Learn about opportunities and challenges for accelerating big data middleware on modern high-performance computing (HPC) clusters by exploiting HPC technologies.
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
HPC Technologies to Accelerate Big Data Processing
1. Exploiting HPC Technologies to Accelerate Big Data
Processing (Hadoop, Spark, and Memcached)
Dhabaleswar K. (DK) Panda
The Ohio State University
E-mail: panda@cse.ohio-state.edu
http://www.cse.ohio-state.edu/~panda
Talk at Intel HPC Developer Conference (SC ‘16)
by
Xiaoyi Lu
The Ohio State University
E-mail: luxi@cse.ohio-state.edu
http://www.cse.ohio-state.edu/~luxi
2. Intel HPC Dev Conf (SC ‘16) 2Network Based Computing Laboratory
• Big Data has become the one of the most
important elements of business analytics
• Provides groundbreaking opportunities for
enterprise information management and
decision making
• The amount of data is exploding; companies
are capturing and digitizing more information
than ever
• The rate of information growth appears to be
exceeding Moore’s Law
Introduction to Big Data Applications and Analytics
3. Intel HPC Dev Conf (SC ‘16) 3Network Based Computing Laboratory
• Substantial impact on designing and utilizing data management and processing systems in multiple tiers
– Front-end data accessing and serving (Online)
• Memcached + DB (e.g. MySQL), HBase
– Back-end data analytics (Offline)
• HDFS, MapReduce, Spark
Data Management and Processing on Modern Clusters
4. Intel HPC Dev Conf (SC ‘16) 4Network Based Computing Laboratory
0
10
20
30
40
50
60
70
80
90
100
0
50
100
150
200
250
300
350
400
450
500
PercentageofClusters
NumberofClusters
Timeline
Percentage of Clusters
Number of Clusters
Trends for Commodity Computing Clusters in the Top 500
List (http://www.top500.org)
85%
5. Intel HPC Dev Conf (SC ‘16) 5Network Based Computing Laboratory
Drivers of Modern HPC Cluster Architectures
Tianhe – 2 Titan Stampede Tianhe – 1A
• Multi-core/many-core technologies
• Remote Direct Memory Access (RDMA)-enabled networking (InfiniBand and RoCE)
• Solid State Drives (SSDs), Non-Volatile Random-Access Memory (NVRAM), NVMe-SSD
• Accelerators (NVIDIA GPGPUs and Intel Xeon Phi)
Accelerators / Coprocessors
high compute density, high
performance/watt
>1 TFlop DP on a chip
High Performance Interconnects -
InfiniBand
<1usec latency, 100Gbps Bandwidth>Multi-core Processors SSD, NVMe-SSD, NVRAM
6. Intel HPC Dev Conf (SC ‘16) 6Network Based Computing Laboratory
• Advanced Interconnects and RDMA protocols
– InfiniBand
– 10-40 Gigabit Ethernet/iWARP
– RDMA over Converged Enhanced Ethernet (RoCE)
• Delivering excellent performance (Latency, Bandwidth and CPU Utilization)
• Has influenced re-designs of enhanced HPC middleware
– Message Passing Interface (MPI) and PGAS
– Parallel File Systems (Lustre, GPFS, ..)
• SSDs (SATA and NVMe)
• NVRAM and Burst Buffer
Trends in HPC Technologies
7. Intel HPC Dev Conf (SC ‘16) 7Network Based Computing Laboratory
Interconnects and Protocols in OpenFabrics Stack for HPC
(http://openfabrics.org)
Kernel
Space
Application /
Middleware
Verbs
Ethernet
Adapter
Ethernet
Switch
Ethernet
Driver
TCP/IP
1/10/40/100
GigE
InfiniBand
Adapter
InfiniBand
Switch
IPoIB
IPoIB
Ethernet
Adapter
Ethernet
Switch
Hardware
Offload
TCP/IP
10/40 GigE-
TOE
InfiniBand
Adapter
InfiniBand
Switch
User
Space
RSockets
RSockets
iWARP
Adapter
Ethernet
Switch
TCP/IP
User
Space
iWARP
RoCE
Adapter
Ethernet
Switch
RDMA
User
Space
RoCE
InfiniBand
Switch
InfiniBand
Adapter
RDMA
User
Space
IB Native
Sockets
Application /
Middleware Interface
Protocol
Adapter
Switch
InfiniBand
Adapter
InfiniBand
Switch
RDMA
SDP
SDP
8. Intel HPC Dev Conf (SC ‘16) 8Network Based Computing Laboratory
• 205 IB Clusters (41%) in the Jun’16 Top500 list
– (http://www.top500.org)
• Installations in the Top 50 (21 systems):
Large-scale InfiniBand Installations
220,800 cores (Pangea) in France (11th) 74,520 cores (Tsubame 2.5) at Japan/GSIC (31st)
462,462 cores (Stampede) at TACC (12th) 88,992 cores (Mistral) at DKRZ Germany (33rd)
185,344 cores (Pleiades) at NASA/Ames (15th) 194,616 cores (Cascade) at PNNL (34th)
72,800 cores Cray CS-Storm in US (19th) 76,032 cores (Makman-2) at Saudi Aramco (39th)
72,800 cores Cray CS-Storm in US (20th) 72,000 cores (Prolix) at Meteo France, France (40th)
124,200 cores (Topaz) SGI ICE at ERDC DSRC in US (21st ) 42,688 cores (Lomonosov-2) at Russia/MSU (41st)
72,000 cores (HPC2) in Italy (22nd) 60,240 cores SGI ICE X at JAEA Japan (43rd)
152,692 cores (Thunder) at AFRL/USA (25th) 70,272 cores (Tera-1000-1) at CEA France (44th)
147,456 cores (SuperMUC) in Germany (27th) 54,432 cores (Marconi) at CINECA Italy (46th)
86,016 cores (SuperMUC Phase 2) in Germany (28th) and many more!
9. Intel HPC Dev Conf (SC ‘16) 9Network Based Computing Laboratory
• Introduced in Oct 2000
• High Performance Data Transfer
– Interprocessor communication and I/O
– Low latency (<1.0 microsec), High bandwidth (up to 12.5 GigaBytes/sec -> 100Gbps), and
low CPU utilization (5-10%)
• Multiple Operations
– Send/Recv
– RDMA Read/Write
– Atomic Operations (very unique)
• high performance and scalable implementations of distributed locks, semaphores, collective
communication operations
• Leading to big changes in designing
– HPC clusters
– File systems
– Cloud computing systems
– Grid computing systems
Open Standard InfiniBand Networking Technology
10. Intel HPC Dev Conf (SC ‘16) 10Network Based Computing Laboratory
How Can HPC Clusters with High-Performance Interconnect and Storage
Architectures Benefit Big Data Applications?
Bring HPC and Big Data processing into a
“convergent trajectory”!
What are the major
bottlenecks in current Big
Data processing
middleware (e.g. Hadoop,
Spark, and Memcached)?
Can the bottlenecks be
alleviated with new
designs by taking
advantage of HPC
technologies?
Can RDMA-enabled
high-performance
interconnects
benefit Big Data
processing?
Can HPC Clusters with
high-performance
storage systems (e.g.
SSD, parallel file
systems) benefit Big
Data applications?
How much
performance benefits
can be achieved
through enhanced
designs?
How to design
benchmarks for
evaluating the
performance of Big
Data middleware on
HPC clusters?
