SlideShare une entreprise Scribd logo
1  sur  23
Télécharger pour lire hors ligne
INTRODUCING A NON-VOLATILE GENERIC OBJECT
PROGRAMMING MODEL FOR IN-MEMORY COMPUTING
(SPARK) PERFORMANCE IMPROVEMENT
YANPING WANG
APACHE MNEMONIC (INCUBATING) PROJECT LEADER
YANPINGW@APACHE.ORG
STORAGE MEDIA
EVOLUTION IS
CHANGING FUTURE SW
PROGRAMING
STORAGE MEDIA EVOLUTION
 Storage Media
Evolution: NAND
and 3D XPoint are
filling the Speed
and Capacity
gap between
DRAM and Hard
Disk Drives
 Software is facing
the challenge on
how to use them
STORAGE MEDIA EVOLUTION IS CHANGING FUTURE
SW PROGRAMING
FROM IN-MEMORY TO IN-PLACE
 Moving from In-Memory
Computing to In-Place
Computing
 Mnemonic is a
pioneering project to
extend in-memory
computing to in-place
computing using next
generation NVM
storage media.
(Picture was taken from Ken
Gibson’s Storage Industry Summit
January, 20, 2016)
A NON-VOLATILE
PROGRAMMING MODEL
FOR JAVA APPLICATION
DEVELOPERS
WHY JAVA?
• There are 300+ open source initiatives at the ASF. 221 of them are
Java based, accounts for 59.9%.
• Top Level Big Data Projects are Java based projects, include
Hadoop, CloudStack, Spark, Cassandra, Hive, HBase, Flink, Solr…..
• Java related optimization has significantly contributed to Big Data
applications’ performance improvement.
TRADITIONAL JAVA BASED BIG DATA
FRAMEWORKS AND APPLICATIONS
OUR PROPOSAL: OFF-HEAP, LEAVE DATA IN-PLACE
NON-VOLATILE JAVA OBJECT MODEL
• Provides Java programmers a
mechanism to define objects and
structures as Mnemonic-managed
objects and structures.
• Objects identified as @DurableEntity
store their durable fields in volatile/non-
volatile memory space.
• @DurableGetter and @DurableSetter are
used to access those fields.
• Handlers are used to refer their
connected durable object.
• https://docs.oracle.com/javase/tutorial
/java/annotations/basics.html
AN EXAMPLE OF NON-VOLATILE LINKED LIST
(LAZY LOADING)
APACHE MNEMONIC INFRASTRUCTURE
APACHE MNEMONIC STRUCTURE
• Performance oriented architecture
• Unified platform enabling Framework
• Unique non-volatile object model
• Flexible & extensible focal point for
optimization
MNEMONIC SEARCH EXAMPLE
• Traditional search:
SerDes, caching,
pack/unpack, burst
temp objects
creation and
destruction cost can
be as high as 70-80%
• Mnemonic Search: in-
place creation &
update, no SerDes,
load on demand, no
need to normalize
MNEMONIC PROJECT POSITIONING
• Apache Mnemonic (Incubating) is a
Java based non-volatile memory library
(API) for in-place structured data
processing and computing
• Addresses Big Data performance issues
• High cost of Serialization/De-serialization
• Frequent unpredicted long Garbage Collection
pauses
• Burst temporary object creation/destruction
• Frequent high cost JNI call with stack
marshalling
• System paging and kernel structure updating
due to massive data processing
USE CASE EXAMPLE:
SPARK + MNEMONIC
=IN-PLACE MACHINE
LEARNING COMPUTING
NON-VOLATILE RDD CODE FOR SPARK
• ~600 line of Non-Volatile
RDD code is added into
Spark, to alter data flow
from Spark RDD via JVM
heap to Mnemonic.
• Spill RDDs to off-heap and
disk can be avoid.
• RDD data cache and
local checkpoints can also
be avoided when using
Nonvolatile or Durable
objects.
INTERNAL STRUCTURED DATA REPRESENTATION
 Spark experiment uses: 2D Linked List
data structure to store data in
persistent memory and storage.
 N nodes: represent the first level of
linked list.
 V nodes: represent the values of first
level of linked list that still is a linked
list.
 Only N2 and V1 (methods & volatile
fields) are instantiated on heap.
SPARK MLLIB KMEANS
SPARK MLLIB KMEANS PERFORMANCE IMPROVEMENT
• SPARK ML Kmeans implements one of the most
popular clustering algorithms
• Exp1: spark.storage.memoryFraction 0.6
• Exp2: spark.storage.memoryFraction 0.2
CONFIGURATION FOR PERFORMANCE MEASUREMENT
 Platform Software Configuration
 OS: CentOS 7
 JVM: JDK8 update 60
 ParallelOldGC is used for Spark ML throughput
workload Kmeans
 Spark1.5.0
 Platform Hardware
 Haswell EP (72 cores, E5-2699 v3 @ 2.3GHz)
 256GB DDR4-2133 on-board Memory
 Storage media: SATA SSD
 Device Model: INTEL SSDSC2BB016T4
 Serial Number: BTWD442601311P6HGN
 User Capacity: 1.60 TB
 SATA Version is: SATA 3.0, 6.0 Gb/s
 Spark Kmeans Experiments Setting:
– System memory is reduced to 128G via
static dummy ramfile occupancy
– 20 executors with 4 GB JVM heap, 3 cores
– Both of datasets are stored on SSD
– Experiment 1
– spark.storage.memoryFraction 0.6
– 100,000,000 records with vector in 12
dimension (20 partitions)
– Experiment 2
– spark.storage.memoryFraction 0.2
– 100,000,000 records with vector in 6
dimension (20 partitions)
DISCLAIMERS
Apache Mnemonic is an effort undergoing incubation at The Apache Software Foundation
(ASF), sponsored by Apache Incubator PMC. Incubation is required of all newly accepted
projects until a further review indicates that the infrastructure, communications, and
decision making process have stabilized in a manner consistent with other successful ASF
projects. While incubation status is not necessarily a reflection of the completeness or
stability of the code, it does indicate that the project has yet to be fully endorsed by the
ASF.
While Mnemonic is operating within the time period where the application is processing
their durable data, Mnemonic guarantees the persistency once Mnemonic is closed
gracefully.
PROJECT LINKS
Mnemonic Community: dev@mnemonic.incubator.apache.org
Source Code: https://github.com/apache/incubator-mnemonic
JIRA Tracking: https://issues.apache.org/jira/browse/MNEMONIC
Website: http://mnemonic.incubator.apache.org/
Contact:
yanpingw@apache.org; garyw@apache.org

