Open Employment Positions | Technical Blog Site Search Virtual News   Escaping the Data Center Matrix By Steve Kaplan, AccessFlow LLC   You are reading this because you know something. You can’t explain it, but you feel it. You’ve felt it your entire career working with Intel servers. There’s something wrong with a world in which each new application requires a new server, where provisioning cycles take weeks, where scheduled downtime increasingly must be done on nights and weekends, and where cost-effective business continuity is only a dream.   You can take the blue pill or the red pill. If you go for the blue pill, the story ends. You go back to the familiar complexity of server infrastructure - and you can continue believing in the inevitability of endless Intel sprawl and its associated inefficiencies. But if you take the red pill, there’s no turning back. Virtualization will rock your world and rationalize your data center. Its sheer elegance will never again let you look the same way at the unwieldy, inconsistent and very expensive physical alternative.     Problems with the Physical Computing World Now that you’ve taken the red pill, you will know that what you’ve been feeling is the data center matrix. It is all around you. It grew organically as IT organizations shifted away from mainframe environments to PC-dominated architectures. As IT professionals dealt with problems of resource contention and vendor support, they learned to deploy a single application for each operating system. The resulting endless growth in Intel servers consumes an ever increasing amount of rack space, power consumption, network connections, host bus adapters and IT staff time.   The matrix has lulled IT into believing that long provisioning cycles are an inherent component of ordering new servers. When a business unit needs a new application, the associated new server must be sized to handle worst-case loads. It must be specified and bid out to suppliers. Once it arrives, it must be set up and configured with the organization’s best practices. Corporate users have consequently become conditioned to expect weeks of waiting before their applications can be deployed.   Server consistency is difficult as well. Small discrepancies between even supposedly identical servers combined with the difficulty of utilizing policies to manually build a server results in variable server builds. This variability in turn makes patch roll-out more difficult because of the increased probability of a patch unexpectedly breaking a server type.   In the matrix, application development and testing requires an isolated environment in order not to corrupt the existing IT infrastructure. Developers often require several servers in order to validate applications on multiple versions and service packs of Windows operating systems. In order to start a QA cycle on an application, they need to configure the test servers from scratch – meaning the set-up and configuration period can be extensive. Challenges in validating compatibility between servers other elements of the infrastructure make it difficult to create realistic test environments.   Troubleshooting production servers can be very challenging in the matrix because of the difficulty of replicating the exact system parameters and hardware environment at the time of failure.   IT personnel accept the matrix’ illusion that maintenance windows grow smaller while their need for scheduled downtime grows larger – they believe the only solution is to work nights, weekends and holidays in order to minimize user disruption.   Providing high availability can be a very expensive endeavor requiring duplicate servers and expensive load-balancing or clustering software. Disaster recovery is similarly challenging because the increasing interdependencies between servers frequently mandates maintaining duplicates of all production servers, not just mission critical servers, at a remote DR site. In addition to the capital expense, it is extremely resourse intensive to continually update DR servers with the hardware updates, patches and drivers to keep them perfectly synchronized with their production server counterparts.   When resources begin running low, IT professionals must purchase more servers, more network switches and more storage - in the matrix there does not seem to be any other way.   Perhaps most-surprisingly in our cost-conscious times, management accepts the matrix illusory call for ever increasing IT budgets. This demand shows no signs of abatement due to increasing multi-tier applications, multiple generations and editions of Windows servers, rapid growth of Linux and escalating demand for more storage.     Virtualization - Down the Rabbit Hole The matrix is beginning to unravel as IT organizations around the world wake up to the realization that data center practices so long believed as essential are no longer relevant. Virtualization has irrevocably altered the IT environment.   Studies show that Intel servers have an average prime shift usage of only between 5-10%. Virtual infrastructure changes this dynamic by enabling many virtual servers to run on the same physical hardware. ESX Server accomplishes this by placing a thin layer of software code on the physical Intel server which interfaces directly with CPU, memory, storage and network resources. Multiple operating systems can then run on top of this virtual infrastructure which allocates these resources to enable optimum utilization.   Each individual operating system, along with its applications, runs in its own virtual machine which looks just like the hardware server for which it was designed. These virtual servers are guaranteed specific amounts of processor cycles, memory access, disk I/O and network bandwidth. When a server requires more or less resources, the allocations can be changed, or VMoton technology can be used to move it to another physical server with no disruption to the user sessions.   Virtual infrastructure blows apart the central tenant of the matrix – that new applications or greater resources demand requires new servers. Hardware management is now treated separately from software management, and hardware can be viewed as a single pool of processing, storage and networking power – which can be allocated to various software services on the fly.   With VMware VirtualCenter management software, an administrator can manage thousands of Windows NT, Windows 2000, Windows 2003, Linux and NetWare servers from a single point of control. Administrators can manage a versioned library of server templates that can be used to instantly provision new virtual servers.     Virtual Infrastructure Benefits IT organizations tend to be resistant to change which, since it has a perceived risk. So why should you disrupt the physical world you’ve mastered in order to deploy virtual infrastructure?   