IBM Global Mailbox - Message replication configuration trade-offs
Immediate replication and delayed replication provide different levels of availability, durability, and performance functions. Based on your business requirements, you must carefully evaluate the required functions and accordingly configure payload replication.
Consider the following items when configuring payload replication:
- Immediate (also known as synchronous) replication increases the response time for message uploads compared to delayed (or asynchronous) replication. This might cause some upload requests to time out, especially for large payloads.
- Delayed replication provides similar response time to traditional mailboxing.
- Messages cannot be uploaded in immediate replication if at least one remote data center is not online. For a two data center deployment of Global Mailbox, immediate payload replication is unavailable if either data center fails.
- If delayed replication is configured, it is possible for a payload to be lost. This might happen if the shared storage within a data center permanently fails before the payload is replicated to a remote data center.
- If shared storage fails within a particular data center, some Sterling B2B Integrator operations might fail to run within that data center.
- Due to the preceding two points, if the shared storage implementation is itself not fault tolerant, it might be a single point of failure. Refer the file system vendor documentation for information about fault tolerance of shared storage and how to configure it.
- Throughput does not differ markedly between immediate and delayed replication, assuming a constant-rate of payload uploads to Global Mailbox.
Important: After the payload of a message is uploaded, the payload cannot be changed. Therefore, all message payloads are strongly consistent.