By: Aaron Spink (aaronspink.delete@this.notearthlink.net), January 26, 2017 8:37 pm
Room: Moderated Discussions
RichardC (tich.delete@this.pobox.com) on January 26, 2017 10:47 am wrote:
> Maybe that depends on whether you're defining "current gen" from the perspective
> of a hardware designer, i.e. "what are we designing and selling in Jan 2017 ?", or from the
> perspective of a user, i.e. "if I provision a cluster on AWS in Jan 2017 what network
> bandwidth will I get between the VMs ?".
>
I describe it was FB had already publicly commented on as their infrastructure as of 2 years ago with intention and need for even faster networks. FB and MS as part of their public OCP infrastructure were already at 40 Gbe in 2015 for new installs.
> Maybe that depends on whether you're defining "current gen" from the perspective
> of a hardware designer, i.e. "what are we designing and selling in Jan 2017 ?", or from the
> perspective of a user, i.e. "if I provision a cluster on AWS in Jan 2017 what network
> bandwidth will I get between the VMs ?".
>
I describe it was FB had already publicly commented on as their infrastructure as of 2 years ago with intention and need for even faster networks. FB and MS as part of their public OCP infrastructure were already at 40 Gbe in 2015 for new installs.