Reducing costs   Organizations around the globe are slashing IT hardware costs by up to 53% and reducing operating costs up to 79%. Total cost of ownership is dropping by up to 64%. Organizations who virtualize therefore gain a significant competitive advantage. Cost reductions result from:     Reducing capital expenditures for servers Elimination of server support contracts for all virtualized servers Reducing required rack space Reducing power consumption Reduced staff time requirements on server provisioning and management Reduced costs trying to manage latency from network traffic Much less expensive business continuity options Much less expensive application availability options Reduced storage costs     Instant server provisioning Instant server provisioning is a key benefit of virtual infrastructure. VirtualCenter enables administrators to select a “gold” template for a new server deployment from a library of standard server templates, and deploy it to the hardware pool in seconds. IT organizations can implement just-in-time server provisioning schemes to allow business units to provision their own servers when needed. And all servers running in the environment match the current best practice for security and configuration. Troubleshooting becomes easier, and the likelihood of an accidental open port, or vulnerable service left active decreases to near-zero levels.   Elimination of maintenance window concerns   IT no longer needs to negotiate hardware downtime windows for business unit applications. Hardware requiring maintenance is emptied of running virtual machines using VMotion, then repaired or upgraded and put back into service.   Troubleshooting and patch roll-out management   Snapshot copies of running production systems can be taken at any time for debugging or patch testing and worked out without taking the server down for maintenance. These are exact duplicates of current production, including the virtual hardware layer. And with the checkpointing and rollback capabilities included in virtual infrastructure, virtual machines with patches that fail in production can be instantly rolled back to the last known good state IT managers also have the security of knowing that if a patch does not break one server type, it will not break any of the others.   Superior development environment   Developers can have multiple versions of Microsoft Windows, Linux, NetWare and DOS on the same server and can move between operating system environments without repartitioning or rebooting. Several developers can simultaneously access a virtual machine. This approach results in significantly improved fault isolation because each virtual machine operates independently of other virtual machines which run on the same physical hardware. Any communication between virtual machines must be established explicitly. A complete network of distributed server virtual machines can be run on the same physical machine thus simulating a network. Crashing one virtual machine dos not crash other virtual machines.   Superior testing environment   A library of virtual machines with standard test environments and tools for ach OS and network configuration can be instantly deployed to any physical machine with spare hardware capacity. Virtualization thus allows testing teams to both eliminate hardware resource conflicts and to significantly shorten test cycles by eliminating configuration and set-up periods. By maintaining staging environments in virtual machines, IT departments always have servers available to conduct pre-deployment testing and to ensure compatibility of the new application with the rest of the IT environment. Once the testing and staging is completed, the virtual machine can be easily moved into production with no user downtime.   Dramatically enhanced high availability   Application availability is ensured by utilizing Microsoft or other clustering software to cluster virtual servers on one ESX server with virtual servers on another. This enables real-time server fail-over in the event of hardware or software failure without having to double a large number of physical servers.   True disaster recovery / business continuity   The disaster recovery site can utilize ESX Server to enable many virtual servers to be recovered on one physical box. This makes it inexpensive to set up an active/passive disaster recovery site utilizing very little hardware. And because the servers are all virtual, there is no need for synchronizing the DR site servers with those in the production environment.   Reduced network traffic   Servers with interdependencies such as a Web server communicating with an application server which communicates with a database server, no longer need to tie up network resources. Once virtualized, these servers now communicate with each other through the VMware virtual switch on the server. Communication speeds between the virtual machines are thereby increased while overall network traffic is reduced.   Reduced backup and restore times   By virtualizing a backup system, the virtual servers are now backed up as single files and placed on the SAN or tape library. This reduces the bottleneck of network latency and consequently enables faster back-ups. Restore times are also accelerated as single large files are recovered rather than individual small files with date and time stamping requirements.   Reduced requirement for network devices   In a physical world, each server requires at least two HBAs (including one for redundancy). With virtual infrastructure, however, each physical server can share HBAs among all of the existing servers, thereby reducing network costs. Similarly, the need for storage switches is reduced or eliminated.   Improved resources management   The complexity of the data center matrix requires substantial up front planning in order to get server resources right. Virtualization lets you think about a pool of hardware resources instead of a bunch of servers. You can guarantee specific amounts of CPU, memory, I/O and network bandwidth to the server. If a server requires more or less resources, you can change the allocations or use Vmotion to move it to another physical machine   Improved server administration   The data center matrix leads to disparate management systems across Windows, Linux and NetWare servers. VirtualCenter lets you monitor and control all of your virtual servers from a single “dashboard”. Threshold and event alerts can be configured for server failures and over/under utilization among other events.   Enhanced storage optimization   Each server in a physical environment requires its own LUN in the SAN. This inevitably leads to inefficiencies in the SAN which can no longer pool storage resources across servers. Virtual infrastructure changes this by allowing one LUN to be shared across two clustered ESX servers. In this manner, storage capacity is maximized and administration is simplified.     Escape from the Matrix Virtualization means IT professionals can now scale up with CPUs and memory rather than struggle to budget how many new servers to buy. It means no more weekends and holidays performing scheduled server maintenance. It means economical high availability for all servers, not just critical machines. It means buckle your seat belt, Dorothy, because Kansas is going bye-bye.     Copyright 2006-2007 Access Flow Inc. All Rights Reserved.