Geo-distributed computing

ullbeking

n00b
Joined
Aug 17, 2017
Messages
32
Hi, I'm interested in geo-distributed computing for the purposes of resilience, redundancy, and HA. I know that this is very difficult problem to solve due to latency, but having servers that are sync'ed up across the world is actually the point of my project, and to see how feasible and difficult it is.

Each geographical "node" will be implemented as a cluster with redundancy, etc, but I really need these concepts to stretch across the world. I understand that data updates cannot be instantaneous, but as long as the far-away nodes are in a consistent state, i.e., they know whether or not they're up to do date, then that's fine.

Can anybody please point me to some essential resources, e.g., on the web, blogs, books, seminars or videos, etc. Thanks!!
 

Thanks for the link, ChristianVirtual. Yes, this is the sort of thing that I am starting out with, i.e., an article like this might describe the configuration of the machines at one data centre or at one site. I also like this article personally, as it's easy to read and I've not looked into MPI in much detail before.

What I'm really getting at is extending the connectivity of the LANs to other geographical sites, yet still maintaining consistency in the data. The point is that your compute and storage can continue even if one geographical site becomes unavailable, breaks down, or gets too slow for some reason. So you might have a very similar or identical setup in another location and some clever synchronization and replication algorithms that are able to mitigate the effects of the inevitably high latency between the two sites. This is known to be a difficult problem. But your article is a very good starting point, a building block for more complex structures that may emerge over time.
 
Back
Top