IT Management Articles

A New Way to Measure Cloud Metrics – Review of Librato Metrics

by Rakesh Sharma
Published on 26 January 2012

Just as we expect telephone service to be available 24/7/365, we increasingly expect Internet or Intranet based services to be available without interruption.

To provide this quality of service, Software as a Service (SaaS) and Platform as a Service (PaaS) providers have to be able to detect and respond instantly to any unforeseen events that may affect their service, and resolve problems before anything breaks.

This means that they have to be able to monitor all aspects of their operation that can alert them to potential service degradations and help them find and fix the root cause of problems.

Librato Metrics is designed for use by DevOps (a new function in IT organizations that bridges the gap between development and operations and a major movement, specifically in PaaS and SaaS companies) and operations teams to monitor their SaaS and PaaS infrastructure and applications.

It enables them to store any and all metrics they need to track the health of their operation, to configure alerts, visualize the metrics in graphs, build dashboards, and perform root-cause analysis. We will look at its interface, features, and see how it can be of use to you.


Librato spun out of Virginia Tech University with a Checkpoint/Restart product for High Performance Computing. "In 2010, we changed direction and used our technology to launch the "Silverline" Application Resource Management service. After numerous requests from customers we followed that up in 2011 with the launch of the Metrics service," says Fred Van Den Bosch, CEO of the company, "we saw an urgent need with SaaS and PaaS providers for a uniform environment to manage all of their metrics, and a similar need in the emerging "Internet of Things", where sensor networks are expected to generate volumes of metrics that far exceed those of the Internet today"


What differentiates Librato Metrics from more traditional monitoring services is that it is a "platform" for storing, visualizing, correlating, and acting on any and all metrics organizations collect. This means that all of its capabilities are accessible through its RESTful API's, allowing its users to send it all the data they collect, be it with existing open source or commercial collection agents, environment-specific collection agents, or directly from applications that are instrumented for data collection.

Since Metrics is delivered as a service, it can be deployed in minutes, and does not require any hardware and software resources to be deployed. By using Metrics for monitoring data, DevOps and Operations teams need to be familiar with just one way of storing and visualizing data, and setting up alerts. By viewing all of their metrics through a single set of graphs and dashboards they can more easily correlate between metrics from different sources.

Metrics are grouped based on the sources they originate from. For example, sources can be server instances or application instances, and CPU usage and memory usage can be metrics associated with a server instance. In addition, we were told that it will soon be possible to us tags to organize your sources in whatever way you like and configure graphs with metrics you want to see together. You can also configure alerts to send a message or invoke an action when thresholds on a metric are crossed. Alerts can be tied to standard service hooks like email, Pagerduty or Campfire, or custom Webhooks you set up yourself.

To collect data, you can either use Librato's library of open-source collection agents, write your own custom collection agents or send data directly from your applications through language bindings. Currently, Librato supports Statsd, Collectd, Rack-statsd, and Sensu collection agents and language bindings for Clojure, Python, Go, Ruby and Java.

Van Den Bosch says that SaaS and PaaS companies can also embed Librato Metrics in their service to allow their users to monitor their use of the service. The Metrics APIs allow them to either just use the storage capability, the graphs, or the full dashboards offered by Metrics, depending on their needs.

Van Den Bosch says that the application is especially useful for companies that use a combination of open source, commercial and custom built tools to monitor and manage their SaaS, PaaS or IT infrastructure and companies that deploy sensor networks. "With its library of open source collection agents and language bindings, Metrics makes it possible to have a unified environment for storing visualizing, analyzing and acting on all data that needs to be monitored," he says. "Our users can focus on using their expertise to build solutions for their customers, and can leave the management of the monitoring data to Metrics."


With its surreal black and green interface, this one's a winner. For a technical application, such as this, a simple interface can considerably help non-technical people like me. And, I was not disappointed. Each task or action is not more than two clicks away. What's more, the interface is task-driven. This means each click is rationed to perform a task. I especially liked the fact that the interface gets right down to business with the attributes and alerts tab.


The application comes with a comprehensive knowledge base. In addition, customers can access their live chat room, where they can interact directly with support and development engineers.


Yes if you are interested in measuring data and run a SaaS or PaaS service, then this is an invaluable and scalable service to help you track and measure service data.


Apps mentioned in this article