What you don't think about when it comes to cloud services

When looking at expanding hosting services and offering cloud infrastructure, service providers think about how to offer automated provisioning, but sometimes neglect to think about the technical issues that later become a hindrance to meeting SLAs and bringing new customers onboard.

When doing dynamic provisioning of resources, it is difficult to foresee the amount of monitoring capacity that will be required at any given time to analyze performance, availability, security, changes and business services—all of which take considerable processing power in a dynamically provisioned virtual environment. So when systems are being provisioned and come up and move due to resource contentions or other issues, there is a bursty need for processing and storage capacity for monitoring event messages (operational event log data).

The bursty nature of the operational data used to monitor and analyze to meet SLAs and other requirements often means existing capacity is insufficient, which causes companies to fail and drop events, putting them at an operational risk and in a position of contractual liability. If service providers don’t adequately plan for how to manage bursty operational data, they will experience negative impacts on their service delivery capabilities.

But this is a challenge to resolve, as companies want to pay only for what is required; they don’t want to buy excess capacity that is only used some of the time. So what can a company do if existing monitoring platforms lack the capacity to accommodate a true pay-per-use model because they don’t scale enough or support bursts in operational event messages?

To be able to maintain a burst of event log data without dropping or limiting events, or the analysis of events, a software-driven approach is needed. Putting forth the type of money and effort required for excess licenses and capital infrastructure, as well as resources to configure those systems, would become too burdensome in terms of money and time.

This is exactly the type of solution AccelOps today announced it will apparently demonstrate at the Gartner Data Center 2010 conference next month.
In essence, it is proposing what it calls a “virtual cluster architecture” that offers “elastic monitoring.” AccelOps is attempting to go beyond providing a virtual appliance and onto providing an architecture that accommodates bursts without forcing service providers to change their licenses and management of all different components as things change on the fly.

That can be a boon, if it works, for service providers that want to avoid having to buy excess capacity. They of course would rather buy on an “as-needed” basis, which is what this proposes to allow.

Whether unlimited scaling or quick implementations for bursts in monitoring traffic will truly be accommodated with this demonstration is hard to know until we see it in action, but it would be really “cool” to see a monitoring infrastructure that can accommodate the alerts, changes and constant vicissitudes of monitoring cloud environments without requiring enormous and repetitive investment in hardware.