11. Intel HPC Dev Conf (SC ‘16) 11Network Based Computing Laboratory
Designing Communication and I/O Libraries for Big Data Systems:
Challenges
Big Data Middleware
(HDFS, MapReduce, HBase, Spark and Memcached)
Networking Technologies
(InfiniBand, 1/10/40/100 GigE
and Intelligent NICs)
Storage Technologies
(HDD, SSD, and NVMe-SSD)
Programming Models
(Sockets)
Applications
Commodity Computing System
Architectures
(Multi- and Many-core
architectures and accelerators)
Other Protocols?
Communication and I/O Library
Point-to-Point
Communication
QoS
Threaded Models
and Synchronization
Fault-ToleranceI/O and File Systems
Virtualization
Benchmarks
Upper level
Changes?
12. Intel HPC Dev Conf (SC ‘16) 12Network Based Computing Laboratory
• Sockets not designed for high-performance
– Stream semantics often mismatch for upper layers
– Zero-copy not available for non-blocking sockets
Can Big Data Processing Systems be Designed with High-
Performance Networks and Protocols?
Current Design
Application
Sockets
1/10/40/100 GigE
Network
Our Approach
Application
OSU Design
10/40/100 GigE or
InfiniBand
Verbs Interface
13. Intel HPC Dev Conf (SC ‘16) 13Network Based Computing Laboratory
• RDMA for Apache Spark
• RDMA for Apache Hadoop 2.x (RDMA-Hadoop-2.x)
– Plugins for Apache, Hortonworks (HDP) and Cloudera (CDH) Hadoop distributions
• RDMA for Apache HBase
• RDMA for Memcached (RDMA-Memcached)
• RDMA for Apache Hadoop 1.x (RDMA-Hadoop)
• OSU HiBD-Benchmarks (OHB)
– HDFS, Memcached, and HBase Micro-benchmarks
• http://hibd.cse.ohio-state.edu
• Users Base: 195 organizations from 27 countries
• More than 18,500 downloads from the project site
• RDMA for Impala (upcoming)
The High-Performance Big Data (HiBD) Project
Available for InfiniBand and RoCE
14. Intel HPC Dev Conf (SC ‘16) 14Network Based Computing Laboratory
• High-Performance Design of Hadoop over RDMA-enabled Interconnects
– High performance RDMA-enhanced design with native InfiniBand and RoCE support at the verbs-level for HDFS, MapReduce, and
RPC components
– Enhanced HDFS with in-memory and heterogeneous storage
– High performance design of MapReduce over Lustre
– Memcached-based burst buffer for MapReduce over Lustre-integrated HDFS (HHH-L-BB mode)
– Plugin-based architecture supporting RDMA-based designs for Apache Hadoop, CDH and HDP
– Easily configurable for different running modes (HHH, HHH-M, HHH-L, HHH-L-BB, and MapReduce over Lustre) and different
protocols (native InfiniBand, RoCE, and IPoIB)
• Current release: 1.1.0
– Based on Apache Hadoop 2.7.3
– Compliant with Apache Hadoop 2.7.1, HDP 2.5.0.3 and CDH 5.8.2 APIs and applications
– Tested with
• Mellanox InfiniBand adapters (DDR, QDR, FDR, and EDR)
• RoCE support with Mellanox adapters
• Various multi-core platforms
• Different file systems with disks and SSDs and Lustre
RDMA for Apache Hadoop 2.x Distribution
http://hibd.cse.ohio-state.edu
15. Intel HPC Dev Conf (SC ‘16) 15Network Based Computing Laboratory
• HHH: Heterogeneous storage devices with hybrid replication schemes are supported in this mode of operation to have better fault-tolerance as well
as performance. This mode is enabled by default in the package.
• HHH-M: A high-performance in-memory based setup has been introduced in this package that can be utilized to perform all I/O operations in-
memory and obtain as much performance benefit as possible.
• HHH-L: With parallel file systems integrated, HHH-L mode can take advantage of the Lustre available in the cluster.
• HHH-L-BB: This mode deploys a Memcached-based burst buffer system to reduce the bandwidth bottleneck of shared file system access. The burst
buffer design is hosted by Memcached servers, each of which has a local SSD.
• MapReduce over Lustre, with/without local disks: Besides, HDFS based solutions, this package also provides support to run MapReduce jobs on top
of Lustre alone. Here, two different modes are introduced: with local disks and without local disks.
• Running with Slurm and PBS: Supports deploying RDMA for Apache Hadoop 2.x with Slurm and PBS in different running modes (HHH, HHH-M, HHH-
L, and MapReduce over Lustre).
Different Modes of RDMA for Apache Hadoop 2.x
16. Intel HPC Dev Conf (SC ‘16) 16Network Based Computing Laboratory
• High-Performance Design of Spark over RDMA-enabled Interconnects
– High performance RDMA-enhanced design with native InfiniBand and RoCE support at the verbs-level
for Spark
– RDMA-based data shuffle and SEDA-based shuffle architecture
– Non-blocking and chunk-based data transfer
– Easily configurable for different protocols (native InfiniBand, RoCE, and IPoIB)
• Current release: 0.9.1
– Based on Apache Spark 1.5.1
– Tested with
• Mellanox InfiniBand adapters (DDR, QDR and FDR)
• RoCE support with Mellanox adapters
• Various multi-core platforms
• RAM disks, SSDs, and HDD
– http://hibd.cse.ohio-state.edu
RDMA for Apache Spark Distribution
17. Intel HPC Dev Conf (SC ‘16) 17Network Based Computing Laboratory
• RDMA for Apache Hadoop 2.x and RDMA for Apache Spark are installed and
available on SDSC Comet.
– Examples for various modes of usage are available in:
• RDMA for Apache Hadoop 2.x: /share/apps/examples/HADOOP
• RDMA for Apache Spark: /share/apps/examples/SPARK/
– Please email help@xsede.org (reference Comet as the machine, and SDSC as the
site) if you have any further questions about usage and configuration.
