtual machine. This console connection provides the same "out of operating system" access as an IP based KVM (an administrator may watch the VM power up from BIOS to OS), but provides performance that is comparable to using a local system. Moreover, the administrator has full access to all power functions remotely. This accessibility enables an administrator to fully control her environment, be it from her desk in the office or from home.
Adding Server Hardware
Adding additional hardware to a physical data center server can be an awkward process. If a server needs more memory, for example, an administrator must first open the server to determine how many (if any) RAM slots are available. The administrator then needs to determine the most cost effective way to achieve the desired amount of memory given the particular motherboard restrictions. An approval process and purchase order drafting follows along with a subsequent wait in shipping time for RAM or, in some cases, for a technician to come and install the memory. After the hardware is acquired, the installation often must take place after hours in order to minimize user downtime. Administrators have learned to compensate for this process by over allocating server resources in order to minimize the need for hardware changes.
Adjusting a virtual machine's memory, on the other hand, is as simple as shutting the machine down, then adjusting a slider bar. Additional hard disks or NICs are a short wizard away. Additional CPUs can be added via a pull down menu. These hardware changes can be accomplished in mere minutes whereas their physical counterparts might take weeks. Access to the servers' hardware allows administrators to get away from the wasteful "over allocate" paradigm and allows them to size their virtual servers accurately.
The stress becomes palpable when an unanticipated IT problem forces administrators to scramble in an attempt to quickly determine and fix the cause. The administrators are handicapped, though, in their ability to attempt different fixes by their mandate to never do anything that could make the situation worse. Registry edits, for example, are common examples of changes that carry this burden; a registry edit can often fix a problem on a Windows server but, if deployed incorrectly, can kill the entire system. This risk ties the administrators' hands - forcing them to find an alternate, less risky, solution.
A common trick is to make a backup copy of a configuration file prior to making changes in order to allow an easy fallback point if things go wrong. Since virtual machines are just a collection of files, administrators can take this technique and kick it up a notch - the entire server can be copied prior to making changes. With VMware snapshots, every aspect of the current state of the machine is saved including the memory state, the hard drive state and the virtual hardware state. Creating a snapshot frees an administrator to attempt any fix desired. In a worst case scenario, the administrator simply rolls the virtual machine back to the point where the snapshot was taken and every aspect of the machine is reverted. Windows, or whatever operating system runs the virtual machine, has no idea that those changes were ever made, and the administrator is free to try something else.
Administrators can freely size and resize their virtual servers, giving each server the hardware that it needs without wasting resources on a server that won't utilize those resources. Administrators can make changes to their virtual servers with the confidence of an absolute fallback point. VMware's VMotion allows the IT staff to upgrade or trouble-shoot servers in the middle of the day with no disruption to user sessions. They can simply VMotion the virtual machines to another ESX host, troubleshoot or upgrade the original host, then VMotion back.
A task such as migrating from Exchange 2003 to Exchange 2007 is typically difficult and expensive to test effectively and inevitably incurs risk. In a virtualized data center, an administrator can snap a copy of the production Exchange server, and then do the migration in a test sandbox that duplicates the production environment. Even the network can be replicated by utilizing different NICs for VLANs. VMware's rollback capabilities enable reversion to earlier stages of the migration if problems are encountered. Once the migration has been successfully completed, the administrator can point to the snapped server as the new production box. This process eliminates much of the cost, time and effort...and all of the risk of the migration. In fact, this virtual QA process is so easy yet beneficial, that administrators use it for essentially everything of any importance. Test cycles are shortened while hardware conflicts, configuration and set-up periods are all eliminated.
Patching servers in a physical data center is dangerous because administrators can never be certain that individual server hardware or driver disparities won't cause a new patch to break the server or application. Virtualized data centers enable the application of a server patch during the middle of the day with 100% confidence of no downtime because if it worked on the test server, it will work on all of the servers. The servers are, after all, identical - provisioned from software templates maintained on a SAN library. VMware's Update Manager further automates the process of patches and updates, including patch tracking and management. It includes options to snapshot virtual machines prior to patch application, enabling administrators to rollback the virtual machine to a known state if the patch has unforeseen side effects. Update Manager can even patch offline virtual machines without exposing them to the network.
Capacity Planning and Server Budgeting
A virtualized data center is not only much less expensive than the physical equivalent, it is easier to plan from a capacity standpoint, and to budget. The IT staff can use VMware VCenter to view server resources as an aggregate pool of CPU, memory and I/O. Utilization trends may show, for example, that aggregate server utilization will rise to 70% in 6 months - and therefore indicate a budget requirement for a new server at that time.
A virtualized data center affords better purview into the "Core Four" Resources: CPU, Memory, Network I/O, Disk I/O. This in turn enables administrators to more easily tune their environments to enhance performance while minimizing hardware resources.
Other Virtualized Data Center Administration Benefits
Data center virtualization extends beyond just servers - it includes storage, network, security, disaster recovery, desktops and applications. A virtual infrastructure provides a unifying platform with interdependencies among each function that enables a much more efficient and elegant architecture. VMware continues to add new capabilities to this environment that go far beyond anything possible in a physical realm. Its growing suite of automation products, for example, provide enhancements such as workflow approval for server provisioning, slashing time to market for software development and automating data center failover and migration. VMware's new View 3 VDI solution enables a whole new dimension of simplifying IT administration by virtualizing both the desktops and the applications, then moving them to the data center.
Virtualized data centers are benefiting network administration as well. The Nexus 1000V virtual switch appliance will debut in the spring of 2009. It enhances the operations of both the physical and networking infrastructures. The Nexus 1000V enables greater efficiencies between the server and network teams by assigning Cisco service, security and operational policies across each VMware virtual machine. It provides network administrators with all the features that a physical Ethernet switch enables including QOS, Tagged VLANS, rate limiting and debugging commands.
IT Staff Converts
Despite management concerns, a properly envisioned, designed and deployed virtualized data center reduces demands on the IT staff. Moreover, they inevitably become fans. This is evidenced by the 12,000 enthusiastic attendees at VMworld this year, even though it was only the 5th such event. Virtualization lets administrators move away from the drudgery of routine server issues and midnight maintenance windows. They can instead spend more time on learning and implementing interesting technology that adds true value to their organizations.
Written by Steve Kaplan & Jason Coleman