Contenu connexe

Tendances

No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...
No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...
No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...Databricks
 
Ceph Day Shanghai - Recovery Erasure Coding and Cache Tiering
Ceph Day Shanghai - Recovery Erasure Coding and Cache TieringCeph Day Shanghai - Recovery Erasure Coding and Cache Tiering
Ceph Day Shanghai - Recovery Erasure Coding and Cache TieringCeph Community
 
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...Odinot Stanislas
 
Red Hat Ceph Storage Acceleration Utilizing Flash Technology
Red Hat Ceph Storage Acceleration Utilizing Flash Technology Red Hat Ceph Storage Acceleration Utilizing Flash Technology
Red Hat Ceph Storage Acceleration Utilizing Flash Technology Red_Hat_Storage
 
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...Red_Hat_Storage
 
Announcing Amazon EC2 F1 Instances with Custom FPGAs
Announcing Amazon EC2 F1 Instances with Custom FPGAsAnnouncing Amazon EC2 F1 Instances with Custom FPGAs
Announcing Amazon EC2 F1 Instances with Custom FPGAsAmazon Web Services
 
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Patrick McGarry
 
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...In-Memory Computing Summit
 
Ceph Day Tokyo - Delivering cost effective, high performance Ceph cluster
Ceph Day Tokyo - Delivering cost effective, high performance Ceph clusterCeph Day Tokyo - Delivering cost effective, high performance Ceph cluster
Ceph Day Tokyo - Delivering cost effective, high performance Ceph clusterCeph Community
 
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA Architecture
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA ArchitectureCeph Day Beijing - Ceph All-Flash Array Design Based on NUMA Architecture
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA ArchitectureDanielle Womboldt
 