• RDMA for Apache Hadoop is also available on Chameleon Cloud as an
appliance
– https://www.chameleoncloud.org/appliances/17/
HiBD Packages on SDSC Comet and Chameleon Cloud
M. Tatineni, X. Lu, D. J. Choi, A. Majumdar, and D. K. Panda, Experiences and Benefits of Running RDMA Hadoop and Spark on SDSC
Comet, XSEDE’16, July 2016
18. Intel HPC Dev Conf (SC ‘16) 18Network Based Computing Laboratory
• High-Performance Design of HBase over RDMA-enabled Interconnects
– High performance RDMA-enhanced design with native InfiniBand and RoCE support at the verbs-level
for HBase
– Compliant with Apache HBase 1.1.2 APIs and applications
– On-demand connection setup
– Easily configurable for different protocols (native InfiniBand, RoCE, and IPoIB)
• Current release: 0.9.1
– Based on Apache HBase 1.1.2
– Tested with
• Mellanox InfiniBand adapters (DDR, QDR, FDR, and EDR)
• RoCE support with Mellanox adapters
• Various multi-core platforms
– http://hibd.cse.ohio-state.edu
RDMA for Apache HBase Distribution
19. Intel HPC Dev Conf (SC ‘16) 19Network Based Computing Laboratory
• High-Performance Design of Memcached over RDMA-enabled Interconnects
– High performance RDMA-enhanced design with native InfiniBand and RoCE support at the verbs-level for Memcached and
libMemcached components
– High performance design of SSD-Assisted Hybrid Memory
– Non-Blocking Libmemcached Set/Get API extensions
– Support for burst-buffer mode in Lustre-integrated design of HDFS in RDMA for Apache Hadoop-2.x
– Easily configurable for native InfiniBand, RoCE and the traditional sockets-based support (Ethernet and InfiniBand with
IPoIB)
• Current release: 0.9.5
– Based on Memcached 1.4.24 and libMemcached 1.0.18
– Compliant with libMemcached APIs and applications
– Tested with
• Mellanox InfiniBand adapters (DDR, QDR, FDR, and EDR)
• RoCE support with Mellanox adapters
• Various multi-core platforms
• SSD
– http://hibd.cse.ohio-state.edu
RDMA for Memcached Distribution
20. Intel HPC Dev Conf (SC ‘16) 20Network Based Computing Laboratory
• Micro-benchmarks for Hadoop Distributed File System (HDFS)
– Sequential Write Latency (SWL) Benchmark, Sequential Read Latency (SRL) Benchmark,
Random Read Latency (RRL) Benchmark, Sequential Write Throughput (SWT) Benchmark,
Sequential Read Throughput (SRT) Benchmark
– Support benchmarking of
• Apache Hadoop 1.x and 2.x HDFS, Hortonworks Data Platform (HDP) HDFS, Cloudera Distribution of
Hadoop (CDH) HDFS
• Micro-benchmarks for Memcached
– Get Benchmark, Set Benchmark, and Mixed Get/Set Benchmark, Non-Blocking API Latency
Benchmark, Hybrid Memory Latency Benchmark
• Micro-benchmarks for HBase
– Get Latency Benchmark, Put Latency Benchmark
• Current release: 0.9.1
• http://hibd.cse.ohio-state.edu
OSU HiBD Micro-Benchmark (OHB) Suite – HDFS, Memcached, and HBase
21. Intel HPC Dev Conf (SC ‘16) 21Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
22. Intel HPC Dev Conf (SC ‘16) 22Network Based Computing Laboratory
• Enables high performance RDMA communication, while supporting traditional socket interface
• JNI Layer bridges Java based HDFS with communication library written in native code
Design Overview of HDFS with RDMA
HDFS
Verbs
RDMA Capable Networks
(IB, iWARP, RoCE ..)
Applications
1/10/40/100 GigE, IPoIB
Network
Java Socket Interface Java Native Interface (JNI)
WriteOthers
OSU Design
• Design Features
– RDMA-based HDFS write
– RDMA-based HDFS
replication
– Parallel replication support
– On-demand connection
setup
– InfiniBand/RoCE support
N. S. Islam, M. W. Rahman, J. Jose, R. Rajachandrasekar, H. Wang, H. Subramoni, C. Murthy and D. K. Panda , High Performance RDMA-Based Design of HDFS
over InfiniBand , Supercomputing (SC), Nov 2012
N. Islam, X. Lu, W. Rahman, and D. K. Panda, SOR-HDFS: A SEDA-based Approach to Maximize Overlapping in RDMA-Enhanced HDFS, HPDC '14, June 2014
23. Intel HPC Dev Conf (SC ‘16) 23Network Based Computing Laboratory
Triple-H
Heterogeneous Storage
• Design Features
– Three modes
• Default (HHH)
• In-Memory (HHH-M)
• Lustre-Integrated (HHH-L)
– Policies to efficiently utilize the heterogeneous
storage devices
• RAM, SSD, HDD, Lustre
– Eviction/Promotion based on data usage
pattern
– Hybrid Replication
– Lustre-Integrated mode:
• Lustre-based fault-tolerance
Enhanced HDFS with In-Memory and Heterogeneous Storage
Hybrid Replication
Data Placement Policies
Eviction/Promotion
RAM Disk SSD HDD
Lustre
N. Islam, X. Lu, M. W. Rahman, D. Shankar, and D. K. Panda, Triple-H: A Hybrid Approach to Accelerate HDFS on HPC Clusters
with Heterogeneous Storage Architecture, CCGrid ’15, May 2015
Applications
24. Intel HPC Dev Conf (SC ‘16) 24Network Based Computing Laboratory
Design Overview of MapReduce with RDMA
MapReduce
Verbs
RDMA Capable Networks
(IB, iWARP, RoCE ..)
OSU Design
Applications
1/10/40/100 GigE, IPoIB
Network
Java Socket Interface Java Native Interface (JNI)
Job
Tracker
Task
Tracker
Map
Reduce
• Enables high performance RDMA communication, while supporting traditional socket interface
• JNI Layer bridges Java based MapReduce with communication library written in native code
• Design Features
– RDMA-based shuffle
– Prefetching and caching map output
– Efficient Shuffle Algorithms
– In-memory merge
– On-demand Shuffle Adjustment
– Advanced overlapping
• map, shuffle, and merge
• shuffle, merge, and reduce
– On-demand connection setup
– InfiniBand/RoCE support
M. W. Rahman, X. Lu, N. S. Islam, and D. K. Panda, HOMR: A Hybrid Approach to Exploit Maximum Overlapping in
MapReduce over High Performance Interconnects, ICS, June 2014
25. Intel HPC Dev Conf (SC ‘16) 25Network Based Computing Laboratory
0
50
100
150
200
250
300
350
400
80 120 160
ExecutionTime(s)
Data Size (GB)
IPoIB (EDR)
OSU-IB (EDR)
0
100
200
300
400
500
600
700
800
80 160 240
ExecutionTime(s)
Data Size (GB)
IPoIB (EDR)
OSU-IB (EDR)
Performance Numbers of RDMA for Apache Hadoop 2.x –
RandomWriter & TeraGen in OSU-RI2 (EDR)
Cluster with 8 Nodes with a total of 64 maps
• RandomWriter
– 3x improvement over IPoIB
for 80-160 GB file size
• TeraGen
– 4x improvement over IPoIB for
80-240 GB file size
RandomWriter TeraGen
Reduced by 3x Reduced by 4x
26. Intel HPC Dev Conf (SC ‘16) 26Network Based Computing Laboratory
0
100
200
300
400
500
600
700
800
80 120 160
ExecutionTime(s)
Data Size (GB)
IPoIB (EDR)
OSU-IB (EDR)
