Will the partition adapt to the change of workload?
Jan 13, 2017
Hi, I have seen the discussion of https://forum.voltdb.com/showthread.php?301-How-does-partitioning-work and checked the source code of TheHashinator.java.
It seems that each new coming row is written to a partition that is determined by a given hash algorithm.
However, if some partition is overloaded, e.g., some some rows of this partition are read frequently in the last interval, will VoltDB migrate some hot data to the other partitions dynamically?
I have read the paper presented in VLDB 2016, "Clay: FineGrained Adaptive Partitioning for General Database Schemas", which propose an adaptive partitioning strategy. Will Clay be used in VoltDB?
Thanks.
It seems that each new coming row is written to a partition that is determined by a given hash algorithm.
However, if some partition is overloaded, e.g., some some rows of this partition are read frequently in the last interval, will VoltDB migrate some hot data to the other partitions dynamically?
I have read the paper presented in VLDB 2016, "Clay: FineGrained Adaptive Partitioning for General Database Schemas", which propose an adaptive partitioning strategy. Will Clay be used in VoltDB?
Thanks.