Elastify Cloud-Native Spark Application with Persistent Memory
Elastify Cloud-Native Spark Application with Persistent MemoryElastify Cloud-Native Spark Application with Persistent Memory
Elastify Cloud-Native Spark Application with Persistent MemoryDatabricks
 
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...inwin stack
 
Deep Learning Computer Build
Deep Learning Computer BuildDeep Learning Computer Build
Deep Learning Computer BuildPetteriTeikariPhD
 
Red Hat Storage Day New York - New Reference Architectures
Red Hat Storage Day New York - New Reference ArchitecturesRed Hat Storage Day New York - New Reference Architectures
Red Hat Storage Day New York - New Reference ArchitecturesRed_Hat_Storage
 
Ceph Day San Jose - Object Storage for Big Data
Ceph Day San Jose - Object Storage for Big Data Ceph Day San Jose - Object Storage for Big Data
Ceph Day San Jose - Object Storage for Big Data Ceph Community
 
Spark meetup feb 2016
Spark meetup feb 2016Spark meetup feb 2016
Spark meetup feb 2016Todd Niven
 
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSD
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSDHigh-Performance Big Data Analytics with RDMA over NVM and NVMe-SSD
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSDinside-BigData.com
 
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Odinot Stanislas
 
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient Ceph Day Seoul - Ceph on Arm Scaleable and Efficient
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient Ceph Community
 
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server Ceph Community
 

Tendances (20)

No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...
No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...
No More Cumbersomeness: Automatic Predictive Modeling on Apache Spark with Ma...
 
Ceph Day Shanghai - Recovery Erasure Coding and Cache Tiering
Ceph Day Shanghai - Recovery Erasure Coding and Cache TieringCeph Day Shanghai - Recovery Erasure Coding and Cache Tiering
Ceph Day Shanghai - Recovery Erasure Coding and Cache Tiering
 
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...
Scale-out Storage on Intel® Architecture Based Platforms: Characterizing and ...
 
Red Hat Ceph Storage Acceleration Utilizing Flash Technology
Red Hat Ceph Storage Acceleration Utilizing Flash Technology Red Hat Ceph Storage Acceleration Utilizing Flash Technology
Red Hat Ceph Storage Acceleration Utilizing Flash Technology
 
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...
Red Hat Storage Day New York -Performance Intensive Workloads with Samsung NV...
 
Announcing Amazon EC2 F1 Instances with Custom FPGAs
Announcing Amazon EC2 F1 Instances with Custom FPGAsAnnouncing Amazon EC2 F1 Instances with Custom FPGAs
Announcing Amazon EC2 F1 Instances with Custom FPGAs
 
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
Using Recently Published Ceph Reference Architectures to Select Your Ceph Con...
 
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
IMC Summit 2016 Breakout - Noah Arliss - The Truth: How to Test Your Distribu...
 
Ceph Day Tokyo - Delivering cost effective, high performance Ceph cluster
Ceph Day Tokyo - Delivering cost effective, high performance Ceph clusterCeph Day Tokyo - Delivering cost effective, high performance Ceph cluster
Ceph Day Tokyo - Delivering cost effective, high performance Ceph cluster
 
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA Architecture
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA ArchitectureCeph Day Beijing - Ceph All-Flash Array Design Based on NUMA Architecture
Ceph Day Beijing - Ceph All-Flash Array Design Based on NUMA Architecture
 
Elastify Cloud-Native Spark Application with Persistent Memory
Elastify Cloud-Native Spark Application with Persistent MemoryElastify Cloud-Native Spark Application with Persistent Memory
Elastify Cloud-Native Spark Application with Persistent Memory
 
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...
Intel - optimizing ceph performance by leveraging intel® optane™ and 3 d nand...
 
Deep Learning Computer Build
Deep Learning Computer BuildDeep Learning Computer Build
Deep Learning Computer Build
 
Red Hat Storage Day New York - New Reference Architectures
Red Hat Storage Day New York - New Reference ArchitecturesRed Hat Storage Day New York - New Reference Architectures
Red Hat Storage Day New York - New Reference Architectures
 
Ceph Day San Jose - Object Storage for Big Data
Ceph Day San Jose - Object Storage for Big Data Ceph Day San Jose - Object Storage for Big Data
Ceph Day San Jose - Object Storage for Big Data
 
Spark meetup feb 2016
Spark meetup feb 2016Spark meetup feb 2016
Spark meetup feb 2016
 
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSD
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSDHigh-Performance Big Data Analytics with RDMA over NVM and NVMe-SSD
High-Performance Big Data Analytics with RDMA over NVM and NVMe-SSD
 
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
Hands-on Lab: How to Unleash Your Storage Performance by Using NVM Express™ B...
 
