If a key has is described multiple times it can be explained as multiple ways.
| Key | Description |
|---|---|
| HB_NUM | The heartbeat used when this poet apply was activated. |
| HB_RATE | The current Window Heartbeat Rate. Can check the Heartbeat with Tag equal to HB_NUM. |
| X_HAT_MINUS | The previous X_HAT. |
| X_HAT | NONE |
| P_MINUS | Previous P plus 0.00001. |
| H | The Speedup from the previous poet apply. |
| K | Seems to be Kalman specific. |
| P | NONE |
| SPEEDUP | The new Speedup to apply. |
| ERROR | Difference between wanted HB_RATE (MAX window size - MIN window size) and gotten HB_RATE |
| WORKLOAD | Estimate of the performance workload |
| WORKLOAD | Estimate of time between heartbeats given minimum amount of resources |
| LOWER_ID | Poet sets to configurations each apply. This is the Id of config with lowest speedup changed to this round. |
| UPPER_ID | Id of config with highest speedup changed to after this round. |
| NUM_HBS | The number of heartbeats spent in lower configuration. 0 if lower_id == upper_id. |
NOTE: Setting the pole as described in the poet paper can be done in the poet/src/poet_constands.h file, by setting FAST, SLOW or neither. If neither is set, it uses a pole of zero (0) as default. Changing P1 changes the pole as described in the poet paper. The other variables are for using poet in a more advanced way.
NOTE: A small pole means that the poet is more reactive. Zero (0) is unmitigated, while One (1) is rigid.
NOTE: When using only one pole, as is described in the poet paper, in the poet implementation provided, speedup can be calculated as: SPEEDUP = H + 'pole' * WORKLOAD * ERROR
NOTE: See more in paper:
- https://ieeexplore.ieee.org/document/7108419/.
- Docs/poet.pdf
For More information check your documentation folder
| Key | Description |
|---|---|
| Beat | Incrementing counter for each time heartbeat has been called |
| Tag | Index for loggin of heartbeat. Increases for each heartbeat log |
| Timestamp | Time when current heartbeat happened. On unix systems the time is given as nanosecond timestamps. See python time.time() |
| Global_Rate | Average rate of heartbeats per second from first timestamp |
| Global_Rate | Returns the heart rate over the life of the entire application |
| Window_Rate | Used by POET |
| Window_Rate | Average rate of heartbeats per second this window |
| Window_Rate | Returns the heart rate over the last window (as specified to init) |
| Instant_Rate | The rate of heartbeats per second this if one only takes a look at the current heartbeat and its predecessor taking only this heartbeat and its predecessor into consideration |
| Instant_Rate | Returns the heart rate for the last heartbeat. |
| Global_Accuracy | NONE |
| Window_Accuracy | NONE |
| Instant_Accuracy | NONE |
| Global_Power | Total power from the first heartbeat to the current heartbeat tag |
| Global_Power | The power (double) over the entire life of the application |
| Window_Power | Used by POET |
| Window_Power | The power (double) over the last window |
| Instant_Power | The power (double) for the last heartbeat |
_NOTE: Information regarding Power description has been collected from heartbeats/inc/heartbeat-accuracy-power.h and heartbeats/src/heartbeat-accuracy-power-shared.c
Documentation stored as refman.pdf.
Note: Field descriptions are usless/non-existent.
- Go into your
hearbeatsfolder. - Run
doxygen heartbeats_doc. - Go into latex documentation folder for heartbeat:
cd cd doc/latex. - Run make:
make. File is now stored asrefman.pdf.
Kalman filtering, also known as linear quadratic estimation (LQE), is an algorithm that uses a series of measurements observed over time, containing statistical noise and other inaccuracies, and produces estimates of unknown variables that tend to be more accurate than those based on a single measurement alone, by estimating a joint probability distribution over the variables for each timeframe. The filter is named after Rudolf E. Kálmán, one of the primary developers of its theory.
From https://en.wikipedia.org/wiki/Kalman_filter