Deuce STM Painless Java concurrency

1
Deuce STM Painless Java concurrency

• Guy Korland
• TAU GigaSpaces

2
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

3
Motivation
4
Moores Law
Transistor count still rising
Clock speed flattening sharply
(hat tip Simon Peyton-Jones)
5
The Problem

• Cannot exploit cheap threads
• Todays Software
• Non-scalable methodologies
• Todays Hardware
• Poor support for scalable synchronization

6
Why Locking Doesnt Scale?

• Not Robust
• Relies on conventions
• Hard to Use
• Conservative
• Deadlocks
• Lost wake-ups
• Not Composable

7
The Brief History of STM
8
TM Overview

• synchronized
• ltsequence of instructionsgt
• atomic
• ltsequence of instructionsgt

9
What is a transaction?

• Atomicity all or nothing
• Consistency consistent state (after before)
• Isolation Other cant see intermediate.
• Durability - persistent

10
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

11
Deuce

• Java STM framework
• _at_Atomic methods
• Field based access
• More scalable than Object bases.
• More efficient than word based.
• Support external libraries
• Can be part of a transaction
• No reserved words
• No need for new compilers (Existing IDEs can be
  used)
• Research tool
• API for developing and testing new algorithms.

12
Deuce - API

• public class Bank
• final private static double MAXIMUM_TRANSACTION
  1000
• private double commission 0
• _at_Atomic(retries64)
• public void transaction( Account ac1, Account
  ac2, double amount)
• if( amount lt 0 amount gt MAXIMUM_TRANSACTION
  )
• throw new IllegalArgumentException(Illega
  l transaction)
• ac1.balance - (amount commission)
• ac2.balance amount
• _at_Atomic
• public void update( double value)
• commission value

13
Deuce - Overview
14
Deuce - Running

• javaagentdeuceAgent.jar
• Dynamic bytecode manipulation.
• -Xbootclasspath/prt.jar
• Offline instrumentation to support boot
  classloader.
• -Dorg.deuce.excludeorg.junit.,org.eclipse.
• Exclude class from being instrumented.
• _at_Exclude
• java javaagentdeuceAgent.jar cp
  ex.jarmyjar.jar MyMain

15
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

16
Implementation

• ASM Bytecode manipulation
• Online Offline
• Fields
• private double commission
• final static public long commission__ADDRESS...
• Relative address (-1 if final).
• final static public Object __CLASS_BASE__ ...
• Mark the class base for static fields access.

17
Implementation

• Method
• _at_Atomic methods.
• Replace the with a transaction retry loop.
• Add another instrumented method.
• Non-Atomic methods
• Duplicate each with an instrumented version.

18
Implementation
_at_Atomic public void update ( double value)
commission value
_at_Atomic public void update ( double value)
double tmp commission commission tmp
value
19
Implementation
public void update( double value, Context c)
double tmp if( commission__ADDRESS lt 0 ) //
final field tmp commission else
c.beforeRead( this, commission__ADDRESS)
tmp c.onRead( this, commission,
commission__ADDRESS) if(
commission__ADDRESS lt 0 ) // final field
commission tmp value else
c.onWrite( this, tmp value, commission__ADDRESS)

20
Implementation
public void update( double value) Context
context ContextDelegetor.getContext() for(
int i retries i gt 0 --i)
context.init() try update(
value, context) if(
context.commit()) return
catch ( TransactionException e )
context.rollback() continue
catch ( Throwable t ) if(
context.commit()) throw t
throw new TransactionException()
21
Implementation
public interface Context void init ( int
atomicBlockId) boolean commit() void
rollback () void beforeReadAccess( Object
obj , long field ) Object onReadAccess(
Object obj, Object value , long field ) int
onReadAccess( Object obj, int value , long field
) long onReadAccess( Object obj, long value ,
long field ) void onWriteAccess( Object
obj , Object value , long field ) void
onWriteAccess( Object obj , int value , long
field ) void onWriteAccess( Object obj , long
value , long field )
22
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

23
TL2 (Transaction Locking II)Dave Dice, Ori
Shalev and Nir Shavit DISC 2006

• CTL - Commit-time locking
• Start
• Sample global version-clock
• Run through a speculative execution
• Collect write-set read-set
• End
• Lock the write-set
• Increment global version-clock
• Validate the read-set
• Commit and release the locks

24
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

25
LSA (Lazy Snapshot Algorithm)Torvald Riegel,
Pascal Felber and Christof Fetzer DISC 2006

• ETL - Encounter-time locking
• Start
• Sample global version-clock
• Run through a speculative execution
• Lock on write access
• Collect read-set write-set
• End
• Increment global version-clock
• Validate the read-set
• Commit and release the locks

26
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

27
Benchmarks (Azul Vega2 2 x 46)
28
Benchmarks (SuperMicro 2 x Quad Intel)
29
Benchmarks (Sun UltraSPARC T2 Plus 2 x Quad
x 8HT)
30
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

31
Summary

• Simple API
• _at_Atomic
• No change to Java
• No reserved word
• OpenSource
• On Google code
• Shows nice scalabilty
• Field based

32
Outline

• Motivation
• Deuce
• Implementation
• TL2
• LSA
• Benchmarks
• Summary
• References

33
References

• Homepage - http//sites.google.com/site/deucestm/
• Project - http//code.google.com/p/deuce/
• Wikipedia -http//en.wikipedia.org/wiki/Software_t
  ransactional_memory
• TL2 http//research.sun.com/scalable/
• LSA-STM - http//tmware.org/lsastm

34
Appendix

• Can you just replace synchronized blocks with
  _at_Atomic blocks?
• Privatization
• wait() (Message passing)