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient Ceph Day Seoul - Ceph on Arm Scaleable and Efficient
Ceph Day Seoul - Ceph on Arm Scaleable and Efficient
 
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
Ceph Day San Jose - All-Flahs Ceph on NUMA-Balanced Server
 

En vedette

Prestação de Contas do Deputado Dr Sinval Malheiros - número 1
Prestação de Contas do Deputado Dr Sinval Malheiros - número 1Prestação de Contas do Deputado Dr Sinval Malheiros - número 1
Prestação de Contas do Deputado Dr Sinval Malheiros - número 1SinvalMalheiros
 
Aktiviti budaya india
Aktiviti budaya indiaAktiviti budaya india
Aktiviti budaya indiaalikhmal
 
хърватска кристина
хърватска кристинахърватска кристина
хърватска кристинаCristina Drăniceanu
 
Using Problem Solving to Resolve Behaviour Problems
Using Problem Solving to Resolve Behaviour ProblemsUsing Problem Solving to Resolve Behaviour Problems
Using Problem Solving to Resolve Behaviour ProblemsEcem Cengiz
 
성공사례로 알아보는 B2C 마케팅
성공사례로 알아보는 B2C 마케팅성공사례로 알아보는 B2C 마케팅
성공사례로 알아보는 B2C 마케팅John UE
 
애자일활용사례
애자일활용사례애자일활용사례
애자일활용사례Dexter Jung
 
성공하는 애자일을 위한 짧은 이야기
성공하는 애자일을 위한 짧은 이야기성공하는 애자일을 위한 짧은 이야기
성공하는 애자일을 위한 짧은 이야기종범 고
 
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기 구글 heart 프레임워크
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기   구글 heart 프레임워크서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기   구글 heart 프레임워크
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기 구글 heart 프레임워크John UE
 

En vedette (13)

Panis angelicus
Panis angelicusPanis angelicus
Panis angelicus
 
Prestação de Contas do Deputado Dr Sinval Malheiros - número 1
Prestação de Contas do Deputado Dr Sinval Malheiros - número 1Prestação de Contas do Deputado Dr Sinval Malheiros - número 1
Prestação de Contas do Deputado Dr Sinval Malheiros - número 1
 
Aktiviti budaya india
Aktiviti budaya indiaAktiviti budaya india
Aktiviti budaya india
 
хърватска кристина
хърватска кристинахърватска кристина
хърватска кристина
 
2nd meeting in poland
2nd meeting in poland2nd meeting in poland
2nd meeting in poland
 
Invertebrates
Invertebrates Invertebrates
Invertebrates
 
Using Problem Solving to Resolve Behaviour Problems
Using Problem Solving to Resolve Behaviour ProblemsUsing Problem Solving to Resolve Behaviour Problems
Using Problem Solving to Resolve Behaviour Problems
 
성공사례로 알아보는 B2C 마케팅
성공사례로 알아보는 B2C 마케팅성공사례로 알아보는 B2C 마케팅
성공사례로 알아보는 B2C 마케팅
 
Meeting in romania
Meeting in romaniaMeeting in romania
Meeting in romania
 
Dear student
Dear studentDear student
Dear student
 
애자일활용사례
애자일활용사례애자일활용사례
애자일활용사례
 
성공하는 애자일을 위한 짧은 이야기
성공하는 애자일을 위한 짧은 이야기성공하는 애자일을 위한 짧은 이야기
성공하는 애자일을 위한 짧은 이야기
 
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기 구글 heart 프레임워크
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기   구글 heart 프레임워크서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기   구글 heart 프레임워크
서비스 및 제품기획자가 알아야 할 사용자경험 kpi 선정하기 구글 heart 프레임워크
 

