Using a cloud network monitoring service to monitor a cloud platform sounds redundant, but the organization that oversees Alberta’s CyberaNet research network has found it works.
Cybera Inc., a not-for-profit, has been using Boundary Inc.’s service for keeping an eye on application traffic between nodes in OpenStack-based clouds it has been creating.
One is a pilot for several Alberta institutions running classrooms in the cloud for online learning using the open source Moodle platform, which Cybera calls a learning management cloud (LMC).
The University of Alberta, the Edmonton-based Northern Alberta Institute of Technology (NAIT) and NorQuest College are currently involved, and are about to be joined by Athabaska University.
Like all its projects, the LMS effort has a limit: There is a three-year agreement to run the network, in part to show OpenStack is reliable and can be used to build a useful infrastructure.
Open source tools like Puppet are used to manage the physical infrastructure and Chef to manage the virtual machines.
“We can create a new production environment for the LMC in about 10 or 15 minutes,” said LukeTymowski, a Cybera systems administrator who helps run the LMC.
But automatic and real-time monitoring of the system is essential because the LMC operates 24 hours a day, so students take a lecture any time.
Tymowski first came across Boundary on the Internet last fall, then learned more about the company in March at an open source monitoring conference in Boston.
An agent is installed on every physical or virtual server and captures application flow –Netflow, Sflow and IPFIX — which it sends back in the cloud to Boundary for almost real-time processing and analysis. Network and application status can be viewed by customers online through dashboards. The service can also send alerts.
According to Chris O’Connell, Boundary’s director of marketing, the agents can poll devices every second.
After trying it on a personal server, Tymowski installed Boundary on a Cybera system, and a month ago received the okay to do a one-year trial.
“Without a tool like Boundary, you know that your app is up and running, but don’t know how well it’s working,” Tymowksi said. Flow data can be analyzed, he said, but it can be hard to trace a problem back to particular hardware or virtual machine.
Boundary uses a protocol that collects some 200 types of traffic for analysis.
Early on it identified an LMC problem related to a University of Alberta backup server that was consuming more bandwidth that had been expected. It wasn’t a critical problem, Tymowski said – changing the servers’ configuration resolved things. But, “if it wasn’t for Boundary we wouldn’t have easily noticed that.”
O’Connell acknowledged that Boundary isn’t for everyone. Governments and financial institutions have shied away from it because they don’t want put any network data in the cloud, he said.
Boundary says its service has complete visibility across any infrastructure irrespective of the underlying language and technology stack used, including private, public hybrid clouds.
The company says it also has off-the-shelf integration with two monitoring tools, AppDynamics and New Relic. Through support for RESTful APIs, it can be integrated with others.
Coming soon, it says, is an Active Flows metric for tracking the number of in-use connections or interactions between hosts.
The service is priced by the amount of flow data the customer transmits a day. It’s free for up to 1 GB a day, and US$495 a month for 5 to 25 GB a day. More than that – or, for unlimited data — the company will give custom pricing.