Performance Numbers of RDMA for Apache Hadoop 2.x – Sort & TeraSort
in OSU-RI2 (EDR)
Cluster with 8 Nodes with a total of
64 maps and 32 reduces
• Sort
– 61% improvement over IPoIB for
80-160 GB data
• TeraSort
– 18% improvement over IPoIB for
80-240 GB data
Reduced by 61%
Reduced by 18%
Cluster with 8 Nodes with a total of
64 maps and 14 reduces
Sort TeraSort
0
100
200
300
400
500
600
80 160 240
ExecutionTime(s)
Data Size (GB)
IPoIB (EDR)
OSU-IB (EDR)
27. Intel HPC Dev Conf (SC ‘16) 27Network Based Computing Laboratory
Evaluation of HHH and HHH-L with Applications
HDFS (FDR) HHH (FDR)
60.24 s 48.3 s
CloudBurstMR-MSPolyGraph
0
200
400
600
800
1000
4 6 8
ExecutionTime(s)
Concurrent maps per host
HDFS Lustre HHH-L Reduced by 79%
• MR-MSPolygraph on OSU RI with
1,000 maps
– HHH-L reduces the execution time
by 79% over Lustre, 30% over HDFS
• CloudBurst on TACC Stampede
– With HHH: 19% improvement over
HDFS
28. Intel HPC Dev Conf (SC ‘16) 28Network Based Computing Laboratory
Evaluation with Spark on SDSC Gordon (HHH vs. Tachyon/Alluxio)
• For 200GB TeraGen on 32 nodes
– Spark-TeraGen: HHH has 2.4x improvement over Tachyon; 2.3x over HDFS-IPoIB (QDR)
– Spark-TeraSort: HHH has 25.2% improvement over Tachyon; 17% over HDFS-IPoIB (QDR)
0
20
40
60
80
100
120
140
160
180
8:50 16:100 32:200
ExecutionTime(s)
Cluster Size : Data Size (GB)
IPoIB (QDR) Tachyon OSU-IB (QDR)
0
100
200
300
400
500
600
700
8:50 16:100 32:200
ExecutionTime(s)
Cluster Size : Data Size (GB)
Reduced
by 2.4x
Reduced by 25.2%
TeraGen TeraSort
N. Islam, M. W. Rahman, X. Lu, D. Shankar, and D. K. Panda, Performance Characterization and Acceleration of In-Memory File
Systems for Hadoop and Spark Applications on HPC Clusters, IEEE BigData ’15, October 2015
29. Intel HPC Dev Conf (SC ‘16) 29Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
30. Intel HPC Dev Conf (SC ‘16) 30Network Based Computing Laboratory
Performance Numbers of RDMA for Apache HBase – OHB in
SDSC-Comet
0
2000
4000
6000
8000
10000
12000
14000
2 8 32 128 512 2k 8k 32k 128k 512k
Latency(us)
Message Size
Put
IPoIB RDMA
0
1000
2000
3000
4000
5000
2 8 32 128 512 2k 8k 32k 128k512k
Latency(us)
Message Size
Get
IPoIB RDMA
• Up to 8.6x improvement over
IPoIB
Reduced by
8.6x
• Up to 5.3x improvement over
IPoIB
Reduced by
3.8x
Reduced by
5.3x
Reduced by
8.6x
Evaluation with OHB Put and Get Micro-Benchmarks (1 Server, 1 Client)
31. Intel HPC Dev Conf (SC ‘16) 31Network Based Computing Laboratory
Performance Numbers of RDMA for Apache HBase –
YCSB in SDSC-Comet
0
20000
40000
60000
80000
100000
120000
4 8 16 32 64 128
Throughput(Operations/sec)
Number of clients
Workload A (50% read, 50% update)
IPoIB RDMA
0
100000
200000
300000
400000
500000
4 8 16 32 64 128
Throughput(Operations/sec)
Number of clients
Workload C (100% read)
IPoIB RDMA
• Up to 2.4x improvement over
IPoIB
Increased by
83%
• Up to 3.6x improvement over
IPoIB
Increased by
3.6x
Evaluation with YCSB Workloads A and C (4 Servers)
32. Intel HPC Dev Conf (SC ‘16) 32Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
33. Intel HPC Dev Conf (SC ‘16) 33Network Based Computing Laboratory
• Design Features
– RDMA based shuffle plugin
– SEDA-based architecture
– Dynamic connection
management and sharing
– Non-blocking data transfer
– Off-JVM-heap buffer
management
– InfiniBand/RoCE support
Design Overview of Spark with RDMA
• Enables high performance RDMA communication, while supporting traditional socket interface
• JNI Layer bridges Scala based Spark with communication library written in native code
X. Lu, M. W. Rahman, N. Islam, D. Shankar, and D. K. Panda, Accelerating Spark with RDMA for Big Data Processing: Early Experiences, Int'l Symposium on High
Performance Interconnects (HotI'14), August 2014
X. Lu, D. Shankar, S. Gugnani, and D. K. Panda, High-Performance Design of Apache Spark with RDMA and Its Benefits on Various Workloads, IEEE BigData ‘16, Dec. 2016.
Spark Core
RDMA Capable Networks
(IB, iWARP, RoCE ..)
Apache Spark Benchmarks/Applications/Libraries/Frameworks
1/10/40/100 GigE, IPoIB Network
Java Socket Interface Java Native Interface (JNI)
Native RDMA-based Comm. Engine
Shuffle Manager (Sort, Hash, Tungsten-Sort)
Block Transfer Service (Netty, NIO, RDMA-Plugin)
Netty
Server
NIO
Server
RDMA
Server
Netty
Client
NIO
Client
RDMA
Client
34. Intel HPC Dev Conf (SC ‘16) 34Network Based Computing Laboratory
• InfiniBand FDR, SSD, 64 Worker Nodes, 1536 Cores, (1536M 1536R)
• RDMA-based design for Spark 1.5.1
• RDMA vs. IPoIB with 1536 concurrent tasks, single SSD per node.
– SortBy: Total time reduced by up to 80% over IPoIB (56Gbps)
– GroupBy: Total time reduced by up to 74% over IPoIB (56Gbps)
Performance Evaluation on SDSC Comet – SortBy/GroupBy
64 Worker Nodes, 1536 cores, SortByTest Total Time 64 Worker Nodes, 1536 cores, GroupByTest Total Time
0
50
100
150
200
250
300
64 128 256
Time(sec)
Data Size (GB)
IPoIB
RDMA
0
50
100
150
200
250
64 128 256
Time(sec)
Data Size (GB)
IPoIB
RDMA
74%80%
35. Intel HPC Dev Conf (SC ‘16) 35Network Based Computing Laboratory
• InfiniBand FDR, SSD, 32/64 Worker Nodes, 768/1536 Cores, (768/1536M 768/1536R)
• RDMA-based design for Spark 1.5.1
• RDMA vs. IPoIB with 768/1536 concurrent tasks, single SSD per node.
– 32 nodes/768 cores: Total time reduced by 37% over IPoIB (56Gbps)
– 64 nodes/1536 cores: Total time reduced by 43% over IPoIB (56Gbps)
Performance Evaluation on SDSC Comet – HiBench PageRank
32 Worker Nodes, 768 cores, PageRank Total Time 64 Worker Nodes, 1536 cores, PageRank Total Time
0
50
100
150
200
250
300
350
400
450
Huge BigData Gigantic
Time(sec)
Data Size (GB)
IPoIB
RDMA
0
100
200
300
400
500
600
700
800
Huge BigData Gigantic
Time(sec)
Data Size (GB)
IPoIB
RDMA
43%37%
36. Intel HPC Dev Conf (SC ‘16) 36Network Based Computing Laboratory
Performance Evaluation on SDSC Comet: Astronomy Application
• Kira Toolkit1: Distributed astronomy image
processing toolkit implemented using Apache Spark.
• Source extractor application, using a 65GB dataset
from the SDSS DR2 survey that comprises 11,150
image files.
• Compare RDMA Spark performance with the
standard apache implementation using IPoIB.
1. Z. Zhang, K. Barbary, F. A. Nothaft, E.R. Sparks, M.J. Franklin, D.A.
Patterson, S. Perlmutter. Scientific Computing meets Big Data Technology: An
Astronomy Use Case. CoRR, vol: abs/1507.03325, Aug 2015.