Similaire à Introduce_non-volatile_generic_object_programming_model_for_In-Memory_Computing

Ceph Day Taipei - Accelerate Ceph via SPDK
Ceph Day Taipei - Accelerate Ceph via SPDK Ceph Day Taipei - Accelerate Ceph via SPDK
Ceph Day Taipei - Accelerate Ceph via SPDK Ceph Community
 
Boosting spark performance: An Overview of Techniques
Boosting spark performance: An Overview of TechniquesBoosting spark performance: An Overview of Techniques
Boosting spark performance: An Overview of TechniquesAhsan Javed Awan
 
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based HardwareRed hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based HardwareRed_Hat_Storage
 
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...Odinot Stanislas
 
Virtualizing Apache Spark with Justin Murray
Virtualizing Apache Spark with Justin MurrayVirtualizing Apache Spark with Justin Murray
Virtualizing Apache Spark with Justin MurrayDatabricks
 
Storage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, WhiptailStorage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, WhiptailInternet World
 
Deview 2013 rise of the wimpy machines - john mao
Deview 2013   rise of the wimpy machines - john maoDeview 2013   rise of the wimpy machines - john mao
Deview 2013 rise of the wimpy machines - john maoNAVER D2
 
Node Architecture Implications for In-Memory Data Analytics on Scale-in Clusters
Node Architecture Implications for In-Memory Data Analytics on Scale-in ClustersNode Architecture Implications for In-Memory Data Analytics on Scale-in Clusters
Node Architecture Implications for In-Memory Data Analytics on Scale-in ClustersAhsan Javed Awan
 
Profiling & Testing with Spark
Profiling & Testing with SparkProfiling & Testing with Spark
Profiling & Testing with SparkRoger Rafanell Mas
 
Představení produktové řady Oracle SPARC S7
Představení produktové řady Oracle SPARC S7Představení produktové řady Oracle SPARC S7
Představení produktové řady Oracle SPARC S7MarketingArrowECS_CZ
 
QCon2016--Drive Best Spark Performance on AI
QCon2016--Drive Best Spark Performance on AIQCon2016--Drive Best Spark Performance on AI
QCon2016--Drive Best Spark Performance on AILex Yu
 
FPGAs in the cloud? (October 2017)
FPGAs in the cloud? (October 2017)FPGAs in the cloud? (October 2017)
FPGAs in the cloud? (October 2017)Julien SIMON
 
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...Redis Labs
 
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance Ceph Community
 
AWS (Hadoop) Meetup 30.04.09
AWS (Hadoop) Meetup 30.04.09AWS (Hadoop) Meetup 30.04.09
AWS (Hadoop) Meetup 30.04.09Chris Purrington
 
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.J On The Beach
 
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory Computing
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory ComputingIMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory Computing
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory ComputingIn-Memory Computing Summit
 
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...Pawel Serwan
 
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...Spark Summit
 

Similaire à Introduce_non-volatile_generic_object_programming_model_for_In-Memory_Computing (20)

Ceph Day Taipei - Accelerate Ceph via SPDK
Ceph Day Taipei - Accelerate Ceph via SPDK Ceph Day Taipei - Accelerate Ceph via SPDK
Ceph Day Taipei - Accelerate Ceph via SPDK
 
Boosting spark performance: An Overview of Techniques
Boosting spark performance: An Overview of TechniquesBoosting spark performance: An Overview of Techniques
Boosting spark performance: An Overview of Techniques
 
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based HardwareRed hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
Red hat Storage Day LA - Designing Ceph Clusters Using Intel-Based Hardware
 
Ceph
CephCeph
Ceph
 
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...
Ceph: Open Source Storage Software Optimizations on Intel® Architecture for C...
 
