Glenn Dekheyser

They were using private interconnection.

The HPC group wasn't the network group.

We didn't have the network group involved.

Probably should have.

And they just weren't working.

So it turned out that the connectivity between the compute, that whole VPC, and our on-prem, when you did a traceroute on it,

it was taking upwards of 300 milliseconds of latency.

Is that bad?

It's not good.

It's not good at all.

When you're at Equinix and you have your storage in the same region, cloud adjacent, you're typically talking, let's just say less than five, but usually less than two milliseconds.

So you're talking about 300.

That's getting ugly.

But the Amazon was showing all green lights and the NetApp stuff was all showing green lights.

but nothing was working.

The customer had no idea how their network was set up.

The network group, when we finally brought them in, really didn't have much of an idea of how cloud networking worked and how it was routing even inside their own network.

And then getting out to Amazon, they didn't understand where they were connecting to from Amazon.

And to make things worse, it was going cross region twice in Amazon, which created an enormous cost.

They got a bill that they didn't know where it was coming from or why.