Google says it’s going to open-source its Maglev load-balancing tool to developers, in a move that will also provide a boost to its own infrastructure.

Google engineers Daniel Eisenbud and Paul Newson explained in a blog post that Google has a long history of building its own networking gear, including its own load balancers, which have handled traffic to Google’s services since 2008.

They claimed that Maglev, which is named after Japan’s super-fast bullet train, is able to serve a million requests every second without any “pre-warming”.

Google explains in a research paper that Maglev runs on commodity Linux servers and can be setup without any specialized rack deployment. That’s because Maglev uses an Equal Cost Multipath (ECMP) approach for distributing network packets equally to Maglev machines in a cluster, which use connection tracking and hacking techniques to ensure that all packets are sent to the right destination. Should a Maglev system become unavailable, the rest of the machines in the cluster will compensate by handling the additional traffic.

This ECMP design optimizes the load-balancing capacity of Google’s networks in such a way that they’re far more cost-effective than traditional hardware load balances, which use an active-passive configuration that means half of all available capacity is reserved for failover purposes, the company said.

“All Maglevs in a cluster are active, performing useful work,” the Google engineers said. “This N+1 redundancy is more cost effective than the active-passive configuration of traditional hardware load balancers, because fewer resources are intentionally sitting idle at all times.”

Maglev is powered by Borg, Google’s internally-built cluster management technology that lets engineers switch service workloads between different clusters as they see fit, allowing for better utilization of unused capacity. It’s similar to how Google’s cloud platform is able to move workloads between different regions and zones, the engineers said.

“Recently, the industry has been moving toward Network Function Virtualization (NFV), providing network functionality using ordinary servers,” they wrote. “Maglev is proof [of] how NFV can be leveraged to enable easier addition and removal of networking capacity. It also shows how NFV approaches can be used to enable additional networking services without the need for new and custom hardware.”

Image crdedit: Katherine Lawrence via pixabay