Virtualizing Apache Spark with Justin Murray
Virtualizing Apache Spark with Justin MurrayVirtualizing Apache Spark with Justin Murray
Virtualizing Apache Spark with Justin Murray
 
Storage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, WhiptailStorage and performance- Batch processing, Whiptail
Storage and performance- Batch processing, Whiptail
 
Deview 2013 rise of the wimpy machines - john mao
Deview 2013   rise of the wimpy machines - john maoDeview 2013   rise of the wimpy machines - john mao
Deview 2013 rise of the wimpy machines - john mao
 
Node Architecture Implications for In-Memory Data Analytics on Scale-in Clusters
Node Architecture Implications for In-Memory Data Analytics on Scale-in ClustersNode Architecture Implications for In-Memory Data Analytics on Scale-in Clusters
Node Architecture Implications for In-Memory Data Analytics on Scale-in Clusters
 
Profiling & Testing with Spark
Profiling & Testing with SparkProfiling & Testing with Spark
Profiling & Testing with Spark
 
Představení produktové řady Oracle SPARC S7
Představení produktové řady Oracle SPARC S7Představení produktové řady Oracle SPARC S7
Představení produktové řady Oracle SPARC S7
 
QCon2016--Drive Best Spark Performance on AI
QCon2016--Drive Best Spark Performance on AIQCon2016--Drive Best Spark Performance on AI
QCon2016--Drive Best Spark Performance on AI
 
FPGAs in the cloud? (October 2017)
FPGAs in the cloud? (October 2017)FPGAs in the cloud? (October 2017)
FPGAs in the cloud? (October 2017)
 
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...
Scaling Redis Cluster Deployments for Genome Analysis (featuring LSU) - Terry...
 
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance
Ceph Day Shanghai - SSD/NVM Technology Boosting Ceph Performance
 
AWS (Hadoop) Meetup 30.04.09
AWS (Hadoop) Meetup 30.04.09AWS (Hadoop) Meetup 30.04.09
AWS (Hadoop) Meetup 30.04.09
 
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.
A Java Implementer's Guide to Boosting Apache Spark Performance by Tim Ellison.
 
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory Computing
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory ComputingIMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory Computing
IMCSummit 2015 - Day 2 General Session - Flash-Extending In-Memory Computing
 
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...
W jak sposób architektura hipekonwergentna cisco simplivity usprawni działani...
 
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...
Accelerating Spark Genome Sequencing in Cloud—A Data Driven Approach, Case St...
 

