问题
We are building a Cassandra (2.1.5) cluster for storing large amount of timeseries data, and we are planning to utilize existing hardware, problem is the hardware available is really different.
2 machines with: 4 core, 8 GB, SSD
2 machines with: 8 core, 16 GB, SSD
2 machines with: 32 core, 64 GB, HDD
Obviously, the 32 core machine can handle much larger load than the 4 core machines, how should we configure Cassandra to handle this. We are using RF 3 and the latest datastax java driver.
Any suggestions?
回答1:
When you configure a Cassandra node you tell it the number of vnodes to use for that node. The load on a node is roughly proprtional to its number of vnodes. Hence by choosing different numbers of vnodes for your computers you can distribute the load unevenly among them.
来源：https://stackoverflow.com/questions/30648981/cassandra-with-uneven-hardware-how-to-configure