You can provision as much or as little storage and compute capacity as you need, with limited impact to your data warehouse operations when you implement those changes. Changing the size of your Db2® Warehouse cluster is a straightforward and fast task that involves just a few commands.
Your MPP cluster must have one head node and at least two data nodes. The maximum number of nodes
that you can scale out to is determined as follows:
If you deployed Db2 Warehouse 1.11.0 or later
and had more than 7.68 TB of cluster total RAM, 60 data partitions were allocated. The maximum
number of nodes that you can scale out to is therefore 60 (one head node and 59 data nodes).
If you deployed Db2 Warehouse 1.11.0 or later
and had 7.68 TB or less of cluster total RAM, 24 data partitions were allocated. The maximum number
of nodes that you can scale out to is therefore 24 (one head node and 23 data nodes). Even if you
increase the cluster total RAM after deployment, you cannot scale out to more than 24 nodes.
If you deployed Db2 Warehouse 1.7.0 - 1.9.0
and had at least 960 GB of cluster total RAM when you deployed, 60 data partitions were allocated.
The maximum number of nodes that you can scale out to is therefore 60 (one head node and 59 data
nodes).
If you deployed Db2 Warehouse 1.7.0 - 1.9.0
and had less than 960 GB of cluster total RAM when you deployed, 24 data partitions were allocated.
The maximum number of nodes that you can scale out to is therefore 24 (one head node and 23 data
nodes). Even if you increase the cluster total RAM after deployment, you cannot scale out to more
than 24 nodes.
As Figure 1 illustrates, after a scale out or scale in,
the data partitions, along with their workloads, are automatically rebalanced across the nodes in
the current setup. If you have a warehouse with 24 data partitions, the most balanced setup has 3,
4, 6, 8, 12, or 24 nodes. If you have a warehouse with 60 data partitions, the most balanced setup
has 3, 4, 5, 6, 10, 12, 15, 20, 30, or 60 nodes.Figure 1. Scaling a Db2 Warehouse deployment
To scale, you add storage capacity to the clustered file system or remove storage capacity from
that clustered file system. When you restart, any memory changes are detected, and your data
warehouse is automatically tuned accordingly.