0
20
40
60
80
100
120
RDMA Spark Apache Spark
(IPoIB)
21 %
Execution times (sec) for Kira SE
benchmark using 65 GB dataset, 48 cores.
M. Tatineni, X. Lu, D. J. Choi, A. Majumdar, and D. K. Panda, Experiences and Benefits of Running RDMA Hadoop and Spark on SDSC
Comet, XSEDE’16, July 2016
37. Intel HPC Dev Conf (SC ‘16) 37Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs and Studies
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
38. Intel HPC Dev Conf (SC ‘16) 38Network Based Computing Laboratory
1
10
100
1000
1
2
4
8
16
32
64
128
256
512
1K
2K
4K
Time(us)
Message Size
OSU-IB (FDR)
0
200
400
600
800
16 32 64 128 256 512 102420484080
Thousandsof
Transactionsper
Second(TPS)
No. of Clients
• Memcached Get latency
– 4 bytes OSU-IB: 2.84 us; IPoIB: 75.53 us, 2K bytes OSU-IB: 4.49 us; IPoIB: 123.42 us
• Memcached Throughput (4bytes)
– 4080 clients OSU-IB: 556 Kops/sec, IPoIB: 233 Kops/s, Nearly 2X improvement in throughput
Memcached GET Latency Memcached Throughput
Memcached Performance (FDR Interconnect)
Experiments on TACC Stampede (Intel SandyBridge Cluster, IB: FDR)
Latency Reduced
by nearly 20X
2X
J. Jose, H. Subramoni, M. Luo, M. Zhang, J. Huang, M. W. Rahman, N. Islam, X. Ouyang, H. Wang, S. Sur and D. K. Panda, Memcached Design on High
Performance RDMA Capable Interconnects, ICPP’11
J. Jose, H. Subramoni, K. Kandalla, M. W. Rahman, H. Wang, S. Narravula, and D. K. Panda, Scalable Memcached design for InfiniBand Clusters using Hybrid
Transport, CCGrid’12
39. Intel HPC Dev Conf (SC ‘16) 39Network Based Computing Laboratory
• Illustration with Read-Cache-Read access pattern using modified mysqlslap load testing
tool
• Memcached-RDMA can
- improve query latency by up to 66% over IPoIB (32Gbps)
- throughput by up to 69% over IPoIB (32Gbps)
Micro-benchmark Evaluation for OLDP workloads
0
1
2
3
4
5
6
7
8
64 96 128 160 320 400
Latency(sec)
No. of Clients
Memcached-IPoIB (32Gbps)
Memcached-RDMA (32Gbps)
0
1000
2000
3000
4000
64 96 128 160 320 400
Throughput(Kq/s)
No. of Clients
Memcached-IPoIB (32Gbps)
Memcached-RDMA (32Gbps)
D. Shankar, X. Lu, J. Jose, M. W. Rahman, N. Islam, and D. K. Panda, Can RDMA Benefit On-Line Data Processing Workloads
with Memcached and MySQL, ISPASS’15
Reduced by 66%
40. Intel HPC Dev Conf (SC ‘16) 40Network Based Computing Laboratory
• RDMA-based Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs and Studies
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
41. Intel HPC Dev Conf (SC ‘16) 41Network Based Computing Laboratory
• The current benchmarks provide some performance behavior
• However, do not provide any information to the designer/developer on:
– What is happening at the lower-layer?
– Where the benefits are coming from?
– Which design is leading to benefits or bottlenecks?
– Which component in the design needs to be changed and what will be its impact?
– Can performance gain/loss at the lower-layer be correlated to the performance
gain/loss observed at the upper layer?
Are the Current Benchmarks Sufficient for Big Data?
42. Intel HPC Dev Conf (SC ‘16) 42Network Based Computing Laboratory
Big Data Middleware
(HDFS, MapReduce, HBase, Spark and Memcached)
Networking Technologies
(InfiniBand, 1/10/40/100 GigE
and Intelligent NICs)
Storage Technologies
(HDD, SSD, and NVMe-SSD)
Programming Models
(Sockets)
Applications
Commodity Computing System
Architectures
(Multi- and Many-core
architectures and accelerators)
Other Protocols?
Communication and I/O Library
Point-to-Point
Communication
QoS
Threaded Models
and Synchronization
Fault-ToleranceI/O and File Systems
Virtualization
Benchmarks
RDMA Protocols
Challenges in Benchmarking of RDMA-based Designs
Current
Benchmarks
No Benchmarks
Correlation?
43. Intel HPC Dev Conf (SC ‘16) 43Network Based Computing Laboratory
Big Data Middleware
(HDFS, MapReduce, HBase, Spark and Memcached)
Networking Technologies
(InfiniBand, 1/10/40/100 GigE
and Intelligent NICs)
Storage Technologies
(HDD, SSD, and NVMe-SSD)
Programming Models
(Sockets)
Applications
Commodity Computing System
Architectures
(Multi- and Many-core
architectures and accelerators)
Other Protocols?
Communication and I/O Library
Point-to-Point
Communication
QoS
Threaded Models
and Synchronization
Fault-ToleranceI/O and File Systems
Virtualization
Benchmarks
RDMA Protocols
Iterative Process – Requires Deeper Investigation and Design for
Benchmarking Next Generation Big Data Systems and Applications
Applications-Level
Benchmarks
Micro-
Benchmarks
44. Intel HPC Dev Conf (SC ‘16) 44Network Based Computing Laboratory
• HDFS Benchmarks
– Sequential Write Latency (SWL) Benchmark
– Sequential Read Latency (SRL) Benchmark
– Random Read Latency (RRL) Benchmark
– Sequential Write Throughput (SWT) Benchmark
– Sequential Read Throughput (SRT) Benchmark
• Memcached Benchmarks
– Get, Set and Mixed Get/Set Benchmarks
– Non-blocking and Hybrid Memory Latency Benchmarks
• HBase Benchmarks
– Get and Put Latency Benchmarks
• Available as a part of OHB 0.9.1
OSU HiBD Benchmarks (OHB)
N. S. Islam, X. Lu, M. W. Rahman, J. Jose, and D.
K. Panda, A Micro-benchmark Suite for
Evaluating HDFS Operations on Modern
Clusters, Int'l Workshop on Big Data
Benchmarking (WBDB '12), December 2012
D. Shankar, X. Lu, M. W. Rahman, N. Islam, and
D. K. Panda, A Micro-Benchmark Suite for
Evaluating Hadoop MapReduce on High-
Performance Networks, BPOE-5 (2014)
X. Lu, M. W. Rahman, N. Islam, and D. K. Panda,
A Micro-Benchmark Suite for Evaluating Hadoop
RPC on High-Performance Networks, Int'l
Workshop on Big Data Benchmarking (WBDB
'13), July 2013
To be Released
45. Intel HPC Dev Conf (SC ‘16) 45Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
46. Intel HPC Dev Conf (SC ‘16) 46Network Based Computing Laboratory
• Design Features
– Memcached-based burst-buffer
system
• Hides latency of parallel file
system access
• Read from local storage and
Memcached
– Data locality achieved by writing data
to local storage
– Different approaches of integration
with parallel file system to guarantee
fault-tolerance
Accelerating I/O Performance of Big Data Analytics
through RDMA-based Key-Value Store
Application
I/O Forwarding Module
Map/Reduce Task DataNode
Local Disk
Data LocalityFault-tolerance
Lustre
Memcached-based Burst Buffer System
47. Intel HPC Dev Conf (SC ‘16) 47Network Based Computing Laboratory
Evaluation with PUMA Workloads
Gains on OSU RI with our approach (Mem-bb) on 24 nodes
• SequenceCount: 34.5% over Lustre, 40% over HDFS
• RankedInvertedIndex: 27.3% over Lustre, 48.3% over HDFS
• HistogramRating: 17% over Lustre, 7% over HDFS
0
500
1000
1500
2000
2500
3000
3500
4000
4500
SeqCount RankedInvIndex HistoRating
ExecutionTime(s)
Workloads
HDFS (32Gbps)
Lustre (32Gbps)
Mem-bb (32Gbps)
