FJ: When you talked to BMC last year we were rounding out our offerings to have full life cycle management to help customers plan their VMware consolidation efforts using our BMC Performance Assurance. Over the course of the last year, we’ve introduced BMC Performance Manager for Virtual Servers to deliver operational monitoring and management of Virtualized environments, and we recently extended our long-standing BMC Performance Assurance for Virtual Servers solution to include predictive performance modeling for Virtual Server environments.

Together, these solutions complete the operational and management lifecycle approach where customers can monitor servers they already have, understand what opportunities for physical or virtual consolidation might be, as well as comprehensively evaluate how their applications would perform if and when virtualized via our modeling capabilities. This ensures that customers can “right size” their VMWare environments to ensure they can support their application workload resource requirements.

We’ll also have some additional announcements around products that will use end-to-end performance statistics to drive VMotion and move things from one physical server to another.

FJ: Yes. If you look at virtualization today, it’s still done primarily at a system level, virtualizing and putting an application in a virtual box. Server virtualization is the primary focus, along with scale out virtualization for clusters and Web farms.

Even with load balancers based on system-level traffic, the decision-making is not looking at how the business services are performing and if resources are aligned based on the needs and priorities of business services.

We’re going to provide a solution that shows how your business service is performing, and also looks at the system end to see if you are getting the most out of the resources, if there’s a potential performance impact, we’ll prevent the impact from occurring.

FJ: Even though there’s been some over-hyping of utility or on-demand computing, I believe it’s inevitable that customers will want to align their infrastructure to optimize services and utilization.

It’s where we’re targeted now, the pragmatic steps toward the utility computing environment. A typical customer environment is running at 10-20% utilization rates. We’re looking at how we can help you get from 10-20% utilization to incrementally step up to 50-60% utilization quickly and easily, and ultimately approach “mainframe class” levels of utilization, say, 80-90% or more..

The more you get out of it, the fewer human resources you need to manage it, so you can deploy those resources to implement new systems rather than managing old systems. It helps free resources for the business without having to expand your infrastructure.

That’s what I mean by pragmatic. If you can do that by reallocating servers appropriately through static provisioning then implementing targeted dynamic provisioning, you can start moving things fluidly based on business needs. You can have a shared pool of servers and swap them in and out based on failures or on performance requirements.

Another ripple effect is that with this more dynamic environment, how do you figure out not just utilization, but per cost allocation? Even for internal use, you want to have some proof of how much you’re spending on your application or business service. We have products coming out to help customers figure out those kinds of things in a more dynamic environment.

FJ: Especially in a virtual environment. If you have four or five images stacked on a server, and those images are getting different levels of resources during the day because there’s not a static dedication of CPU time, how do you measure that and then charge back? And then how do you do that for a three-tiered application?

FJ: What’s going to change is the concept of element management. BMC is helping customers manage growth, so we’ll use agentless monitoring to keep track of the state of images. You want to know the overall state, and when something is going to impact the service you’re trying to provide. Then you want to very quickly address that impact before anything bad happens.

In today’s world, do you consider it a bad thing if a server fails? With virtualization, a server failing is bad at a micro level, but as long as it doesn’t impact the business, you don’t necessarily care in real-time that it failed. You just care in real-time that the service is performing, and you accumulate the bad occurrences and then fix them over time. Virtualization helps manage failure if you have the proper tracking tools.

FJ: The products that work with VMware are named for virtual servers. The core online monitoring comes from BMC Performance Manager for Virtual Servers which we introduced earlier this year. It covers operational monitoring and management of VMware, but we’re anticipating that as customers start implementing Microsoft Virtual Server or Xen in production, we’ll start adapting our technology to Xen and Virtual Server too.

For over a year, we have had BMC Performance Analyzer which was the data collection and analytics tool that collected and analyzed performance data for VMWare environments. In October we created BMC Performance Assurance Suite for Virtual Servers, by adding predictive analytic and modeling capabilities to BMC Performance Analyzer. The predictive analysis capabilities allow you to understand – in advance – what configurations will be required to support VMWare, as well as to optimize performance of VMWare environments in an ongoing fashion..

FJ: Yes, so you can get an inside-out and outside-in view, element-level or image-level data collection, and also collect at the server or physical level, and then correlate the two.

Just as an aside, we rebranded our classic PATROL solutions in this area, so classic PATROL is now called BMC Performance Manager. PATROL Perform and PATROL Predict are called BMC Performance Assurance for Virtual Servers. We’re trying to move away from sub-brands like PATROL and move towards saying what the product does.

