I’d like to follow up this week’s announcements of VMware’s IT as a Service strategy and VMware vFabric by zeroing in on the challenges surrounding cloud application performance management, and how vFabric Hyperic can help you meet those challenges.
As our CEO Paul Maritz mentioned at yesterday’s VMworld keynote, our industry has hit a tipping point where virtualization has surpassed the physical computing paradigm, due to a need for IT to quickly respond to dynamic business needs at ever-improving price points. To ensure even greater agility and value, the next destination for our industry is cloud computing. Making this shift requires a pragmatic, evolutionary approach that leverages investments in existing architecture.
Yesterday, Rod Johnson, SVP of VMware’s Cloud Application Platform Division, described how VMware vFabric is a key element of enabling our customers to reach this next destination of cloud computing. Delivering the agility and value promised by cloud computing requires a new kind of application — cloud applications — that have their own unique characteristics:
- Dynamic architectures
- Elastic capacity
- Extreme scalability
- Open choice
How do cloud applications change performance monitoring?
These characteristics of cloud applications bring new requirements to application performance monitoring. For instance, dynamic architectures and elastic capacity imply a datacenter defined by constant flux, with pools of hundreds, even thousands of VMs continually being started and stopped, vMotioned, reverted to snapshots, and so on.
This blistering rate of change is a natural outgrowth of responsiveness to business needs. But it is impossible to manage with manually maintained, complex configuration files used by legacy monitoring tools. The only way to get ahead of it is to use a monitoring product that can automatically discover changes to your entire application infrastructure — everything from the application code itself to the vSphere host. Hyperic Autodiscovery does exactly that, updating itself within moments of app infrastructure changes.
The extreme scalability required by cloud applications requires a lot of virtual machines – which leads to a firehose of performance data. For instance, a typical Hyperic customer will collect a million performance metrics per minute. It’s not at all hard to get to this volume of metrics, since we have a number of customers running 1000 (or more!) virtual machines, each with a Hyperic agent collecting about 1000 metrics. So, even though we’re not talking about Google-level scalability, application performance data become a performance problem in itself if not managed properly. Thankfully, Hyperic is engineered to handle high volumes of application performance data.
Open choice means that cloud applications can be built from a wide range of components. For instance, as Rod pointed out yesterday, cloud applications might use WebSphere, WebLogic, JBoss, and our own vFabric tc Server, as well as public cloud platforms, for their Java application server tier. It’s critical to have a monitoring tool that supports a range of popular web application technologies out of the box. But no one tool can support every conceivable technology. It’s understandable – there’s no way one vendor, no matter how innovative, can keep up with the thousands of other innovative companies in our industry. So in addition, it’s critical to have a monitoring tool that makes it easy to build custom monitoring plug-ins, and makes its built-in plugins open source to provide a wide range of code to reference and leverage. Hyperic delivers on this front, providing approximately 50,000 performance metrics for 75 web application technologies, as well as open source plugin code and a fully-supported plugin API that has been used by third parties to extend the range of technologies monitored by Hyperic.
What’s next for Hyperic?
We’re demonstrating vFabric Hyperic live at VMworld, in the main VMware booth. Come see for yourself how Hyperic adjusts to changes in virtual infrastructure in near real-time. Later this month, we’ll demonstrate Hyperic at Oracle Open World in San Francisco, September 19-23, and after that, we’ll be at Spring One 2GX in Chicago, October 19-22.
We’re committed to making Hyperic the leading choice for monitoring cloud applications, with their inherent dynamicism, scale, and openness, and we’re looking forward to working with you — our open source community, our users, and our customers — to make this happen.