48.3%
40%
17%
N. S. Islam, D. Shankar, X. Lu, M.
W. Rahman, and D. K. Panda,
Accelerating I/O Performance of
Big Data Analytics with RDMA-
based Key-Value Store, ICPP
’15, September 2015
48. Intel HPC Dev Conf (SC ‘16) 48Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
49. Intel HPC Dev Conf (SC ‘16) 49Network Based Computing Laboratory
• ohb_memlat & ohb_memthr latency & throughput micro-benchmarks
• Memcached-RDMA can
- improve query latency by up to 70% over IPoIB (32Gbps)
- improve throughput by up to 2X over IPoIB (32Gbps)
- No overhead in using hybrid mode when all data can fit in memory
Performance Benefits of Hybrid Memcached (Memory + SSD) on
SDSC-Gordon
0
2
4
6
8
10
64 128 256 512 1024
Throughput(milliontrans/sec)
No. of Clients
IPoIB (32Gbps)
RDMA-Mem (32Gbps)
RDMA-Hybrid (32Gbps)
0
100
200
300
400
500
Averagelatency(us)
Message Size (Bytes)
2X
50. Intel HPC Dev Conf (SC ‘16) 50Network Based Computing Laboratory
– Memcached latency test with Zipf distribution, server with 1 GB memory, 32 KB key-value pair size, total
size of data accessed is 1 GB (when data fits in memory) and 1.5 GB (when data does not fit in memory)
– When data fits in memory: RDMA-Mem/Hybrid gives 5x improvement over IPoIB-Mem
– When data does not fit in memory: RDMA-Hybrid gives 2x-2.5x over IPoIB/RDMA-Mem
Performance Evaluation on IB FDR + SATA/NVMe SSDs (Hybrid Memory)
0
500
1000
1500
2000
2500
Set Get Set Get Set Get Set Get Set Get Set Get Set Get Set Get
IPoIB-Mem RDMA-Mem RDMA-Hybrid-SATA RDMA-Hybrid-
NVMe
IPoIB-Mem RDMA-Mem RDMA-Hybrid-SATA RDMA-Hybrid-
NVMe
Data Fits In Memory Data Does Not Fit In Memory
Latency(us)
slab allocation (SSD write) cache check+load (SSD read) cache update server response client wait miss-penalty
51. Intel HPC Dev Conf (SC ‘16) 51Network Based Computing Laboratory
– Data does not fit in memory: Non-blocking Memcached Set/Get API Extensions can achieve
• >16x latency improvement vs. blocking API over RDMA-Hybrid/RDMA-Mem w/ penalty
• >2.5x throughput improvement vs. blocking API over default/optimized RDMA-Hybrid
– Data fits in memory: Non-blocking Extensions perform similar to RDMA-Mem/RDMA-Hybrid and >3.6x
improvement over IPoIB-Mem
Performance Evaluation with Non-Blocking Memcached API
0
500
1000
1500
2000
2500
Set Get Set Get Set Get Set Get Set Get Set Get
IPoIB-Mem RDMA-Mem H-RDMA-Def H-RDMA-Opt-Block H-RDMA-Opt-NonB-
i
H-RDMA-Opt-NonB-
b
AverageLatency(us)
MissPenalty(BackendDBAccessOverhead)
ClientWait
ServerResponse
CacheUpdate
Cachecheck+Load(Memoryand/orSSDread)
SlabAllocation(w/SSDwriteonOut-of-Mem)
H = Hybrid Memcached over SATA SSD Opt = Adaptive slab manager Block = Default Blocking API
NonB-i = Non-blocking iset/iget API NonB-b = Non-blocking bset/bget API w/ buffer re-use guarantee
52. Intel HPC Dev Conf (SC ‘16) 52Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
53. Intel HPC Dev Conf (SC ‘16) 53Network Based Computing Laboratory
• Locality and Storage type-aware data
access
• Enhanced data access strategies for
Hadoop and Spark for clusters with
heterogeneous storage characteristics
– different nodes have different types of
storage devices
• Re-design HDFS to incorporate the
proposed strategies
– Upper-level frameworks can
transparently leverage the benefits
Efficient Data Access Strategies with Heterogeneous Storage
Hadoop/Spark Workloads
Access
Strategy
Selector
Weight
Distributor
Locality
Detector
Storage Type
Fetcher
DataNode Selector
DataNode
SSD HDD
Storage Monitor
Connection
Tracker
N. Islam, M. W. Rahman, X. Lu, and D. K. Panda, Efficient Data Access Strategies for Hadoop and Spark on HPC Clusters with
Heterogeneous Storage, accepted at BigData ’16, December 2016
54. Intel HPC Dev Conf (SC ‘16) 54Network Based Computing Laboratory
Evaluation with Sort Benchmark
OSU RI2 EDR Cluster
• Hadoop
– 32% improvement for 320GB
data size on 16 nodes
• Spark
– 18% improvement for 320GB
data size on 16 nodes
Hadoop Sort Spark Sort
Reduced by 32%
Reduced by 18%
0
100
200
300
400
500
600
700
4:80 8:160 16:320
ExecutionTime(s)
Cluster Size: Data Size (GB)
HDFS OSU-Design
0
50
100
150
200
250
300
350
400
450
4:80 8:160 16:320
ExecutionTime(s)
Cluster Size: Data Size (GB)
55. Intel HPC Dev Conf (SC ‘16) 55Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
56. Intel HPC Dev Conf (SC ‘16) 56Network Based Computing Laboratory
• Hybrid and resilient key-value store-
based Burst-Buffer system Over Lustre
• Overcome limitations of local storage on
HPC cluster nodes
• Light-weight transparent interface to
Hadoop/Spark applications
• Accelerating I/O-intensive Big Data
workloads
– Non-blocking Memcached APIs to
maximize overlap
– Client-based replication for resilience
– Asynchronous persistence to Lustre
parallel file system
Burst-Buffer Over Lustre for Accelerating Big Data I/O (Boldio)
D. Shankar, X. Lu, D. K. Panda, Boldio: A Hybrid and Resilient Burst-Buffer over Lustre for Accelerating Big Data I/O, IEEE Big Data 2016.
DirectoverLustre
Hadoop Applications/Benchmarks (E.g. MapReduce, Spark)
Hadoop File System Class Abstraction (LocalFileSystem)
Burst-Buffer Memcached Cluster
Burst-Buffer Libmemcached Client
RDMA-enhanced Communication Engine
Non-Blocking API Blocking API
RDMA-enhanced
Comm. Engine
Hyb-Mem Manager
(RAM/SSD)
Persistence Mgr.
RDMA-enhanced
Comm. Engine
Hyb-Mem Manager
(RAM/SSD)
Persistence Mgr.
Boldio
…..
