researcherben / structural-simulation-toolkit

Sandia SST
GNU General Public License v3.0
4 stars 0 forks source link

create an example of the thundering herd problem using SST components #37

Open bhpayne opened 2 years ago

bhpayne commented 2 years ago

Thundering herd arises in the context of cache misses. If many things try to access the same thing at the same time, a bottleneck can occur.

Part 1 of this task would be to implement a model using SST that demonstrates the thundering herd problem.

Part 2 of the task would be to show what measurement can be used to indicate a thundering herd is causing the problem.

Why this is useful: Even though this is a solvable problem, it can still occur if protocol architects and network designers are unfamiliar with the issue or due to unforeseen conditions. Having an SST model of this specific problem allows designers to become familiar with how the problem would manifest in their SST-based model.

thundering herd examples in modern software

https://news.ycombinator.com/item?id=1722126

https://ospassist.puppet.com/hc/en-us/articles/360039825574-Diagnose-stop-and-prevent-a-thundering-herd-for-Puppet- and https://ospassist.puppet.com/hc/en-us/articles/360039825994

https://dieswaytoofast.blogspot.com/2018/09/hacking-thundering-herd.html

https://blog.clubhouse.com/reining-in-the-thundering-herd-with-django-and-gunicorn/ and https://news.ycombinator.com/item?id=28190481

https://downtimeproject.com/podcast/how-coinbase-unleashed-a-thundering-herd/

bhpayne commented 2 years ago

I'm not seeking automatic detection in SST

I'm wondering should be measured (logs, statistics) offline

bhpayne commented 2 years ago

indicator to measure is "how many things are checking a cache per time window?"