Virtual Iron: Dynamic Infrastructure for the Data Center By Ann Ernst published: Tuesday, October 04 2005
Virtual Iron Software, Inc. provides an enterprise-class virtual computing and management platform enabling customers to achieve the highest levels of datacenter utilization, flexibility and significantly lower capital and operating expenses.
VSM spoke with Mike Grandinetti, Chief Marketing Officer and Alex Vasilevsky, Co-Founder and Chief Science Officer, about what’s new at the company.
VSM: What does Virtual Iron do that’s different from other virtualization software companies?
MG: The word virtualization has taken on a lot of ambiguity. Many people think about it in very basic terms, primarily single server partitioning and data center consolidation. That value proposition has driven a lot of the activity in the market to date.
TCO benefits do accrue with that, some opportunity to reduce your footprint in the data center and some operational benefits. But you can only take it so far. We’ve had an opportunity to do a lot of market research in this space, in individual meetings with customers and focus groups in all of the early adopter industries in both Linux and virtualization technology.
What we learned has reinforced the founding premise of the company: There are a lot of gaps in managing data center resources and virtualized environments. Virtual Iron’s goal is to help organizations realize the vision of an on-demand or utility computing-based data center infrastructure, what we call “dynamic infrastructure”.
In many cases, companies want to bring a utility computing model to specific applications or workloads. Virtual Iron is the only company that can help enterprise IT departments manage these workloads across multiple systems by bringing the appropriate data center computing, storage and network resources to the application as needed, without pausing or stopping these applications.
We do basic single server partitioning, just as VMware does. We can partition a single box into multiple virtual machines, run multiple workloads and drive utilization of that box to 70% or 80% or higher. When you introduce single server partitioning you may eliminate the utilization problem, but you introduce more complexity by running all the virtual machines.
VSM: Companies are realizing that they may have stopped physical server sprawl, but may be dealing with virtual server sprawl. That may be why we’re moving toward policy-driven management – data center management, instead of just putting together silos of virtual machines.
MG: As people run into virtual server sprawl, and feel that they can’t move from the physical to the virtual world and into production, they’re asking us to supply our basic policy-based management capabilities to a more heterogeneous virtual world and bridge them from the lab into production.
There’s a tremendous opportunity to make a contribution and alleviate those pain points. Based on some early acceptance and traction, and on discussions with the market about their challenges, we’re working to help people go from a homogenous to a heterogeneous environment.
VSM: Can you talk about the policy management?
MG: It’s about being able to set parameters to bring compute power, including additional CPU and memory in addition to storage and network capacity, at specific times to specific workloads. You can set performance utilization or degradation parameters, so if a workload gets to a specific percentage, you can automatically adjust capacity to that server. You can set timing parameters to manage peak usage. Policies can also be set to automatically respond to hardware failures. These are just a few examples of the types of policies that can be applied.
AV: The policy engine is built into our management environment. Through the user interface one can specify what action to take on what triggers and events. The events, triggers and actions can be flexibly defined by the end user, based on time or time and workload for example.
The product ships with a number of default pre-packaged policies installed, like a high availability policy that end users can implement with the click of a mouse. If there’s a hardware failure within one physical server that’s running all the virtual servers, the virtual servers will be migrated and started up on another piece of physical hardware.
We see our product evolving very rapidly in distributed workload management and more advanced policy-based management.
VSM: It sounds like the product would be useful for Disaster Recovery and Business Continuity.
AV: Those are some of the policies we ship by default. And we see adding distributed load balancing across multiple physical servers running many different virtual machines, moving things around to rebalance your virtual data center to run more efficiently.
MG: Disaster Recovery and Business Continuity is an important capability, because very few people are satisfied with their existing high availability solutions. They’re often redundant and therefore very expensive.
Rather than twoN clustering, where you over-provision and add a redundant server to back up each server running a workload, we’ve introduced n + 1 clustering. You’d have a standby machine to fail over to multiple servers simultaneously, leading to a dramatic reduction in capital expense.
VSM: Is provisioning and deployment of virtual machines also policy driven?
AV: Both are policy driven. Anything one can do in the user interface, one can create a policy for. You can specify the virtual server requirements and operating system, and then specify a time to deploy the virtual server and have it boot up.
VSM: How does this compare to VMware’s VirtualCenter?
AV: It’s a very different product. VirtualCenter is designed to manage VMware virtual machines, in particular ESX Server.
We designed our product to be heterogeneous and to include features that manage not just Virtual Iron virtual machines, but also Xen, and to manage operating systems, users, policies and hardware.
VSM: Can you talk about the storage side of the product?
AV: You can dynamically discover the whole SAN infrastructure without powering up a single virtual or physical server. There’s a button on the management console that finds all the LUNs you’re allowed to see. The administrator doesn’t have to manually type those virtual HBA numbers into the SAN network to see the actual disks for every virtual server.
And you can pick up a virtual server and move it to another piece of physical hardware without reprogramming your SAN infrastructure. The worldwide node names are completely transparent to what hardware they’re running on.
We do the same thing for the network – discover what networks a virtual server has available to use, and have the MAC addresses go from one place to another.
VSM: When you’re setting provisioning and deployment policy, can you include storage?
AV: Absolutely. Everything is available to the end user, and not just through the user interface. Our system includes a sophisticated Java-based scripting engine, so the user has a choice to go one level below the user interface and write scripts.
VSM: What about iSCSI?
AV: We are planning support for that later on in the product. It is on our roadmap. First we are supporting SAN and Ethernet, because very few of our customers are asking for iSCSI.
VSM: Right now this is an enterprise-class product?
AV: Yes.
VSM: Only on Linux?
AV: The product is composed of two pieces, the management environment and the virtualization layer. The management environment runs on all standard operating systems, such as Windows and Linux. The virtualization layer today supports Linux as its guest operating system. It is on our roadmap for next year to deliver a Windows solution.
VSM: Can you talk about your recent announcement?
MG: We’re a young, venture-backed company, on a mission to take on some challenging problems. By definition that gives us the opportunity to spend a lot of time talking to early adopters. We continue to hear about virtual server sprawl, sneaker-net and manual labor, and about difficulties in getting out of the lab and into production.
In the financial services industry, where we’re seeing the Xen open source virtual machine monitor get some early acceptance, people are asking us to bring Virtual Iron’s policy-based automated management capability to other virtual environments.
We’re now providing support for the first of a number of heterogeneous environments, including Xen, which will be available in 4Q05. The Xen management module will be put into open source and tightly integrated into Virtual Iron capabilities.
We’re helping organizations take a single consistent, cohesive view of the entire data center, and to apply the virtualization capability across the data center to abstract and separate the virtual layer from the physical layer.
When we think about virtualization it includes all the server components – CPU, memory – as well as the network and storage. You have an infinitely divisible, sharable pool of resources to bring to bear on specific workloads that can be managed from a policy-based, automated console. You can fully align the needs of the business with what IT is doing to support the business.
*****
For more information about Virtual Iron, visit www.virtualiron.com.
|