Co-Design
BoldioFileSystem Abs.
Lustre Parallel File System
MDS OSS OSS OSS
MDT MDT OST OST OST OST OST OST
BoldioServerBoldioClient
57. Intel HPC Dev Conf (SC ‘16) 57Network Based Computing Laboratory
• Based on RDMA-based Libmemcached/Memcached 0.9.3, Hadoop-2.6.0
• InfiniBand QDR, 24GB RAM + PCIe-SSDs, 12 nodes, 32/48 Map/Reduce Tasks, 4-node Memcached cluster
• Boldio can improve
– throughput over Lustre by about 3x for write throughput and 7x for read throughput
– execution time of Hadoop benchmarks over Lustre, e.g. Wordcount, Cloudburst by >21%
• Contrasting with Alluxio (formerly Tachyon)
– Performance degrades about 15x when Alluxio cannot leverage local storage (Alluxio-Local vs. Alluxio-Remote)
– Boldio can improve throughput over Alluxio with all remote workers by about 3.5x - 8 .8x (Alluxio-Remote vs. Boldio)
Performance Evaluation with Boldio
Hadoop/Spark Workloads
21%
0
50
100
150
200
250
300
350
400
450
WordCount InvIndx CloudBurst Spark TeraGen
Latency(sec)
Lustre-Direct Alluxio-Remote Boldio
DFSIO Throughput
0.00
10000.00
20000.00
30000.00
40000.00
50000.00
60000.00
70000.00
20 GB 40 GB 20 GB 40 GB
Write Read
Agg.Throughput(MBps)
Lustre-Direct Alluxio-Local
Alluxio-Remote Boldio
~3x
~7x
58. Intel HPC Dev Conf (SC ‘16) 58Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
59. Intel HPC Dev Conf (SC ‘16) 59Network Based Computing Laboratory
• Challenges
– Operations on Distributed Ordered Table (DOT)
with indexing techniques are network intensive
– Additional overhead of creating and maintaining
secondary indices
– Can RDMA benefit indexing techniques (Apache
Phoenix and CCIndex) on HBase?
• Results
– Evaluation with Apache Phoenix and CCIndex
– Up to 2x improvement in query throughput
– Up to 35% reduction in application workload
execution time
Collaboration with Institute of Computing Technology,
Chinese Academy of Sciences
Accelerating Indexing Techniques on HBase with RDMA
S. Gugnani, X. Lu, L. Zha, and D. K. Panda, Characterizing and Accelerating Indexing Techniques on HBase for Distributed Ordered
Table-based Queries and Applications (Under review)
0
5000
10000
15000
Query1 Query2
THROUGHPUT
TPC-H Query Benchmarks
HBase RDMA-HBase HBase-Phoenix
RDMA-HBase-Phoenix HBase-CCIndex RDMA-HBase-CCIndex
Increased by
2x
0
100
200
300
400
500
600
Workload1 Workload2 Workload3 Workload4
EXECUTIONTIME
Ad Master Application Workloads
HBase-Phoenix RDMA-HBase-Phoenix
HBase-CCIndex RDMA-HBase-CCIndex Reduced by
35%
60. Intel HPC Dev Conf (SC ‘16) 60Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
61. Intel HPC Dev Conf (SC ‘16) 61Network Based Computing Laboratory
Challenges of Tuning and Profiling
• MapReduce systems have different
configuration parameters based on the
underlying component that uses these
• The parameter files vary across
different MapReduce stacks
• Proposed a generalized parameter
space for HPC clusters
• Two broad dimensions: user space and
system space; existing parameters can
be categorized in the proposed spaces
62. Intel HPC Dev Conf (SC ‘16) 62Network Based Computing Laboratory
MR-Advisor Overview
• A generalized framework for
Big Data processing engines
to perform tuning, profiling,
and prediction
• Current framework can work
with Hadoop, Spark, and
RDMA MapReduce (OSU-MR)
• Can also provide tuning for
different file systems (e.g.
HDFS, Lustre, Tachyon),
resource managers (e.g.
YARN), and applications
63. Intel HPC Dev Conf (SC ‘16) 63Network Based Computing Laboratory
Execution Details in MR-Advisor
• Workload Generator maps the
input parameters to the
appropriate configuration and
generates workload for each
tuning test
• Job submitter deploys small
clusters for each experiment and
runs the workload on these
deployments
• Job Tuning Tracker monitors failed
jobs and re-launches, if necessary
Workload Preparation and Deployment Unit
Workload
Generator
Job Submitter
Job Tuning
Tracker
Compute Nodes
Lustre Installation
MDS OSS
InfiniBand/
Ethernet
App Master
Map/Reduce
NM DN
RM/Spark
Master
Tachyon
Master/ NN
Spark Worker
Tachyon
Worker
DN
App Master
Map/Reduce
NM LC
M. W. Rahman , N. S. Islam, X. Lu, D. Shankar, and D. K.
Panda, MR-Advisor: A Comprehensive Tuning Tool for
Advising HPC Users to Accelerate MapReduce Applications on
Supercomputers, SBAC-PAD, 2016.
64. Intel HPC Dev Conf (SC ‘16) 64Network Based Computing Laboratory
Tuning Experiments with MR-Advisor (TACC Stampede)
Apache MR over HDFS OSU-MR over HDFS Apache Spark over HDFS
Apache MR over Lustre OSU-MR over Lustre Apache Spark over Lustre
23% 17% 34%
46% 58% 28%
Performance improvements compared to current best practice values
65. Intel HPC Dev Conf (SC ‘16) 65Network Based Computing Laboratory
• Basic Designs
– HDFS, MapReduce, and RPC
– HBase
– Spark
– Memcached
– OSU HiBD Benchmarks (OHB)
• Advanced Designs
– HDFS+Memcached-based Burst-Buffer
– Memcached with Hybrid Memory and Non-blocking APIs
– Data Access Strategies with Heterogeneous Storage (Hadoop and Spark)
– Accelerating Big Data I/O (Lustre + Burst-Buffer)
– Efficient Indexing with RDMA-HBase
– MR-Advisor
• BigData + HPC Cloud
Acceleration Case Studies and Performance Evaluation
66. Intel HPC Dev Conf (SC ‘16) 66Network Based Computing Laboratory
• Motivation
– Performance attributes of Big Data workloads
when using SR-IOV are not known
– Impact of VM subscription policies, data size, and
type of workload on performance of workloads
with SR-IOV not evaluated in systematic manner
• Results
– Evaluation on Chameleon Cloud with RDMA-
Hadoop
– Only 0.3 – 13% overhead with SR-IOV compared
to native execution
– Best VM subscription policy depends on type of
workload
Performance Characterization of Hadoop Workloads on SR-
IOV-enabled Clouds
S. Gugnani, X. Lu, and D. K. Panda, Performance Characterization of Hadoop Workloads on SR-IOV-enabled Virtualized InfiniBand
