IT Chargeback, Trackback - The Time is Right
Joe Onisick’s blog post, IT Chargeback/Trackback: Yes, You Need It, at Networking Computing, provides a good argument for understanding and utilizing an IT chargeback/trackback tool set to assist in the deployment and operation of private clouds.
In my opinion, the most important part of virtualization technology and cloud computing is one of the least often discussed: a transparent system that shows LOBs (Lines of Business) what their use of the infrastructure costs. With a proper chargeback system and fully transparent cost model, LOBs can understand what portion of hardware and personnel resources they’re consuming. Ideally, by exposing pricing, IT organizations around the world can enter the free market and compete with public cloud providers. Amazon publishes its EC2 prices on the internet, why don’t we?
But to do this, we need different tools. Domain-specific management tools (i.e. traditional tools) typically have no capability to peer through the virtual infrastructure into a domain beyond their own. Embracing these tools now will definitely help companies adopt cloud computing practices. In the case of Amazon EC2, companies are charged based on bandwidth, memory, and CPU, so developers need to build applications with these factors in mind. Similarly, Microsoft Windows Azure charges based on compute instances and resource usage. These are measured, and if they detect an application consuming extensive resources, the application will be put into an isolation mode, which costs more. Ultimately, we need to be aware of the ‘noisy neighbor’ problem that’s exposed in these environments.
Due to the sensitivity of overhead put on VMs, it’s important these days to adopt agent-less management. It’s even more important to manage your infrastructure as you adopt cloud computing solutions. Agent-less harvesting of utilization and response times that provide cross-domain mapping gives insightful performance correlations—not at a single point in time, but over the course of time, accounting for all peaks and troughs in application workloads.