Load Prediction for Best Effort Real Time

1
Load Prediction for Best Effort Real Time
Peter A. Dinda
Execution Model
Execution Time and Host Load
Integration In BBN QuO
Execution Model and Integration into BBN QuO
Choose host based on predicted load
Load is Self Similar
Load Exhibits Epochal Behavior
Load is Highly Variable
Statistical Properties of Host Load
Linear Time Series Models
Realizable Pole-zero Models
Real World Benefits of Prediction
sa is the confidence interval for t1
predictions Map work that would take 100 ms at
zero load axp0 sz0.54, m1.0, sa(ARMA(4,4))
0.109 sa(ARFIMA(4,d,4)) 0.108 no model 1.0
/- 1.06 (95) gt 100 to 306 ms ARMA 1.0 /-
0.22 (95) gt 178 to 222 ms ARFIMA 1.0 /-
0.21 (95) gt 179 to 221 ms axp7 sz0.14,
m0.12, sa(ARMA(4,4)) 0.041 sa(ARFIMA(4,d,4))
0.025 no model 0.12 /- 0.27 (95) gt 100 to
139 ms ARMA 0.12 /- 0.08 (95) gt 104 to 120
ms ARFIMA 0.12 /- 0.05 (95) gt 107 to 117 ms
Load Prediction With Linear Time Series Models
ARFIMA Models Capture Long-range Dependence of
Self-Similar Signals