Chapter 4. Clustering and High Availability
Dan Kegel published his well-known problem, C10K in 1999. The problem simply arose from handling 10k simultaneous clients on the web servers. Currently, we have to handle more than 100k simultaneous clients on our web servers or on our software systems.
C10K is a great start to solve the scalability problem; however, we have a much bigger problem on our hands now. If we return to messaging systems and RabbitMQ, we have to handle lots of simultaneous messages; however, we don't have a chance to handle all simultaneous messages in a single RabbitMQ server.
Anyway, RabbitMQ has great skills to handle lots of messages in a single machine, such as more than 50k messages per second according to VMware Performance ...