Please lower max_decay constant in source to readjust quickly to network activity
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
hardware-monitor (Ubuntu) |
Expired
|
Wishlist
|
Unassigned |
Bug Description
Binary package hint: hardware-monitor
Thanks to the new releases of this package, I have been able to start using the beautiful network monitor in this package.
However, one issue bugs me tremendously - once I've had a peak in network throughput, the monitor takes way too long time to readjust to a lower activity, resulting in a flat line at the bottom, in spite of reporting 100Kb/s...
Having looked at the source, it seems there is a max_decay constant that defines the behavior, scaling down old maximums gradually. This constant needs to be reduced significantly. At the current setting of 0.999, it takes 1000 updates before it reaches 37%. 1000 updates = 1000 seconds = 16.7 minutes. Thus, If I saw a peak of 1Mb/s at one point, 15 minutes later h-m will still consider 370Kb/s as the maximum and scale the axes accordingly, making the monitor pretty unusable for the most part of those 15 mins.
A setting of below 0.99 is more reasonable, meaning that 60 seconds will take you down to 54%. Thus, in the above example, 3 minutes would give you a max of 160Kb/s, which starts to be reasonable. Using 0.98 would give a max of 26Kb/s after three minutes, while 0.97 would give 4Kb/s after 3 mins, so I'd say a value around 0.97-0.99 is reasonable.
Thanks, Mikael
Changed in hardware-monitor: | |
importance: | Undecided → Wishlist |
Actually, trying with a value of 0.975 does not look very good. Maybe a better algorithm is called for?