Two networking industry veterans have launched a network observability startup based on an open source model, with users paying for a fully supported enterprise version if needed.
Dinesh Dutt, former chief scientist at Cumulus Networks and a Cisco fellow who helped design many of the networking giant’s ASICs, and Neela Jacques, whose background includes VMware, Barracuda, the OpenDaylight Project and Isovalent, have co-founded Stardust Systems based on Dutt’s work on the open source Suzieq network management tool.
Jacques, whose experience includes the OpenDaylight, eBPF and Cilium open source projects, sees an open source business model as the best way to get their technology out to the market. His goal is to get Suzieq into the hands of as many users as possible, even if they never become paying customers.
“Some users will never pay for additional functionality, services or support,” Jacques told Enterprise Networking Planet. “We think that’s the price of entry these days. We’re committed to giving users something solid of value, something they can use in production. We’re betting many of them will want more and happily pay for the enterprise version.”
Networks have never been more complicated or harder to understand — witness Facebook’s recent outage — and many network operators are afraid to make changes as a result.
“When a network fails, the outcome is spectacular,” Jacques said. “Most network operators don’t want to admit how fragile their networks are.”
Also read: The Future of Network Management with AIOps
Current Network Monitoring Tools ‘Siloed, Reactive’
Modern networks are essentially “spaghetti soup,” Jacques said. “No individual can understand it,” and no tool can see it all either.
Jacques and Dutt aim to change that. Current network monitoring and management tools are “siloed, reactive, unhelpful, device-centric,” and largely focused on SNMP and flow data.
“Everyone’s doing flow data,” Jacques said. “No one’s doing operational state data the way we are.”
Jacques outlined the limitations of monitoring tools:
- They are siloed, with no uniform way to correlate across logs, metrics and operational states.
- They are reactive, responding to problems rather than preventing them.
- They don’t help with validation, troubleshooting, and change management, or the gap between applications and the network.
Network Observability to the Rescue
Suzieq starts with a list of devices and credentials — discovery isn’t automated yet — and then implements Dutt’s vision of “network observability,” which he literally wrote the book on, Cloud Native Data Center Networking. Dutt also co-authored the TRILL and VXLAN standards.
Suzieq supports devices from Cisco, Arista, Juniper, Sonic and Cumulus. It helps admins “confidently make changes,” target network improvements, and recover quickly from failures and outages.
Suzieq analyzes network architecture and interactions with servers and applications.
The network observability tool isn’t based on a database, instead using the Apache Parquet columnar storage format for better performance.
The network management tool worked so well that users began asking for support and enhancements, and the result is Stardust Systems. Early users include the managed service provider (MSP) division of a Fortune 500 company, a top 10 ecommerce company, a hosting company, an investment management company, and an engineering firm.
Jacques said networking hasn’t had the same innovation as other enterprise systems.
“The network isn’t getting the attention it should,” he said. “There hasn’t been the competition and innovation.”
Suzieq is a step toward bringing dependable automation to networking, Jacques said.
“No one’s done a good job of this from the networking side,” he added. “You can’t avoid automation. It needs to happen. But more automation means increased blast radiuses. To automate confidently, you need better observability. Automation without observability is madness.”
A network engineer at a large global MSP that’s using Suzieq in production shared his views on the tool with Enterprise Networking Planet. “Network observability is a big issue for us,” the engineer said. “We manage networks for some of the biggest companies in the world and needed a better way to identify problems. Suzieq has been fantastic for us, and we’re super excited about their vision and roadmap.”
Read next: NetOps vs DevOps: Bringing Automation to the Network