-
Type:
Task
-
Status: Complete
-
Priority:
Medium
-
Resolution: Done
-
Affects Version/s: None
-
Fix Version/s: 1.6
-
Component/s: None
-
Labels:None
-
Epic Link:
-
Sprint:EV 18.15 Stability/Availabilit
-
Documentation Impact:Unset
In order to find out proper values for ZMQ queue sizes and a number of messages we process by each run of the looper (DEFAULT_LISTENER_QUOTA, DEFAULT_SENDER_QUOTA, ZMQ_CLIENT_QUEUE_SIZE, ZMQ_NODE_QUEUE_SIZE), we need to measure the following:
- How many messages we process in one run in node-to-node and client-to-node stacks
- under the load and not
- average, minimum and maximum values, distribution graph)
- How much time each run of the looper takes (from one call to network stack to another)
- Size of 3PC batches
- Size of messages
- How many node-to-node messages we have for 1 3PC Batch and 1 request
- How many transport Batches we create and their size (on both KB and number of messages)
For each value we need to measure:
- System not under load
- System with a steady-state of significant load (see
INDY-1343)
We need to know the average, minimum and maximum values, distribution graph.
This baseline testing should only follow our recommended production configuration.
This baseline testing does not include:
- Edge cases
- Forced view changes
- Combinations of the various configuration values
Acceptance criteria
- Hooks in the code to perform the measurement
- A spreadsheet with info above
- A summary document (about 1 page) of findings and recommendations for network node resourcing that can be provided to the Sovrin Foundation
- Identified improvements should be raised as separate issues
- relates to
-
INDY-1549 Change default configs for better performance and stability
-
- Complete
-