Title: The Case For Prediction-based Best-effort Real-time
1The Case For Prediction-based Best-effort
Real-time
- Peter A. Dinda
- Bruce Lowekamp
- Loukas F. Kallivokas
- David R. OHallaron
- Carnegie Mellon University
2Overview
- Distributed interactive applications
- Could benefit from best-effort real-time
- Example QuakeViz (Earthquake Visualization) and
the DV (Distributed Visualization) framework - Evidence for feasibility of prediction-based
best-effort RT service for these applications - Mapping algorithms
- Execution time model
- Host load prediction
3Application Characteristics
- Interactivity
- Users initiate tasks with deadlines
- Timely, consistent, and predictable feedback
- Resilience
- Missed deadlines are acceptable
- Distributability
- Tasks can be initiated on any host
- Adaptability
- Task computation and communication can be
adjusted
Shared, unreserved computing environments
4Motivation for QuakeViz
Teora, Italy 1980
5Northridge Earthquake Simulation
Real Event
High Perf. Simulation
40 seconds of an aftershock of Jan 17, 1994
Northridge quake in San Fernando Valley of
Southern California
16,666 40M x 40M SMVPs 15 GBytes of RAM 6.5 hours
on 256 T3D PEs 80 trillion (1012) FLOPs 3.5
sustained GFLOP/s 1.4 peak GB/s
Huge Model
50 x 50 x 10 km region 13,422,563
nodes 76,778,630 tetrahedrons 1 Hz frequency
resolution 20 meter spatial resolution
HUGE OUTPUT
16,666 time steps 13,422,563 3-tuples per step 6
Terabytes
6Must Visualize Massive Remote Datasets
Datasets must be kept at remote supercomputing
site due to their sheer size
Visualization is inherently distributed
Problem
One Month Turnaround Time
7QuakeViz Distributed Interactive
Visualizationof Massive Remote Earthquake
Datasets
Sample 2 host visualization of Northridge
Earthquake
Goal
Interactive manipulation of massive remote
datasets from arbitrary clients
8DV A Framework For Building Distributed
Interactive Visualizations of Massive Remote
Datasets
- Logical View Distributed pipelines of vtk
modules
local display and user
User feedback and quality settings
resolution
contours
ROI
interpolation
isosurface extraction
Dataset
reading
scene synthesis
rendering
interpolation
morphology reconstruction
Display update latency
deadline
Visualization Toolkit, open source C library
9DV A Framework For Building Distributed
Interactive Visualizations of Massive Remote
Datasets
- Logical View Distributed pipelines of vtk
modules
local display and user
User feedback and quality settings
resolution
contours
ROI
interpolation
isosurface extraction
Dataset
reading
scene synthesis
rendering
interpolation
morphology reconstruction
Display update latency
deadline
Visualization Toolkit, open source C library
10Active Frames
Physical View of Example Pipeline
interpolation
isosurface extraction
scene synthesis
deadline
deadline
deadline
Active Frame n2
?
Active Frame n1
?
Active Frame n
?
- Encapsulates data, computation, and path through
pipeline - Launched from server by user interaction
- Dynamically chose on which host each pipeline
stage will execute and what quality settings to
use
11Active Frames
Physical View of Example Pipeline
interpolation
isosurface extraction
scene synthesis
deadline
deadline
deadline
Active Frame n2
?
Active Frame n1
?
Active Frame n
?
- Encapsulates data, computation, and path through
pipeline - Launched from server by user interaction
- Dynamically chose on which host each pipeline
stage will execute and what quality settings to
use
12Active Frame Execution Model
deadline
Active Frame
- pipeline stage
- quality params
Resource Predictions
Mapping Algorithm
Exec Time Model
CMU Remos API
Prediction
Prediction
Host Load Measurement
Network Measurement
Remos Measurement Infrastructure
13Active Frame Execution Model
deadline
Active Frame
- pipeline stage
- quality params
Resource Predictions
Mapping Algorithm
Exec Time Model
CMU Remos API
Prediction
Prediction
Host Load Measurement
Network Measurement
Remos Measurement Infrastructure
14Active Frame Execution Model
deadline
Active Frame
- pipeline stage
- quality params
Resource Predictions
Mapping Algorithm
Exec Time Model
CMU Remos API
Prediction
Prediction
Host Load Measurement
Network Measurement
Remos Measurement Infrastructure
15Feasibility of Best-effort Mapping Algorithms
16Active Frame Execution Model
deadline
Active Frame
- pipeline stage
- quality params
Resource Predictions
Mapping Algorithm
Exec Time Model
CMU Remos API
Prediction
Prediction
Host Load Measurement
Network Measurement
Remos Measurement Infrastructure
17Feasibility of Execution Time Models
18Active Frame Execution Model
deadline
Active Frame
- pipeline stage
- quality params
Resource Predictions
Mapping Algorithm
Exec Time Model
CMU Remos API
Prediction
Prediction
Host Load Measurement
Network Measurement
Remos Measurement Infrastructure
19Why Is Prediction Important?
Bad Prediction No obvious choice
Good Prediction Two good choices
Predicted Exec Time
Predicted Exec Time
deadline
Good predictions result in smaller confidence
intervals Smaller confidence intervals simplify
mapping decision
20Feasibility of Host Load Prediction
21Comparing Prediction Models
Run 1000s of randomized testcases, measure
prediction error for each, datamine results
Inconsistent low error
Consistent high error
97.5
Mean Squared Error
75
Consistent low error
Mean
50
25
Model A
Model B
Model C
2.5
Good models achieve consistently low error
22Comparing Linear Models for Host Load Prediction
15 second predictions for one host
97.5
75
Mean
50
25
2.5
Raw
Very
Cheap
Expensive
23Conclusions
- Identified and described class of applications
that benefit from best-effort real-time - Distributed interactive applications
- Example QuakeViz / DV
- Showed feasibility of prediction-based
best-effort real-time systems - Mapping algorithms, execution time model, host
load prediction
24Status - http//www.cs.cmu.edu/cmcl
- QuakeViz / DV
- Overview PDPTA'99, Aeschlimann, et al
- http//www.cs.cmu.edu/quake
- Currently under construction
- Remos
- Overview HPDC98, DeWitt, et al
- Available from http//www.cs.cmu.edu/cmcl/remulac
/remos.html - Integrating prediction services
- Network measurement and analysis
- HPDC98, DeWitt, et al HPDC99, Lowekamp, et al
- Currently studying network prediction
- Host load measurement and analysis
- LCR98, Dinda SciProg99, Dinda
- Host load prediction
- HPDC99, Dinda, et al
25Feasibility of Best-effort Mapping Algorithms
26Feasibility of Host Load Prediction
27Comparing Linear Models for Host Load Prediction
15 second predictions aggregated over 38 hosts
97.5
75
Mean
50
25
2.5
Raw
Very
Cheap
Expensive