Those are the primary products. We’ll be announcing new capabilities around helping to implement data center virtualization. That will provide policy-based orchestration for virtual environments.

The BMC Virtualizer Family applies to data center infrastructure. There are three different scenarios that you can look at. The first is high-availability. Customers have spent a lot of money implementing high availability, and they have a lot of redundancy built into the infrastructure.

With our high-availability offering, we’ll be able to provide failover from a shared pool. You have the shared pool of servers, you have a failure on Exchange, for example, and we’ll automatically swap in a server from the shared pool to fix the failure.

If you move up the stack a bit, customers have clusters all over in a typical infrastructure, but they’ve provisioned the clusters based on peak needs. They don’t share resources across clusters. If you could pool and share into those clusters, you could have a lot less static allocation of resource to a given cluster. When you start seeing a performance impact, you can allocate more systems into the cluster to prevent it from happening.

FJ: We think the ideal is where you’re being more predictive. Using our current capabilities we can help you identify business cycles like payroll and end-of-month or end-of-quarter processing. But there are other natural business cycles that occur that may be less obvious, and we can help identify those cycles and translate them into policies.

Then you do just-in-time provisioning of resources based on this predicted cycle. Rather than a reactive operation, you can be more of a predictive operation. And if things work right then it’s a very boring system. Our goal is to make management of IT a very boring exercise.

FJ: This is where we see some opportunity for an ecosystem. The policy and orchestration are pretty generic, and we see opportunity to interface with storage provisioning. We’re already interfacing with networked devices, like big IP load balancers.

We’ve taken a more holistic view, where customers don’t want to just provision a server, they want to orchestrate its participation in a production environment, which may include provisioning additional storage resources.

FJ: We think the evolution is a trend toward utility computing. Customers are concerned with the services they’re providing, and they want to look at that service level and only be told when they really need to react and respond.

That’s a different management paradigm than just reacting to changes at the element level. We think that’s going to become a lot less important, because if you can dynamically move resources based on business service needs, you can get another level of optimization.

In fact, virtualization creates a great opportunity to do snapshot capture, so when there’s a failure you can snapshot the app, do a failover process so it’s running in a new space, and still keep the old one so you can do offline diagnostics to figure out what caused the problem.

FJ: A little of both. Last year there wasn’t a lot of VMware in the production environment. Now when we say ”virtualization” or “VMware” they ask, “What have you got?”

FJ: We have an input section and we can take data from any source - BMC Performance Manager of course, but non-BMC inputs also - to help drive the policy. On the orchestration side, which is the action side, it’s an open interface so we can work with partners to build more interfaces.

FJ: Yes. The enterprise-level customer is BMC’s sweet spot. Virtualization is key, but what’s driving it is data center consolidation, server consolidation, and the flexibility or adaptability to be more competitive.

DW: In the context of the virtualization space, we’re announcing a full life cycle solution for management of virtualized environments.

DW: We’re now at the point where 80-90% of all server shipments are based on industry standard server technologies with AMD or Intel in their core. Because of the technology directions that AMD and Intel are taking with multi-core CPUs and with the OEM community with blade servers, these so-called “commodity” technologies now offer the ability to sustain enterprise-class processing. But, it’s a tremendous waste of these technologies capabilities to contine to dedicate them to the “single application, single server” approach. It is highly wasteful.

We view the success of VMware, Xen, and Microsoft Virtual Server as a technology that is enabling industry standard servers, which are most of the market, to become the platform for enterprise computing going forward. Virtualization is one of the key enabling technologies that’s making that possible.

Until now, it’s been very difficult to run applications in a truly consolidated fashion on these architectures, because you don’t have isolation, which results in applications basically stepping on each oather. Further, you haven’t had the ability to manage applications the way you could on higher-scale proprietary UNIX environments or on classical mainframes.

That’s one opportunity. Also, we have had solutions to manage and optimize virtualized environments for 20-plus years, and we saw an opportunity to leverage that experience managing and optimizing mainframe, AS400 and VAX/VMS virtualized operating environments..

With virtualization these environments are going to change very dynamically. There’s going to be a need for change management and change control processes, and for operational management solutions. We came up with what we call Service-Oriented Resource Management (SORM).

DW: SORM is a strategy and a methodology that consists of a variety of solutions, each of which can be adopted modularly and in any order, to realize a point product or point solution value. And, they are part of an overall life cycle management.

