First Impressions: Scalent V/OE
With Scalent, management of a mixed virtual and physical data center can be a one-person operation.
- By Peter Varhol
Part of the appeal of virtualization is that it enables us to abstract much of the complexity of the hardware and network infrastructure from day-to-day data center management. You can detail images to different VMware ESX servers, bring online new instances of virtual Web servers in response to peak demand, or move images from failed hardware to new systems.
In general practice, there are a lot of limitations and exceptions to this highly effective abstraction. For example, you can transfer virtual machines (VMs) from one VMware server to another, but only as long as they're part of the same subnet. You may be able to save images in the event of a data center failure, but reconstituting them on servers in the hot backup site can be a long and tedious process. You may have successfully virtualized many of your software stacks, but the Linux and Solaris servers have you stumped.
Limitations such as these are among the reasons that virtualization is taking time to penetrate data centers, and why there are practical limits to how far many are able to take it. Today, you typically see small pockets of virtualization in the enterprise, and growing use in individual applications and hardware types, but rarely is there a comprehensive and overarching strategy for using virtualization in all IT roles.
This is where Scalent comes in. Scalent V/OE (for Virtual Operating Environment) provides a means of abstracting the virtual and underlying physical data center away from its details. It lets you easily move "persona" -- full disk images and virtual machines -- between different machines, and to configure those images in ways that support the infrastructure. Setting up and managing a virtual and physical data center can be easy with Scalent.
Scalent installs on a controller system on the network. This controller acts as a traffic cop for mapping out new servers, switches and other physical and virtual devices on the network, and for configuring and displaying those devices.
For demo purposes, the full Scalent controller software can be installed on a single computer, along with a simulator that simulates its interaction with the actual network. The UI is Web-based, so you can access it from your local host in that configuration. In a live network, the controller is also the Web server for the UI, so you can access it from any desktop system with a Flash-enabled Web browser.
[Click on image for larger view.]
|Figure 1. The Scalent console is easy to use with drag-and-drop capability.
I used the UI in conjunction with the simulator, and also with live hardware and images. The simulator shows servers and other physical and virtual network devices as you've set them up in the UI. You can bring servers down, transfer payloads from one server to another-even across network boundaries-and a variety of other activities. The only difference is that there's no hardware at the other end. The only apparent difference with the live hardware is the server startup and shutdown times were much faster on the simulator than in real life.
The controller software installs without the need for any manual intervention. Once you have the software installed, you can add physical devices individually to the layout so that V/OE will recognize them. Servers must have the ability to be remotely turned on and off. You define devices, then configure those devices to perform a network boot. That way, Scalent can power up a server through its controller. The server requests a network boot, and the controller responds by booting the requested persona. That persona can be any of the stored server VMs or disk images, from an Apache Web server to a SQL Server database. And the server operating system can be any OS that supports (32-bit or 64-bit) or Solaris SPARC architecture.
Switching an image, or even groups of images, between servers, or loading new images onto servers, is deceptively easy. From within the V/OE management console, you simply visually choose the VMs and click a couple of buttons to load them onto any given server.
You can do much of this with ESX today (at least in part because Scalent has submitted code to the VMware community that's included with ESX), but the environment has to be entirely ESX to do it. That might work in some enterprises, but as virtualization matures, most environments will likely be a combination of ESX, Microsoft Hyper-V and various versions of Xen.
Much Assembly Required
V/OE requires quite a bit of effort, especially at the beginning, where you're defining the physical characteristics of your data center and building up your catalog of persona and devices. You have to be cognizant of each physical box, switch, rack, chassis and persona (a unique image of a software stack), and represent them individually in a Visio-type environment. It can take hours or even days when you add in the need to reconfigure each of your servers for network boot.
Fortunately, once you've done this legwork, you can generate the representation of that environment into a script language. You can load the configuration from the script, take snippets of script (which is highly readable in an XML-like format), copy and paste it to other scripts, modify script information, and place it into executable code that can automatically perform repeatable actions. The script console provides a way to build new configurations or activate personas on the fly using a command-line interface.
You can also use the management console to make adhoc changes in the configuration of a given server or application. If a Web server goes down, for example, you can bring up and assign an image to a new one within a few minutes (in my tests, even a cold start to fully engaged server never took more than two or three minutes). That's more difficult with something like a database server, but if you have the right level of backup and redundancies, you can still re-target a backup image or physical box to a live box with barely a hiccup.
Scalent has overcome many of the technical limitations inherent in broadening the scope and depth of virtualization in the data center. What remains as obstacles are how we think about our infrastructure, the servers and their payloads, and cultural attitudes surrounding physical servers and our mission-critical applications.
To successfully enable a virtualized data center, both IT and line-of-business owners have to get over the concept that their application maintains a static presence on a given server. Virtualization abstracts the payload from the hardware, and Scalent enables that physical or virtual payload to move around anywhere, locally or across the world. You really have no idea where your application may be at any given time.
Scalent also requires a fairly large-scale commitment from the data center. All parts of IT have to be in agreement as to the purpose and goals of this level of flexibility in the data center. In a sense, the persona defines the datacenter, not the box. There has to be an IT culture that internalizes that concept and works to that end. But if you combine an organization commitment to virtualization, the use of Scalent V/OE, and this type of culture, I don't doubt that the result will be an amazing level of flexibility in the data center.
Scalent Virtual Operating Environment (V/OE)
Palo Alto, 650-424-1222
Peter Varhol is a principal at Technology Strategy Research LLC, an industry analysis and consulting firm.