Application Heartbeats Henry Hoffmann, Jonathan Eastep, Marco - - PowerPoint PPT Presentation

Application Heartbeats

Henry Hoffmann, Jonathan Eastep, Marco Santambrogio, Jason Miller, Anant Agarwal

CSAIL Massachusetts Institute of Technology Cambridge, MA 02139 http://groups.csail.mit.edu/carbon/heartbeats

2

Outline

• Introduction/Motivation

– Problem: Monitoring applications in self-tuning systems – Solution: Standard interface expresses performance/goals

• Application Heartbeats
• Experiments
• Conclusion

3

As System Complexity Increases, Self-Tuning Systems Emerge

• System Complexity is Skyrocketing

– Multicore processors – Parallel communication libraries – Heterogeneous architectures – Distributed, deep memory hierarchies – Special-purpose functional units – Unreliable components – New constraints: power, energy, wire delay

• Application programmers must be

experts in systems and apps Possible Solution: Self-Tuning Systems

Systems observe their runtime behavior, learn, and take actions to meet desired goals

4

Self-tuning Systems Must Monitor the Applications They Support

Disk I/O Devices DRAM

App 2 App 1 App 3

miss rate voltage, freq, precision cache size, associativity power IPC, power, temp

App 1

Core

Cache

App 2 App 3

Core

Cache speed

Application Layer Self-Tuning Services Layer

We propose Application Heartbeats as a standard API for applications to specify their goals and performance to self-tuning system services

Scheduler, Memory manager, file system

Operating System

Currently, applications run as performance black-boxes:

5

Outline

• Introduction/Motivation
• Application Heartbeats

– Idea – Interface

• Experiments
• Conclusion

6

The Application Heartbeats Idea

• At key intervals, apps issue a heartbeat using a simple function call
• Apps also register desired performance with other function calls
• The performance (heart rate) can be read within the application (a) or by

another process (b)

• If performance is low the system adapts to increase performance

7

Application Heartbeats Provide Standard API for Expressing Performance & Goals

• Application Heartbeats express goals and current performance
• System software can use Heartbeats to directly measure performance

Application Heartbeats

Disk I/O Devices DRAM

App 2 App 1 App 3

miss rate voltage, freq, precision cache size, associativity power activity, power, temp

App 1

Core

Cache

App 2 App 3

Core

Cache speed heartbeat, goals

Heartbeat

App 1 Min heart rate = 10 Max heart rate = 100 Current heart rate = 75 App 2 Min heart rate = 29.5 Max heart rate = 30 Current heart rate = 29.8 App 3 Min heart rate = 0.5 Max heart rate = 1.5 Current heart rate = .2

Scheduler, Memory manager, file system

Operating System

Apps no longer performance black-boxes

8

Heartbeat API Functions

Function Parameters Description

heartbeat_initialize [int] window_size

Initialize the heartbeat object to collect heartbeats. Uses a sliding window of window_size to calculate current hear trate

heartbeat [int] tag

Records a heartbeat with a given tag

hb_get_current_rate

Returns the current heart rate averaged over the last window_size heartbeats

hb_set_target_rate [float] min, [float] max

Sets the desired min and max heart rates for this app

hb_get_target_min_rate

Returns the minimum desired heart rate

hb_get_target_max_rate

Returns the maximum desired heart rate

hb_set_target_latency [float] min, [float] max, [int] tag1, [int] tag2

Sets the desired latency between heartbeats with tags tag1 and tag2

hb_get_min_latency [int] tag1, [int] tag2

Returns the minimum desired latency between two tags

hb_get_max_latency [int] tag1, [int] tag2

Returns the maximum desired latency between two tags

hb_get_history [int] n

Returns all heartbeat information for the last n heartbeats

Heartbeat API allows direct communication of performance and goals

9

Heartbeats Reference Implementations

http://groups.csail.mit.edu/carbon/heartbeats

• Files for distributed computing
• Performance1

– Throughput: ~0.900 Kbeat/s – Latency: ~1000 µs

• Shared Memory for multicore
• Performance2

– Throughput: ~1500 Kbeat/s – Latency: ~1.5 µs

• 1. Intel Xeon servers @3.16 GHz with :Linux NFS
• 2. Intel Xeon servers @ 3.16 GHz with Linux and POSIX shared memory