There are key obstacles to accomplishing virtualization that represent organizational and operational paradigm shifts. The ability to change has gone from days or weeks to minutes or seconds, which implies a need for automation. The ability to move resources to shared pools and service-based pools of resources, quite simply, requires virtualization. And, if multiple applications are going to virtually share underlying hardware resources, this requires capacity management discipline.

The final dimension is business processing and intelligent automation. Processes have to move from being stable and defined to dynamic and based on capacity management. SORM moves the automation of virtualized environments to minutes and seconds and makes capacity management values available to customers in near-real-time.

DW: Absolutely. Certain environments should be hosted physically. How do customers identify which applications are candidates for virtualization? Once they’ve identified them how can they identify what best technologies to host them on, how can they ensure that they’ll deliver service levels in an ongoing operational environment, and reduce the cost while doing it?

You have the commoditization of server platforms and virtualization technologies available on these platforms, but the business risk of not doing capacity planning still remains. It doesn’t matter how cheap your hardware is, if you don’t have sufficient capacity, you’re going to violate your service levels. If you have too much capacity, you’re going to waste money. Finally, response time and throughput - service levels - don’t directly relate to resource utilization.

Capacity planning has to become real-time, and automated provisioning of resources needs to be driven by intelligence and foresight, not reactively. Service-Oriented Resource Management allows customers to set up business policies that will deliver just-in-time the right amount of computing resources to applications and do it based on business importance.

That’s the marriage of technologies, taking our long experience in predictive performance management of physical and virtualized environments, and marrying that to automation technologies that will provision the resources.

DW: If you have a set of applications running on a set of physical servers, we can predictively model what will the response time of those applications be if you move them to a virtual environment, and what hardware configurations would be required to do so. And we can factor in the business cycle and historical behavior.

DW: The key one is BMC Virtualizer, the new software family that allows customers to automatically orchestrate and provision resources on demand. There are two products within it, BMC Virtualizer for High Availability and BMC Virtualizer for Capacity on Demand. Both provide the ability to implement business policies to orchestrate and provision resources, and both allow customers to consolidate and cost-optimize their data center and operational availability of applications.

The distinction between them is that BMC Virtualizer for High Availability is designed to reduce redundancy in the failover environment. BMC Virtualizer for Capacity on Demand is an upgrade from High Availability that extends the provisioning capabilities to create a pool of servers that can be shared dynamically across any number of application clusters. It can be invoked to deliver incremental capacity whenever there is a performance situation that would require it.

DW: In HA you’d create one pool for each application, but those pools couldn’t be shared. With Capacity on Demand, if you had three different racks running three different applications, all three of those different application clusters could share the same pool of excess capacity.

The most efficient way to do that would be to have the additional pool of capacity that’s there for either failover or capacity be set up using something like VMware pre-loaded with application images for any application that might need it. You could then do the provisioning in a matter of less than five minutes. Naturally, we can also support provisioning of bare-metal environments, where we basically grab a physical server platform, provision an OS and application instance on it, and turn it on. This naturally takes longer, typically around 30 minutes. Again, modularity and choice are keys here.

The other major new feature in this release is the extension of our existing performance assurance solutions to include predictive capabilities for distributed virtualized environments. We look at how it will perform if you move from physical to virtual; what happens if you stack multiple applications on the same hardware, running VMware; whether they will contend for the same resources at the same time; what the response time and throughput will be.

DW: You run the product, gather the performance data, and then it automatically builds the models. You can evaluate a model and run it in minutes to get your answers. If you have a different question, or want to change the scenario, within a minute you can have your answer, without it doing any benchmarking or any test labs.

DW: We acquired the Virtualizer product technology last December from Corosoft. It was in beta for a fairly long time. We worked with a lot of customers, and as part of an acquisition, it came with a set of customers.

BMC Performance Assurance has been around for many years. Our 7.3 release of BMC Performance Assurance for Virtual Servers was introduced a year ago, and we started a beta in March with the predictive capabilities. We worked with beta customers worldwide to make sure it was meeting their requirements.

The output of BMC Performance Assurance for Virtual Servers also turns out to be an excellent basis for creating a policy for automation. One approach to automating the delivery of more resources for capacity on demand might be to monitor performance, and if utilization exceeds a certain rate, to automatically add another CPU. We can take it to the next level by discovering what and when are the needs for resources. As part of our performance analytics, we can discover the underlying business cycle behavior and know when repetitive needs for performance and capacity are going to occur, and then give that as a policy to Virtualizer.

