Cluster node considerations for cloud services
This topic describes the cluster considerations that need to be followed before you install cloud services on the IBM Storage Scale cluster.
You can install the core cloud services packages on a maximum of 4 CES or NSD (or a combination of both) nodes that you want as members of your cloud services node class. We support up to 4 node classes. The client package is installed on all other nodes within the same cluster. Cloud data transfer (migration, recall, import, or export) occurs only from the cloud services nodes. However, you can initiate a cloud data transfer from either a server or a client node.
You can install the transparent cloud tiering server package on a non-gateway node. In this client-assisted recall configuration, the non-gateway node can recall from the same node instead of routing the commands through the gateway nodes. Except for recall, all other commands are routed through the gateway nodes. Thus, file-read requests originating on the client nodes are served immediately from the same node.
- The non-gateway node has direct network access to the cloud object storage.
- High CPU and memory resources are available on the node as the client-assist package starts a Java™ virtual machine per node-class managed
- If a large number of recalls are initiated through the non-gateway nodes.
- Do not support transparent data recalling
- Cannot respond to requests that initiate from nodes on which a client is not installed
- Cannot efficiently perform deletions