Impact of Java Compressed Heap on Mobile/Wireless Communication Mayumi KATO and Chia-Tien Dan Lo (itcc’05) Department of Computer Science, University of Texas at San Antonio {mayumik, Speaker : Mayumi KATO
Outline 1.Introduction 2.Related work 3.Proposed architecture 4.Experiment and results 5.Conclusion and future work
Mobile Network Service Providers (network activated) Service Management Component Services Repository Service Archive service Web service Mobile commerce Audio, video animation Mobile/Wireless Communication Client and server models Introduction
Main Issue of Java mobile/wireless computing Introduction Many application demands more memory Mobile/wireless devices suffer from their small memory
Related work 6. Java heap memory compression [Lo and KATO ’03], [Chen et al. ’03], [KATO and Lo ’04] Introduction 1.Java classfile compression [Pugh’99] (small file, but the same info. as of a Jar file: eliminate redundancy) 2. Java bytecode factorization [Clausen et al.’00] (extended instruction set, macro instruction definitions from CAP file) -- bytecode instructions, replace common instruction sequences 3. Java compact bytecode instructions [Evans and Fraser’01] (grammar based method, G a parse tree derivation of the program) -- compression demands a minimum length derivation of the program 4. Java on-the-fly constant pool compaction [Rippert et al. ’04] (class loading, eliminate constant pool entry if not referenced) 5. Java profile-driven code unloading [Zhang and Krintz ’04] (JIT, unloading methods “has-not-been-used recently)
stored accessed Compressed heap Decompressing unit Compressing unit ALB table Cache unit Delayed buffer Java Virtual Machine (JVM) core Memory management module Store compressed block Delayed buffer is full? (compressed form) Local object accessed The proposed architecture consists of: Address Lookaside buffer
1.Reduce memory demands 2.Allow large client applications to run on mobile/wireless embedded devices 3.Minimize the number of active memory banks, and power off unused banks to eliminate the leakage current in memory system The Proposed Architecture Goals
The hardware de/compression engines are integrated into Java virtual machine (software) to de/compress a group (a page) of local and remote objects during Java execution. a group (a page) execution during Javaobjects Different from constant pool [Rippen et al.] native code [zhang and Krintz] Different from classfile [Pugh], bytecode [Clausen et al.], [Evans and Fraser] Different from per-object [Chen et al. 03] Features local and remote
Assumptions 1. Object is created either locally or remotely 2. Objects that come over the Internet have been compressed at the sending side 3. Objects that newly created inside the JVM are not compressed. The Proposed Architecture
stored Compressed heap Decompressing unit Compressing unit ALB table Cache unit Delayed buffer New local object created (uncompressed form) Java VM core Memory management module Store compressed block Delayed buffer is full?
stored Compressed heap Decompressing unit Compressing unit ALB table Cache unit Delayed buffer Remote object created and accessed (compressed form) Java VM core Memory management module Delayed buffer is full? Store compressed block Address Lookaside buffer
stored accessed Compressed heap Decompressing unit Compressing unit ALB table Cache unit Delayed buffer Java VM core Memory management module Delayed buffer is full? Store compressed block Compressed form Address Lookaside buffer
Garbage collection Java memory management system –Garbage collection mechanism Mark, sweep, compaction phases We redesigned it to handle compressed objects –Mark, similar to the original version –Sweep and compaction phases Migrated into de/compression modules Delayed until de/compression is invoked The Proposed Architecture
Garbage collection mechanism The Proposed Architecture From the caching unit
In-memory compression algorithms Popular compression algorithm LZ family –Designed for human text –Not suitable for data in memory/cache because of its regularity modeling Most in-memory/cache data –Word aligned integers and pointers –Contains many repeating zero values We use Wilson-Kaplan (WK) compression family –A dictionary-based algorithm The Proposed Architecture
WK algorithms Coding format [4 bits] [10 bits] [22 bits] Dictionary indexlow upper Match type Coding specification ZERO EXACT PARTIAL MISS high low
WK Example A0129FAE A0129CAE A0129DAE A01290AE A0129FAE no match partial exact Input Dictionary output
Experiment and Results 1.Examined compression techniques on mobile/wireless devices (CS LAN) 2.Show their impact using space and time efficiencies
W gc : watermark on the original architecture (gc) W comp+gc : watermark on the proposed architecture (compression + gc) T gc : total execution time (including gc time) on the original architecture T comp+gc : total execution time (including comp. and gc times) on the proposed architecture W gc spaceEfficiency = W comp+gc T gc timeEfficiency = T comp+gc
Summaries of Experiment Results Application Space Efficiency Viewer 256KB heap 2.05 HTTP demo 64KB heap 1.80 Stock 64KB heap 2.50 Audiodemo 64KB heap 2.20 Manyballs 32KB heap 2.09 Space efficiency 2.0 –Reduce heap memory demand to 50% or more on average –Independent of the size of Java dynamic heap –Half of the memory banks for Java heap may never be turned on –More than 50% of the memory leakage can be saved Experiment and results
Time efficiency 1.0 –HTTP demo, Audio demo, many balls No time overhead Good data and code locality and less invocation of garbage collection Time efficiency 0.99 –Stock and Viewer Time overhead is within 1 % The use of local database and disk accesses? Application time Efficiency Viewer 256KB heap 0.99 HTTP demo 64KB heap 1.00 Stock 64KB heap 0.99 Audiodemo 64KB heap 1.00 Manyballs 32KB heap 1.00
Conclusion and Future Work We have seen the impact of Java compressed heap. Results show The compressed heap –Effective –Ensure small memory footprints for mobile/wireless application with any memory demand. Experiment and results
On-going work Tuning speed Future work Studying the impact of the compressed heap on remote object
Questions URL paper : ieee library 6 pages, but with Dr. Lo’s permission