How frequent are network partitions on cloud services?
- by roja
Much is made of the CAP trade-off for data storage where conflicts can be introduced if there is a network partition.
My question is there any evidence that this is a problem that arises with any significant frequency in modern cloud IAAS services e.g.; EC2, Azure, Rackspace.
Is it a problem which, despite being a theoretical roadblock in constructing idealised distributed systems is, in fact, a non-issue for all practical concerns?
Has anyone experienced a network partition within one of these systems (within a single data-centre?) If so would you be willing to share any details?