Clarification of timestamps in MS_CallGraph_Table (2021 trace dataset)

[yazhuo]

Thanks for offering such a great trace dataset.

I'm wondering the timestamp in MS_CallGraph_Table is collected from the serverside or clientside. For example, is the timestamp recorded from a machine(host) running um or a machine(host) running dm?

columns	Example Entry
timestamp	16397576
traceid	015101cd15919399974329000e
rpcid	0.1.1.2.50
um	35114acfb54c54fb9618f23cd28bbc57c765f597df140977d7030dcc52775ed4
rpctype	rpc
interface	af42b5e3e0eb334d38619733586d78d1414f6549f24d31b39a5294454638bc59
dm	b65fdc9bfef6b4974c3e90e1ec7b92d30e639789da5a78c1d4685857e19c75a0
rt	13

[niewuya]

Thanks for your interest. The timestamp is retrieved from the log collected in the machine. The timestamp is collected from the serverside (clientside) if the log is recorded from the machine running dm (um). You can use the metric rt to judge whether the timestamp is collected from the serverside or clientside. If the rt > 0, it is on the clientside. By contrast, if the rt < 0, it is on the serverside. Please let me know if you have any other questions. Thanks.

[Sachin-A]

Thanks for the response @niewuya!

A few follow up questions to the ones @yazhuo asked:

1. What does it mean for the dm to log the rt? Is it logging the difference between when it received the request and when it sent out the response to the um?
2. How do we determine whether it was logged at um or dm if rt is equal to 0?
3. There are many entries where the um and dm do not agree on the rt. For example, um logs the rt as 48 whereas the dm logs it as -47. Is this because of the network latency between um and dm?