Countless hours are spent evaluating, testing, migrating and building cloud ready and cloud native applications. The new era of cloud computing is driving data center migrations and application development, reluctantly managing operations management. Without a doubt, there are many new avenues to pursue server monitoring from a cloud prospective  Let’s take a look for a moment, at the similarities between the two. The valuable experience that has been gained with years of experience of monitoring traditional infrastructure may still apply, and can be used to inform the new age of operations AKA cloud monitoring.

1. Beyond Compute

First there was physical equipment, running at capacity, costing time and resources – requiring an increase in resources. , . Later, virtualization increased scale and on-demand resource allocation for critical applications. Then, the cloud burst with cleansing rain promising server-less compute, high availability across geographic regions, and most importantly lower barriers to entry and as needed resource allocation. All the while, the application responsiveness and end user experience remained the key to success across these architectures. The compute resources themselves come from any combination of these stages, but what about the application?

While compute has long been the key to accurate performance management, it’s critical to take an extended view of the environment to what’s being run there. Be it SQL server, Apache webservices, or cloud resources like Office 365, having the application view of resource consumption and any latency associated with it completes the picture. Having the agentless and agent based ability to capture those critical metrics is key.

2. Hypervisor Perspective

In much the same way, cloud infrastructures today function as a large virtualized environment. Historically, NOC teams have relied on the agent based function to capture information on bare metal systems. That same window into reality exists in cloud platforms today in the form of AWS Cloudwatch, or other cloud ready monitoring API. We wrote about how critical the agent based approach above, and both strategies have their advantages. By centralizing Cloudwatch information, you’re able to accurately see performance, and calculate metrics such as CPU steal by leveraging both agent and agentless monitors.

3. Database Monitoring

Cloud solutions inherently lend themselves to new service development and broad availability. We’ve seen these advances with AWS and Azure alike, with services such as RDS or SQL on Azure. These services share the supporting infrastructure behind them, making traditional monitoring mechanisms difficult if not impossible. Capturing the “hypervisor perspective” then becomes key. Once again centralizing that information enable the distribution and centralization of that information, such that it can be viewed alongside traditional metrics.

We talked a little about, traditional and new approaches, as well as key areas to keep in mind. I hope you’ve found this post helpful, and I look forward to our next blog.


Skyler Cozad is a Sr. Presales Consultant at CA Technologies

Leave a Reply