You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In ailoads, we had the concept of a "user." From a testing perspective, this makes some sense because (via a loadtest), we're trying to pretend that we are 1 user doing some activity and multiplying that by some scale value: --users n. But it's not entirely intuitive what a "user" is and this could easily get confused with the --name parameter in loads-broker where we specify the username of the person running the tests.
In molotov, we've renamed that to "worker." This is better in one sense as it's more like a thread, but could easily be confused with the "process" parameter which is a thread of the async activity. A process is essentially a thread and a thread is sometimes also referred to as a "worker", so sometimes I find that a bit confusing.
We could rename "workers" to "connections", but in ap-loadtester, a connection is used to refer to an open websocket connection, whereas with molotov our "workers" are essentially agents and the connections may be established and terminated multiple times over the course of a test.
Should we call it an "agent" instead?
NOTE:
When doing a loads-broker test, these params would be added to the loads-broker.json so that your scale would be determined by:
molotov.json
(1) # processes
(2) # agents
loads-broker.json
(3) # attack nodes
TOTAL # AGENTS = # processes / node X # agents / node X # attack nodes
The text was updated successfully, but these errors were encountered:
rpappalax
changed the title
re-name workers?
re-name --workers to --agents??
Feb 23, 2017
I agree the naming is fuzzy. I like the name of agents - in another tool they call it virtual-users
To make sure things are clear, right now:
a process is a real OS process, where molotov is forked and a new program runs.
a worker is a co-routine, that asynchronously calls the scenarii
so when you run with 5 processes and 10 workers, you end up with 50 workers running in parallel.
--agents sounds good as long as it's clear that it's not the total number of concurrent virtual users, but the total number of virtual user PER process.
so when you run with 5 processes and 10 workers, you end up with 50 workers running in parallel.
yea, that's my understanding as well.
--virtual-users (-u) is also quite good. I like the brevity of 'agents', but as you suggest it may still be too vague. i quite like 'virtual-users'. Though a bit long, it does give a very intuitive name to what it does. Whatever you think is best.
@tarekziade, question for you....
In ailoads, we had the concept of a "user." From a testing perspective, this makes some sense because (via a loadtest), we're trying to pretend that we are 1 user doing some activity and multiplying that by some scale value: --users n. But it's not entirely intuitive what a "user" is and this could easily get confused with the --name parameter in loads-broker where we specify the username of the person running the tests.
In molotov, we've renamed that to "worker." This is better in one sense as it's more like a thread, but could easily be confused with the "process" parameter which is a thread of the async activity. A process is essentially a thread and a thread is sometimes also referred to as a "worker", so sometimes I find that a bit confusing.
The loadtesting tool: beeswithmachineguns refers to these "users"/"workers" as "concurrent connections":
https://github.com/igkins/beeswithmachineguns/blob/master/beeswithmachineguns/main.py#L100
We could rename "workers" to "connections", but in ap-loadtester, a connection is used to refer to an open websocket connection, whereas with molotov our "workers" are essentially agents and the connections may be established and terminated multiple times over the course of a test.
Should we call it an "agent" instead?
NOTE:
When doing a loads-broker test, these params would be added to the loads-broker.json so that your scale would be determined by:
molotov.json
(1) # processes
(2) # agents
loads-broker.json
(3) # attack nodes
TOTAL # AGENTS = # processes / node X # agents / node X # attack nodes
The text was updated successfully, but these errors were encountered: