On reading about Cato Networks announcing its $2.5B valuation, I couldn’t help but think back to a recent performance test we ran on the Cato backbone. I think it encapsulates why investors and enterprises have such high hopes for this company.
A manufacturer came to us looking to move away from its MPLS service. Their global network spanned some 150 locations worldwide and you can imagine how much their MPLS charges were running. Aging Riverbeds at each site meant they would need to spend upwards of $1.2 million just to upgrade their WAN optimization appliances.
But like so many of us, cost wasn’t the only issue for getting away from carrier services. They were fed up with delays and the support issues. They also wanted to be nimbler, more efficient. Users were working from home; apps were moving to the cloud. MPLS’s day had passed. Which meant, of course, evaluating SD-WAN and SASE vendors.
Testing Backbone Solutions
As part of that process, we had to test various backbone technologies that would replace MPLS. One approach considered was just the Internet, which was recommended by numerous SD-WAN vendor. Cato came in as well, recommending the Cato Private Backbone that’s built into their SASE service, the Cato SASE Cloud. (Check out our SASE page to learn more about Cato’s unique approach to SASE.)
So, we ran iPerf tests from Singapore to Shanghai and Amsterdam to Singapore across all three networks: MPLS, DIA and Cato. The results were outstanding for Cato. On average, Cato had 3,400% better throughput than the Internet and 1398% better than MPLS, with performance increasing with distance. Cato also outperformed the Internet and MPLS on the path between Amsterdam and Singapore:
Global Backbone Throughput Testing Using IPerf in Mbps
Singapore to Shanghai
Amsterdam to Singapore
Built in Optimization is the Key
Overall, Cato’s performance results were so compelling that my client could avoid their Riverbed refresh and pocket the $1.2 million investment as well as the license fees. Not a bad start towards a compelling ROI.
The big question, of course, was why did iPerf perform so much better over Cato than the Internet. Notice I say the Internet and not MPLS.
The Internet’s crazy routing and unpredictable loss and latency metrics wreak havoc on global connections. They also would explain the significant swings between average and max throughput numbers (more than 200% in the case of Singapore to Shanghai). By contrast, both MPLS and Cato provide private, low latency, low packet loss global networks with stable performance.
So why did iPerf perform so much better across Cato than MPLS? Part of it was capacity. You can simply afford more capacity with Cato than you can MPLS. With Cato, our average pipe was 200 Mbps. With MPLS, we only had 20 Mbps between locations.
But capacity and a stable global network with well-designed routing was only part of the story. Cato also builds in optimization into its network. It proxies the TCP connection on the ingress and egress giving the clients the impression that a destination isn’t across the world but just around the corner, as in just 20ms to their PoP.
Finding a Better Cloud Backbone
With all of that in mind, we went to market looking for another cloud backbone with built-in optimization. We checked out offerings from Megaport, Equinix, AWS, and Azure. All of them stable networks with lower latency and loss than the Internet, but none of them include the kind of optimizations we found with Cato. That optimization makes all the difference.