DW: Exactly. You’re doing it based on intelligence and analytics, as opposed to reacting. Eventually, prior to provisioning a resource, we will be able to run a predictive model in real time that shows if the resource you’re about to provision is going to fix the problem.

We have the initial integrations that build your policies based on intelligence, and the direction is to automate based on predictive analytics. And to round it out, we have Virtualizer integrations with existing BMC products in other solution domains. BMC Performance Manager, formerly known as PATROL, is our monitoring and management solution.

Another integration is with our Marimba Software Configuration Management and Discovery Solutions. BMC Virtualizer can now orchestrate the provisioning of appropriate software using Marimba solutions. You can use it to make sure you’re provisioning the right version with the right patch levels. You can maintain your operational standards and change control policies. And by doing that, you get integration with the CMDB automatically. That keeps our configuration items up to date, which is critical to ensure conformance to Change Management processes..

DW: You can use just our automation solution, or just our performance solution, or just our configuration discovery solution. They deliver intrinsic value on their own, but by integrating them you get the synergy.

DW: Solaris on Intel? It’s supported in Virtualizer. SORM has Solaris on Sparc, Windows and Linux on Intel.

DW: We sell the components of SORM both as a software suite that customers can purchase or rent, and as components for performance analysis, performance reporting, and capacity planning. There are several ways they can purchase the value components of this solution.

Customers can buy our software licenses, or they can lease them effectively as a product license. Or if a customer doesn’t have staff, we will sell them a remote-hosted subscription service. It’s an open and modular approach.

DW: It’s creating a virtualization layer above the resource layer. We don’t care if the servers are physical or virtual. We will also span the networks and storage, so if an existing business service application requires additional capacity on demand, out of a virtualized pool of resources we can reconnect it to additional storage, additional server resources, talk to the load balancer in front of it, to make sure that transactions are rerouted in real time.

We can deliver capacity on demand, or what we call service self-healing. If an application server fails, we can provision another one in near real-time to automatically pick it up out of a shared pool of resources. The advantage is that we can drive out a lot of over-provisioning of physical resources. And because you’re introducing automation, you’re getting rid of the potential for service interruptions that previously existed.

DW: The resources that could be provisioned could be VMware instances on an existing server, or they could be physical servers, or both.

DW: Exactly. We can go physical-physical, virtual-virtual, or any combination. The best availability solution for customers is when they set up a shared pool of server resources running VMware, and then put the different application environments in their virtual machines.

You’re running a hot-standby environment of shared resource on top of VMware, so when there’s a failure, or when you need capacity, you don’t have to go through the time delay of bare metal provisioning. We can orchestrate bare metal provisioning if that’s what customers want to do, but that takes longer.

DW: This could work in a customer environment with no virtualization software. We could go in with our solution and instead of 10 servers active and 10 passive failover, you can have 10 active and two passive failover physical servers, and if any of those 10 fail, we will bring one of those two failovers into service.

With SORM we’re adding a new layer of virtualization above the entire server application layer. We’re doing the virtualization at a business service layer, because we’re not talking about just the servers. We will reconnect that running application to the SAN and NAS, the storage side of the equation, and we will reconnect it to the front end with your network load balancers.

Customers have varying degrees of implementational maturity, different IT pain points. Some don’t know what they have in their infrastructure, they don’t have an audit of their servers or what the servers are doing. They can yield ROI very quickly by using discovery solutions to catalog their servers and the software licenses running on those servers.

Some customers know what servers they have, but don’t know the utilization. They might want to do some performance baselining with performance analysis capabilities. They might just be surprised to find a lot of excess capacity which can be repurposed to drive better efficiency and agility.

A customer might be ready to move to a virtualized environment, and want to evaluate the performance impact, in terms of response time and throughput, of moving existing applications to a virtual environment. They would use predictive modeling to size those things.

Maybe they’re into production now, and want to drive out cost and increase service levels by automatically provisioning resources. They can start at any point in the cycle and yield value at that step, or they can proceed to the next step.

If one assumes that adoption of these virtualized environments will lead to more dynamic configuration change, like VMotion and orchestrated provisioning of resources on demand, those will cause changes in the configuration much more frequently than traditionally existed.

Then it’s even more critical to keep a centralized configuration management database (CMDB) up to date as the audit trail the entity IT uses to ensure that everybody’s operating from the same information.

DW: Virtualization is going to drive that. Once you get into virtualization, how do you keep track of which virtual machine is running which application at what time of the day? If you don’t do that, you can’t troubleshoot, and you can’t audit. The availability and use of a CMDB-based solution is a key to successful adoption of virtualization technologies.