Looking at Single TCP session throughput performance between Amazon AWS regions and comparing it to the throughput between a similar setup across the Internet.
Arista EOS is based on a Linux kernel and provides full and open access to a Linux shell, allowing installation and use of Linux based management and troubleshooting tools. In this short post I want to show you two use cases where this capability comes in extremely handy in the daily network management work.
While working with Amazon AWS S3 it turned out, that it allows users to specify S3 bucket names in the “US Standard” regions that are not allowed like this in any other zone. As most libraries building on top of S3 assume the naming restrictions for all non-“US Standard” regions are also enforced in “US Standard”, this breaks quite a bit of things.
EOS is released as a single image that supports all of their hardware platforms. But that same single image can also be run in a virtual machine! While a great article on “Building a Virtual Lab with Arista vEOS and VirtualBox” already exists, I wanted to accomplish the same with vSphere 5.x.
The topic of measuring network throughput between network devices comes up quite frequently: It ranges from users claiming (and often blaming) that the 100 Mbps Internet uplink in reality is only 10 Mbps or being surprised why they can’t transfer that multi-gigabyte file via FTP faster between data center locations.
Let’s have a look behind the scenes of network throughput measurement and understand why users are actually measuring something completely different, but also how to get more “performance” out of these connections.