Hello, I'm planning to test TIPC scalability in more detail, especially regarding the binding table and topology (service tracking) service with large number of nodes (~1k) and sockets (100-1k). Is it enough to do it with LXC containers with isolated namespaces to get realistic results? My concern here is that if TIPC implementation would not provide full isolation to sockets e.g. for performance reasons inside the same kernel, then this may not provide results matching to real environments. VMs are the other choice, but they tend to be heavy. Are the namespaces fully isolated inside the kernel?
Thanks for the info, -- Juhamatti _______________________________________________ tipc-discussion mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/tipc-discussion