Callable from C/C++

10

Outline

• Introduction/Motivation
• Application Heartbeats
• Experiments

– Heartbeat use within an application – Heartbeat use by an external system – Other systems using Heartbeats

• Conclusion

11

Experiment 1: Internal Heartbeat Usage

• Experiment 1: Adaptive H.264 Encoder
• Goal: produce the highest quality video in real-time
• Method:

– A heartbeat is registered for each frame (frame rate = heart rate) – Encoder reads heartbeat and changes algorithm to reach target

• Results:

– Now the encoder is fast and still high quality – Achieve target performance with barely visible quality loss

12

Example 1: Performance

5 10 15 20 25 30 35 100 200 300 400 500 600

Time (Frame Number) Heart Rate (Frames/s)

Adaptive Encoder Target Heart Rate esa search umh search dia search eliminated I4x4 mode in I-frames eliminated I4x4 mode in P-frames eliminated P8x8 mode in P-frames eliminated sub-16x16 modes in P-frames eliminated rate-distortion optimizations reduced sub-pixel search reduced sub-pixel search

13

Example 1: Image Quality

• 1.2
• 1
• 0.8
• 0.6
• 0.4
• 0.2

0.2 0.4 0.6 0.8 100 200 300 400 500 600

Frame number PSNR Difference (dB)

14

Example 2: External Heartbeat Usage

• Experiment 2: External System Reads Heart Rate and Assigns

Cores

• Goal: Assign cores to keep performance within target range
• Method: Use PARSEC benchmarks

– Target heart rates set to be achievable using less than full number of cores

• Results:

– The scheduler keeps the applications running at the target speed – Scheduler can adapt to changes in the difficulty of the inputs

15

Example 2: bodytrack

0.5 1 1.5 2 2.5 3 3.5 4 4.5 50 100 150 200 250 Time (Heartbeat) Heart Rate (beat/s) 1 2 3 4 5 6 7 8 9 Cores

Heartrate Target Min Target Max Cores

16

Example 2: streammcluster

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 20 40 60 80 Time (Heartbeat) Heart Rate (beat/s) 1 2 3 4 5 6 7 8 Cores

Heartrate Target Min Target Max Cores

17

Example 2: x264

5 10 15 20 25 30 35 40 45 50 200 400 600 Time (Heartbeat) Heart Rate (beat/s) 1 2 3 4 5 6 7 8 9 10 Cores

Heart Rate Target Min Target Max Cores

18

Other Heartbeat Uses

• SpeedPress compiler and SpeedGuard runtime system

– The SpeedPress compiler discovers possible quality-of-service/ performance tradeoffs

• Achieve up to 2x speedup for 5% QoS loss

– The SpeedGuard runtime makes these tradeoffs dynamically in response to maintain a given heart rate in the face of environmental changes

• SmartLocks

– Subject of an upcoming SMART talk

More detail available in: Hoffmann, Misailovic, Sidiroglou, Agarwal, Rinard. Using Code Perforation to Improve Performance, Reduce Energy Consumption, and Respond to Failures. MIT-CSAIL-TR-2209-042. August, 2009.

19

Outline

• Introduction/Motivation
• Application Heartbeats
• Experiments
• Conclusion

– Request for feedback/usage – Summary

20

Request for Feedback

• Thanks to the reviewers for their feedback, but we need more…
• Heartbeat code is available online

http://groups.csail.mit.edu/carbon/heartbeats

• We need your feedback!

– If you have an self-tuning system service that could benefit from being able to directly measure an application’s performance try the interface – Let us know what you think

21

Summary

• Presented the Application Heartbeat interface

– API provides a standard means for an application to make its performance and goals known

• Presented several experiments showing basic usage

– Several other systems at MIT are using Heartbeats in more advanced applications

• Requested feedback from the community

Adaptive Scheduling Algorithm

• Take average heart rate over last 20 beats
• If heartbeat < target min

– Add a core – Wait for 20 beats and reapeat

• Else if heartbeat > target max

– Remove a core – Wait for 20 beats and repeat

• Else

– Repeat

22