Hello,
I am planning to run a small network for scientific computation experiment.
Ideally I should have:
- A main internet connection (About 20Mb)
- A backup connection (could be wired or even an USB 4G connection).
- I would like to be able to have various sites/regions
- On each region I should have from 3 to 24 compute nodes, depending on the size of each region (48 ports switches should work fine).
- The Overall network consumption for each region is low (about 10Mbps) since the nodes do not interact with each other, except for regular network broadcasts.
- Each region should be available through a VPN or similar for both maintenance and remote ssh/web access to the nodes but the nodes are not required to connect with each other to work. Each node only requires a working internet connection (this is why backup internet is important).
- These regions could be dispersed across two countries.
- Ideally a dashboard/tool should show the state of the regions connectivity as the compute nodes.
- Later on I will try to implement out some Graphana dashboard for the application layer
My internet provider does not provide me with a fixed IP address so I either use dynamic DNS for the VPN or use some sort of mesh VPN. I could get a VPS for the sake of having a fixed IP address and have the regions connect to that VPS. Of course the more tolerant to failure, the better. The less bottlenecks or centralization, the better.
Questions:
- Which equipment would you recommend?
- Which configuration would you recommend?
A similar experiment used Cisco 3750 switches (48 ports) and Meraki mx64w routers.
Thank you very much.