Overcast - AWS CloudConnect

Many computer science experiments require increased control of the wide area network, include everything from guaranteed quality of service to low-level network programability using software defined networking. Chameleon provides these capabilities though direct low-level network connection between the research and public clouds, however creating them is challenging. While most public clouds provide low-level networking services (e.g. AWS Direct Connect, Azure ExpressRoute, or Google Dedicated Interconnect), using them is typically expensive; on the research cloud side, they can involve complicated campus network configuration arrangements that often limit access to this type of experimental configuration to a few a few select scientists or campus IT staff themselves.

This Jupyter notebook walks through the deployment of an experiment spanning Chameleon and AWS using CloudConnect. It deploys the network, compute servers, and a fully configured BGP router. Further, the BGP router can, optionally, be deployed on a dedicated OpenFlow networking switch or as software Quagga router existing on a standard x86 compute host. The full networking configuration is depicted in the figure below.

5 1 March 10, 2021, 10:17 a.m.


  • Paul Ruth

Digital Object Identifier (DOI)

Launch on Chameleon

Launching this artifact will open it within Chameleon’s shared Jupyter experiment environment, which is accessible to all Chameleon users with an active allocation.