If I have one Cassandra cluster setup across 5 data centers (3 are private DCs) and 2 are Public (Azure DCs), can I say I have 5 rings or is this 1 cluster and 1 ring ?
Can someone help understanding the term "ring" in this context.
1 Answers
Long answer:
Yes, cluster and ring can be used interchangeably. "Cluster" is certainly used more today. "Ring" comes from the early, pre-vNodes days of Cassandra, where each node was assigned a single, contiguous token range.
We used to have to manually configure the token range for each node. In fact, I would use this code to do it (assuming a 5 node cluster):
python -c 'print [str(((2**64 / 5) * i) - 2**63) for i in range(5)]'
['-9223372036854775808', '-5534023222112865485', '-1844674407370955162', '1844674407370955161', '5534023222112865484']
When computing token ranges of partition keys, each node was responsible for the next, sequential range. Once it got to the last node in the cluster, the range calculation looped back around to the first node, sort of making a ring-like diagram.
In summary, now with multiple non-contiguous token ranges on each node, Cassandra really doesn't form a "ring" anymore. To see what I mean, run a nodetool ring on a Cassandra 3 node with multiple vNodes, and you'll see what I mean.
Short answer:
For all intents and purposes, "cluster" == "ring." You can say either, and people will know what you mean.
Related