Clusters, accepted at BDCAT’16, December 2016
0
200
400
40 GB 60 GB 40 GB 60 GB
EXECUTIONTIME
Hadoop Workloads
Native (1 DN, 1 NM) Native (2 DN, 2 NM)
VM per node VM per socket
TeraGen WordCount
0
500
1000
Sample Data 1 GB 60 GB 90 GB
EXECUTIONTIME
Applications
Native (1 DN, 1 NM) Native (2 DN, 2 NM)
VM per node VM per socket
CloudBurst Self-join
67. Intel HPC Dev Conf (SC ‘16) 67Network Based Computing Laboratory
• Challenges
– Existing designs in Hadoop not virtualization-
aware
– No support for automatic topology detection
• Design
– Automatic Topology Detection using
MapReduce-based utility
• Requires no user input
• Can detect topology changes during
runtime without affecting running jobs
– Virtualization and topology-aware
communication through map task scheduling and
YARN container allocation policy extensions
Virtualization-aware and Automatic Topology Detection
Schemes in Hadoop on InfiniBand
S. Gugnani, X. Lu, and D. K. Panda, Designing Virtualization-aware and Automatic Topology Detection Schemes for Accelerating
Hadoop on SR-IOV-enabled Clouds, CloudCom’16, December 2016
0
2000
4000
6000
40 GB 60 GB 40 GB 60 GB 40 GB 60 GB
EXECUTIONTIME
Hadoop Benchmarks
RDMA-Hadoop Hadoop-Virt
0
100
200
300
400
Default
Mode
Distributed
Mode
Default
Mode
Distributed
Mode
EXECUTIONTIME
Hadoop Applications
RDMA-Hadoop Hadoop-Virt
CloudBurst Self-join
Sort WordCount PageRank
Reduced by
55%
Reduced by
34%
68. Intel HPC Dev Conf (SC ‘16) 68Network Based Computing Laboratory
• Upcoming Releases of RDMA-enhanced Packages will support
– Upgrades to the latest versions of Hadoop and Spark
– Streaming
– MR-Advisor
– Impala
• Upcoming Releases of OSU HiBD Micro-Benchmarks (OHB) will support
– MapReduce, RPC and Spark
• Advanced designs with upper-level changes and optimizations
– Boldio
– Efficient Indexing
On-going and Future Plans of OSU High Performance Big Data
(HiBD) Project
69. Intel HPC Dev Conf (SC ‘16) 69Network Based Computing Laboratory
• Discussed challenges in accelerating Big Data middleware with HPC
technologies
• Presented basic and advanced designs to take advantage of InfiniBand/RDMA
for HDFS, MapReduce, RPC, HBase, Memcached, and Spark
• Results are promising
• Many other open issues need to be solved
• Will enable Big Data community to take advantage of modern HPC
technologies to carry out their analytics in a fast and scalable manner
• Looking forward to collaboration with the community
Concluding Remarks
70. Intel HPC Dev Conf (SC ‘16) 70Network Based Computing Laboratory
Funding Acknowledgments
Funding Support by
Equipment Support by
71. Intel HPC Dev Conf (SC ‘16) 71Network Based Computing Laboratory
Personnel Acknowledgments
Current Students
– A. Awan (Ph.D.)
– M. Bayatpour (Ph.D.)
– S. Chakraborthy (Ph.D.)
– C.-H. Chu (Ph.D.)
Past Students
– A. Augustine (M.S.)
– P. Balaji (Ph.D.)
– S. Bhagvat (M.S.)
– A. Bhat (M.S.)
– D. Buntinas (Ph.D.)
– L. Chai (Ph.D.)
– B. Chandrasekharan (M.S.)
– N. Dandapanthula (M.S.)
– V. Dhanraj (M.S.)
– T. Gangadharappa (M.S.)
– K. Gopalakrishnan (M.S.)
– R. Rajachandrasekar (Ph.D.)
– G. Santhanaraman (Ph.D.)
– A. Singh (Ph.D.)
– J. Sridhar (M.S.)
– S. Sur (Ph.D.)
– H. Subramoni (Ph.D.)
– K. Vaidyanathan (Ph.D.)
– A. Vishnu (Ph.D.)
– J. Wu (Ph.D.)
– W. Yu (Ph.D.)
Past Research Scientist
– S. Sur
Past Post-Docs
– D. Banerjee
– X. Besseron
– H.-W. Jin
– W. Huang (Ph.D.)
– W. Jiang (M.S.)
– J. Jose (Ph.D.)
– S. Kini (M.S.)
– M. Koop (Ph.D.)
– K. Kulkarni (M.S.)
– R. Kumar (M.S.)
– S. Krishnamoorthy (M.S.)
– K. Kandalla (Ph.D.)
– P. Lai (M.S.)
– J. Liu (Ph.D.)
– M. Luo (Ph.D.)
– A. Mamidala (Ph.D.)
– G. Marsh (M.S.)
– V. Meshram (M.S.)
– A. Moody (M.S.)
– S. Naravula (Ph.D.)
– R. Noronha (Ph.D.)
– X. Ouyang (Ph.D.)
– S. Pai (M.S.)
– S. Potluri (Ph.D.)
– S. Guganani (Ph.D.)
– J. Hashmi (Ph.D.)
– N. Islam (Ph.D.)
– M. Li (Ph.D.)
– J. Lin
– M. Luo
– E. Mancini
Current Research Scientists
– K. Hamidouche
– X. Lu
– H. Subramoni
Past Programmers
– D. Bureddy
– M. Arnold
– J. Perkins
Current Research Specialist
– J. Smith
– M. Rahman (Ph.D.)
– D. Shankar (Ph.D.)
– A. Venkatesh (Ph.D.)
– J. Zhang (Ph.D.)
– S. Marcarelli
– J. Vienne
– H. Wang
72. Intel HPC Dev Conf (SC ‘16) 72Network Based Computing Laboratory
The 3rd International Workshop on
High-Performance Big Data Computing (HPBDC)
HPBDC 2017 will be held with IEEE International Parallel and Distributed Processing
Symposium (IPDPS 2017), Orlando, Florida USA, May, 2017
Tentative Submission Deadline
Abstract: January 10, 2017
Full Submission: January 17, 2017
HPBDC 2016 was held in conjunction with IPDPS’16
Keynote Talk: Dr. Chaitanya Baru,
Senior Advisor for Data Science, National Science Foundation (NSF);
Distinguished Scientist, San Diego Supercomputer Center (SDSC)
Panel Moderator: Jianfeng Zhan (ICT/CAS)
Panel Topic: Merge or Split: Mutual Influence between Big Data and HPC Techniques
Six Regular Research Papers and Two Short Research Papers
http://web.cse.ohio-state.edu/~luxi/hpbdc2016
73. Intel HPC Dev Conf (SC ‘16) 73Network Based Computing Laboratory
• Three Conference Tutorials (IB+HSE, IB+HSE Advanced, Big Data)
• HP-CAST
• Technical Papers (SC main conference; Doctoral Showcase; Poster; PDSW-
DISC, PAW, COMHPC, and ESPM2 Workshops)
• Booth Presentations (Mellanox, NVIDIA, NRL, PGAS)
• HPC Connection Workshop
• Will be stationed at Ohio Supercomputer Center/OH-TECH Booth (#1107)
– Multiple presentations and demos
• More Details from http://mvapich.cse.ohio-state.edu/talks/
OSU Team Will be Participating in Multiple Events at SC ‘16
74. Intel HPC Dev Conf (SC ‘16) 74Network Based Computing Laboratory
{panda, luxi}@cse.ohio-state.edu
http://www.cse.ohio-state.edu/~panda
http://www.cse.ohio-state.edu/~luxi
Thank You!
Network-Based Computing Laboratory
http://nowlab.cse.ohio-state.edu/
The High-Performance Big Data Project
http://hibd.cse.ohio-state.edu/