Introduce_non-volatile_generic_object_programming_model_for_In-Memory_Computing

  • 1. INTRODUCING A NON-VOLATILE GENERIC OBJECT PROGRAMMING MODEL FOR IN-MEMORY COMPUTING (SPARK) PERFORMANCE IMPROVEMENT YANPING WANG APACHE MNEMONIC (INCUBATING) PROJECT LEADER YANPINGW@APACHE.ORG
  • 2. STORAGE MEDIA EVOLUTION IS CHANGING FUTURE SW PROGRAMING
  • 3. STORAGE MEDIA EVOLUTION  Storage Media Evolution: NAND and 3D XPoint are filling the Speed and Capacity gap between DRAM and Hard Disk Drives  Software is facing the challenge on how to use them
  • 4. STORAGE MEDIA EVOLUTION IS CHANGING FUTURE SW PROGRAMING
  • 5. FROM IN-MEMORY TO IN-PLACE  Moving from In-Memory Computing to In-Place Computing  Mnemonic is a pioneering project to extend in-memory computing to in-place computing using next generation NVM storage media. (Picture was taken from Ken Gibson’s Storage Industry Summit January, 20, 2016)
  • 6. A NON-VOLATILE PROGRAMMING MODEL FOR JAVA APPLICATION DEVELOPERS
  • 7. WHY JAVA? • There are 300+ open source initiatives at the ASF. 221 of them are Java based, accounts for 59.9%. • Top Level Big Data Projects are Java based projects, include Hadoop, CloudStack, Spark, Cassandra, Hive, HBase, Flink, Solr….. • Java related optimization has significantly contributed to Big Data applications’ performance improvement.
  • 8. TRADITIONAL JAVA BASED BIG DATA FRAMEWORKS AND APPLICATIONS
  • 9. OUR PROPOSAL: OFF-HEAP, LEAVE DATA IN-PLACE
  • 10. NON-VOLATILE JAVA OBJECT MODEL • Provides Java programmers a mechanism to define objects and structures as Mnemonic-managed objects and structures. • Objects identified as @DurableEntity store their durable fields in volatile/non- volatile memory space. • @DurableGetter and @DurableSetter are used to access those fields. • Handlers are used to refer their connected durable object. • https://docs.oracle.com/javase/tutorial /java/annotations/basics.html
  • 11. AN EXAMPLE OF NON-VOLATILE LINKED LIST (LAZY LOADING)
  • 13. APACHE MNEMONIC STRUCTURE • Performance oriented architecture • Unified platform enabling Framework • Unique non-volatile object model • Flexible & extensible focal point for optimization
  • 14. MNEMONIC SEARCH EXAMPLE • Traditional search: SerDes, caching, pack/unpack, burst temp objects creation and destruction cost can be as high as 70-80% • Mnemonic Search: in- place creation & update, no SerDes, load on demand, no need to normalize
  • 15. MNEMONIC PROJECT POSITIONING • Apache Mnemonic (Incubating) is a Java based non-volatile memory library (API) for in-place structured data processing and computing • Addresses Big Data performance issues • High cost of Serialization/De-serialization • Frequent unpredicted long Garbage Collection pauses • Burst temporary object creation/destruction • Frequent high cost JNI call with stack marshalling • System paging and kernel structure updating due to massive data processing
  • 16. USE CASE EXAMPLE: SPARK + MNEMONIC =IN-PLACE MACHINE LEARNING COMPUTING
  • 17. NON-VOLATILE RDD CODE FOR SPARK • ~600 line of Non-Volatile RDD code is added into Spark, to alter data flow from Spark RDD via JVM heap to Mnemonic. • Spill RDDs to off-heap and disk can be avoid. • RDD data cache and local checkpoints can also be avoided when using Nonvolatile or Durable objects.
  • 18. INTERNAL STRUCTURED DATA REPRESENTATION  Spark experiment uses: 2D Linked List data structure to store data in persistent memory and storage.  N nodes: represent the first level of linked list.  V nodes: represent the values of first level of linked list that still is a linked list.  Only N2 and V1 (methods & volatile fields) are instantiated on heap.
  • 20. SPARK MLLIB KMEANS PERFORMANCE IMPROVEMENT • SPARK ML Kmeans implements one of the most popular clustering algorithms • Exp1: spark.storage.memoryFraction 0.6 • Exp2: spark.storage.memoryFraction 0.2
  • 21. CONFIGURATION FOR PERFORMANCE MEASUREMENT  Platform Software Configuration  OS: CentOS 7  JVM: JDK8 update 60  ParallelOldGC is used for Spark ML throughput workload Kmeans  Spark1.5.0  Platform Hardware  Haswell EP (72 cores, E5-2699 v3 @ 2.3GHz)  256GB DDR4-2133 on-board Memory  Storage media: SATA SSD  Device Model: INTEL SSDSC2BB016T4  Serial Number: BTWD442601311P6HGN  User Capacity: 1.60 TB  SATA Version is: SATA 3.0, 6.0 Gb/s  Spark Kmeans Experiments Setting: – System memory is reduced to 128G via static dummy ramfile occupancy – 20 executors with 4 GB JVM heap, 3 cores – Both of datasets are stored on SSD – Experiment 1 – spark.storage.memoryFraction 0.6 – 100,000,000 records with vector in 12 dimension (20 partitions) – Experiment 2 – spark.storage.memoryFraction 0.2 – 100,000,000 records with vector in 6 dimension (20 partitions)
  • 22. DISCLAIMERS Apache Mnemonic is an effort undergoing incubation at The Apache Software Foundation (ASF), sponsored by Apache Incubator PMC. Incubation is required of all newly accepted projects until a further review indicates that the infrastructure, communications, and decision making process have stabilized in a manner consistent with other successful ASF projects. While incubation status is not necessarily a reflection of the completeness or stability of the code, it does indicate that the project has yet to be fully endorsed by the ASF. While Mnemonic is operating within the time period where the application is processing their durable data, Mnemonic guarantees the persistency once Mnemonic is closed gracefully.
  • 23. PROJECT LINKS Mnemonic Community: dev@mnemonic.incubator.apache.org Source Code: https://github.com/apache/incubator-mnemonic JIRA Tracking: https://issues.apache.org/jira/browse/MNEMONIC Website: http://mnemonic.incubator.apache.org/ Contact: yanpingw@apache.org; garyw@apache.org