Google last week announced that it has started using version 2.1 of its Andromeda software-defined networking stack.
Why are we bothering to report that?
A couple of reasons. First, it's fast. Google says it “reduces network latency between Compute Engine VMs by 40% over Andromeda 2.0 and by nearly a factor of eight since we first launched Andromeda in 2014.”
Other clouds don't often discuss this stuff. They all do software-defined networking (SDN), but Google tries harder to makes a virtue and a differentiator of its efforts in the field. Google's also recently added added “Dedicated Interconnect” to its cloud, permitting networks that span on-premises and cloud infrastructure.
Second, like many SDN tools, Andromeda has its roots in Stanford University's Open Networking Research Center, a facility that gave the world OpenFlow, Nicira and NSX, plus many other SDN outfits
Another alumnus of the center is Big Switch Networks. The Register met Big Switch's co-founder Kyle Forster at the Sydney OpenStack Summit this week and he said should the Andromeda team cast their gaze in his company's direction, they would be flattered to see the extent of their influence on its products.
Because so much of the SDN community has worked in similar circles, it is small and tight.
So there's a decent chance that whatever Google's done to Andromeda between version 2.0 and 2.1 will filter out into that community.And not many months afterwards, into products that become available to you.
That's exciting because SDN is a welcome alternative: remember how Electronic Arts' tech director told us he now considers networking vendors a better source of lunches than networks? Forster told The Register his company is being welcomed by Fortune 2000 companies for similar reasons, as users tire of being tied down by networking vendors.
And if 40 per cent reductions in latency result from the SDN industry's observations of Google's new toys, why wouldn't SDN be welcomed? ®