By David Roth, CEO and Co-Founder, AppFirst
As managed services providers (MSPs), you are keenly interested in ensuring your customers’ servers, storage, and network infrastructure are humming along efficiently. After all, you have made significant investments in infrastructure monitoring tools so you are aware of problems well before they impact your customers. You have probably also invested considerable effort in developing scripts to help gather additional metrics. But most MSPs do not expose this level of detail to their users. Why?
However, your users could benefit from valuable insights residing in their systems, application, and business metrics. Today's application developers/owners have a responsibility to their customers, and/or their management, for the performance of the applications running on your hardware. Especially in Dev/Ops circles, application owners want to know everything they can about the moving pieces that make up their application, whether or not they are "responsible" for that level of metrics.
Traditionally it was assumed that to keep the servers and networks humming along, the MSP would monitor the hardware, but the client was responsible for monitoring the application. This is sometimes enforced in a contract, with limitations in place to prevent the client from seeing low level metrics. Users can’t install their own low level monitoring tools because of adverse impact on other users of a shared infrastructure. But the growing numbers of savvy application owners know how valuable this data is to their business, and they want access to it.
At the same time, there is often a technical limitation in place. Most traditional monitoring solutions assume only one group is interested in performance metrics. Even "advanced" solutions claiming to support multiple "personas" assume these people all work for the same company. Allowing clients to see "their" metrics requires you to install a separate instance of the monitoring solution to support each customer. In some cases, this would mean setting up a second tool, because you can’t install two agents from the same company on the same server.
The Growing Partner Ecosystem
There has to be a better way — and savvy MSPs are already doing some strategic partnering. For instance, leading Platform-as-a-Service (PaaS) provider Engine Yard works with us as part of its new platform, enabling the Engine Yard operations team to keep an eye on server availability, watch system logs, and collect important polled data from sources like Nagios, JMX and WMI. Multi-tenant capabilities enable separate access to the same agent for Engine Yard customers. Here, they can see the same metrics, or add application specific features like statsD data and New Relic feeds.
Active State and Rackspace have both tapped New Relic to enable users to monitor application performance. ActiveState’s Stackato users can now, for no extra cost, easily monitor Java, Ruby, PHP, and Python applications and get real-time performance data about transactions, applications and websites. Rackspace has added New Relic to its Cloud Tools Marketplace so users can monitor every detail of their apps from end user experience, through servers, and down to the line of code. These are just a couple of examples, and strategic partnering will get busier.
Virtualization Blind Spot
Today's MSP tends to be very virtualized. One of the traps in multi-tenant virtualization is a concept known as "stolen time." If a shared infrastructure has five virtual machines running on a single host, the hypervisor decides how to allocate resources to the virtual machines. The guest OS is virtually unaware of this and traditional host level monitoring will be oblivious to the effects…except that users’ applications will appear to run very sluggishly. Since most MSPs prevent the user from seeing the hypervisor layer, they could be totally unaware of this problem and point fingers at the MSP — increasing support costs and impacting customer satisfaction.
At AppFirst, our developers don't really trust other OS developers. As a result, we independently track things like stolen time for our customers and learned that once high levels of stolen time start to appear, they don’t go away. The MSP should not be blamed for this — sometimes the hypervisor gets hung or confused. The best thing the user can do is simply kill the instance and spawn another. But without partnering strategically with a performance monitoring expert, there is a lot of finger pointing, wasted debugging time, and slow application performance impacting your revenue-generating customers.
Today’s shrewd application performance management (APM) and data center management vendors are partnering together, weaving technologies into an integrated, elegant, best-of-breed solution. By working together successfully, we can directly and positively impact application performance for MSPs and their customers.
David Roth, CEO and co-founder of AppFirst, brings more than 20 years of experience leading organizations and leveraging partnerships for building successful companies. He sets the AppFirst strategic direction and guides the day-to-day business execution.