Understanding production usage conditions
Consider the flow of data and the actions of clients in your Diffusion™ server. Pre-production testing that closely models the usage you expect to see in production is most useful in understanding how your solution will respond in production.
The following sections contain some of the questions to consider when deciding how to test your solution before going to production. For each of these questions consider both average use values and edge case values.
Client connections
- How many clients do you expect to attempt to connect simultaneously?
- How many clients do you expect to be connected concurrently?
- Do you have session replication enabled and, if so, in a failover situation do you expect all of your concurrently connected clients to attempt reconnect at the same time?
- How long is a client connection expected to last?
- What is the expected geographic distribution of client connections?
- How does your load balancer decide how to distribute incoming client connections?
- How are your expected client connections distributed by platform or API?
- How are your incoming client connections authenticated?
Topics
- At what frequency do you expect to create topics?
- How many topics do you expect to create at the same time?
- At what frequency do you expect to delete topics?
- How many topics do you expect to delete at the same time?
- How many topics do you expect your clients to be subscribed to?
- How many topics do you expect your clients to subscribe to in a single action?
Topic updating
- How many topics is a given client expected to update?
- How frequently do you expect topics to be updated?
- How many topics do you expect to send updates to at the same time?
- How many topics do you expect a given client to send updates to at the same time?
- How big do you expect the data in your topic updates to be?
Other client actions
- How many client authentication requests is a given client expected to handle?
- How many messages sent to a message path is a given client expected to handle?
- How many messages sent directly to the client is a given client expected to receive?
- How many messages is a given client expected to send to a message path?
- How many messages is a given client expected to send directly to another client?
- How often do you expect clients to manage other clients?
- How many clients do you expect a given client to manage?