This section covers which MemDB instances support data striping, instructions for configuring basic and advanced features of MemDB data striping, how to display the current striping configuration, and striping-related statistics.
|
What is MemDB data striping?
MemDB data striping means dividing a MemDB instance into partially independent "stripes". This can remove bottlenecks in MemDB, letting Rhino better use more available cores (in machines with many CPU cores). In other words, the primary purpose of data striping is to increase vertical scalability. MemDB data striping was introduced in Rhino 2.3.1. |
|
Data striping should not be used for replicated MemDB instances. Under some conditions it can corrupt the management database. |
MemDB instances
Rhino includes two types of MemDB (Rhino’s in-memory database):
-
local MemDB — contains state local to the Rhino node, used by non-replicated applications running in "high-availability mode".
-
replicated MemDB — contains state replicated across the cluster, domain, or sub-cluster, used by replicated applications running in "fault-tolerant mode".
|
|
MemDB instances backed by disk storage — including the profile database and management database — do not support striping. |
