ltstudiooo - Fotolia
Cumulus Networks is the latest software-based networking vendor to address the lack of tools for collecting real-time performance data from switching fabrics found in the data centers of large enterprises.
Cumulus released last week a product called NetQ that's aimed at companies using the vendor's Linux-based network operating system. Cumulus' 600 customers include a third of the Fortune 50 and many other large corporations.
Enterprises with massive data centers struggle with network monitoring because of the paucity of tools that can continuously collect and aggregate operational information from switches and servers, said Shamus McGillicuddy, an analyst at Enterprise Management Associates based in Boulder, Colo. Today, the majority of companies use network monitoring systems that poll devices every few minutes using the Simple Network Management Protocol (SNMP).
Shamus McGillicuddyanalyst, Enterprise Management Associates
"SNMP polling is imperfect because it pulls data at intervals," McGillicuddy said. "When you're relying on a pull mechanism, you can often miss events that have a real impact on network health and performance."
The delay in data gathering is a big problem for companies running large cloud-based data centers. Those facilities run hundreds of switches and need tools that can provide immediate feedback on the impact of configuration changes, which frequently occur in those environments.
NetQ data-collecting agents
Cumulus attacks the problem by running data-gathering agents on switches running Cumulus Linux 3.3 or higher and servers running Ubuntu or Red Hat Enterprise Linux. The agents stream information on all events that could affect network state to a Redis database. Cumulus provides the database software on a virtual machine and the customer buys the hardware to run it.
Through a command-line interface (CLI), network operators can get answers to a wide variety of questions. For example, they use NetQ to locate containers running on servers and see which ports are open. Engineers can also check whether network overlays and switches are configured correctly, determine the protocols in use and see whether any devices have crashed.
The collected information lets network managers locate a problem faster and then playback events to determine the cause. Also, operators that use IT automation tools, such as Ansible, Chef or Puppet, can make configuration changes to switches and immediately see the impact. If the modification causes some switches to fail or operate improperly, then engineers can return them to their prior state.
Cumulus is not the first software-based networking vendor to provide customers with real-time telemetry data to spot, diagnose and fix problems. Others include Arista and startup Pluribus Networks.
In 2016, Cisco introduced a network analytics engine, called Tetration, that also performs similar functions. However, the system is "orders of magnitude" more expensive than the other offerings and only works on some Cisco Nexus 9000 switches, McGillicuddy said.
"Basically, the most elite customers of each of these vendors are asking for something like this, and all of them are responding," he said. "In the future, you can expect this technology to trickle down into the mainstream."
Analysts expect more network monitoring tools to hit the market because of the growing number of companies moving applications to cloud environments, such as infrastructure or platform as a service. The shift to the cloud will result in service providers spending roughly $26 billion this year in IT infrastructure, according to IDC. That amount, plus the $16 billion companies will spend on private clouds, represents a 15% increase from last year. Spending on non-cloud IT infrastructure is expected to fall by about 5%.
Avoiding security risks when deploying an SDN controller
SDN's role in cloud computing
SDN use cases in the LAN, WAN and data center