high performance computing with aws
DESCRIPTION
More and more, the scalable on-demand infrastructure provided by AWS is being used by researchers, scientists and engineers in Life Sciences, Finance and Engineering to solve bigger problems, answer complex questions and run larger simulations. In this session we start by talking about the supercomputing class performance and high performance storage available to the scientists and engineers at their fingertips. We will go over examples of how startups are innovating and large enterprises are extending their HPC environments. Finally, we walk through some of the common questions that come up as organizations start leveraging AWS for their high performance computing needs.TRANSCRIPT
Jafar Shameem and David Pellerin
High Performance Computing with AWS
Business Development, HPC
Migrate entire HPC applications
and datacenters to the cloud
Use cloud capabilities to create
entirely new HPC applications
Augment on-premise HPC
resources with cloud capacity
How are Organizations Using Cloud for HPC?
• Security: Deploy applications and store data in a secure,
highly configurable VPC environment
• Agility: Deploy the right infrastructure for each technical
computing job, at the right time
• Scalability: Add and subtract servers in minutes to
optimize time-to-results
• Cost Savings: Pay only for what you use, don’t pay for
idle or outdated servers
Why AWS for High-Performance Computing?
Waste
User/Customer
Dissatisfaction
Actual demand
Predicted Demand
Rigid On-Premise Resources Elastic Resources
Actual demand
Resources scaled to demand
AWS for Agility
On-Demand
Pay for compute
capacity by the hour
with no long-term
commitments
For spiky workloads,
or to define needs
Many purchase models to support different needs
Reserved
Make a low, one-time
payment and receive a
significant discount on
the hourly charge
For committed
utilization
Spot
Bid for unused capacity,
charged at a Spot Price
which fluctuates based
on supply and demand
For time-insensitive or
transient workloads
Dedicated
Launch instances within
Amazon VPC that run
on hardware dedicated
to a single customer
For highly sensitive or
compliance related
workloads
Free Tier
Get Started on AWS
with free usage & no
commitment
For POCs and
getting started
Massive scale allows AWS to constantly reduce
costs, while improving quality and reliability
TCO of cloud is much lower then on-premise IT
when all costs are considered
Result? Large scale datacenter-to-cloud
migrations are in progress every day
AWS for Scale
Scalable Computing: Go From Just One Instance…
To Thousands… in Just Minutes!
Mem
ory
(GiB
)
Small 1.7 GB,
1 EC2 Compute Unit
1 virtual core
Micro 613 MB
Up to 2 ECUs
Large 7.5 GB
4 EC2 Compute Units
2 virtual cores
Extra Large 15 GB
8 EC2 Compute Units
4 virtual cores
Hi-Mem XL 17.1 GB
6.5 EC2 Compute Units
2 virtual cores
Hi-Mem 2XL 34.2 GB
13 EC2 Compute Units
4 virtual cores
Hi-Mem 4XL 68.4 GB
26 EC2 Compute Units
8 virtual cores
High-CPU Med 1.7 GB
5 EC2 Compute Units
2 virtual cores
High-CPU XL 7 GB
20 EC2 Compute Units
8 virtual cores
Cluster GPU 4XL 22 GB
33.5 EC2 Compute Units,
2 x NVIDIA Tesla “Fermi”
M2050 GPUs
Cluster Compute 4XL 23 GB
33.5 EC2 Compute Units
Medium 3.7 GB,
2 EC2 Compute Units
1 virtual core
High I/O 4XL 60.5 GB, 35
EC2 Compute Units,
2*1024 GB SSD-based
local instance storage
High Storage 8XL 117 GB
35 EC2 Compute Units
24 * 2 TB instance store
Cluster High Mem 8XL
89 EC2 Compute Units
244 GB SSD instance storage
EC2 Compute Units
Cluster Compute 8XL 60.5 GB
88 EC2 Compute Units
Choose the Right Instance Type for the Job
On-Premise
Experiment
infrequently
Failure is
expensive
Less Innovation
Cloud
Experiment
often
Fail quickly at a
low cost
More Innovation
$ Millions Nearly $0
AWS for Innovation
Focus on innovation
Leave the muck of infrastructure management to AWS
http://eddie.niese.net/20090313/dont-pity-incompetence/
• Engineering: CAD and CAE for aerospace, defense, structures,
consumer products
• Life Sciences: For basic research, drug discovery, genomics, and
translational medicine
• Energy and Geophysics: Including seismic processing, reservoir
estimation, high-energy simulation, wind energy modeling, GIS
• Financial Services and Insurance: Including valuation and risk
analytics
And Many More!
HPC Applications Running on AWS Today
HPC for Engineering
Scalable Computing for CAD/CAE/EDA
AWS for Engineering
• Computer-Aided Design, Simulation, Analysis, Visualization
– For development of commercial and military products
– Aerospace, automotive, civil, construction, energy, others
– Across industries, the trend is Simulation-Driven Design
• Examples
– Computer-Aided Design (CAD) including 3D models
– Electronic Design Automation (EDA)
– Computational Fluid Dynamics (CFD)
– Finite Element Analysis (FEA) and Thermal Analysis
– Crash Analysis
– Failure and Hazard Analysis
CFD for Turbine Engine Design
• Time accurate fluid dynamics
• SBIR-funded project for the US Air Force Research Laboratory (AFRL)
• SAS 70 Type II certification and VPN-level access required
• Additional security measures:
– Uploaded and downloaded data was encrypted
– Dedicated EC2 cluster instances were provisioned
– Data was purged upon completion of the run
“The results of this case were impressive. Using Amazon EC2 the large-scale, time accurate simulation was turned around in just 72 hours with computing infrastructure costs well below $1,000.”
http://aws.amazon.com/solutions/case-studies/aerodynamic-solutions/
• Commercial provider of mixed-signal ASICs for X-ray and gamma ray
detection and imaging
• Needs to perform very large Monte Carlo simulations using as many as 4000
server nodes
• Computing workloads are highly variable, project-driven
• Building an on-premise cluster to handle peak loads would be cost prohibitive
• Solution: EC2 3rd-generation High-Memory instances
• Up to 80% savings by using Spot instances on EC2
Radiation Simulation for ASIC Design
1) Customer Managed Application Hosting
• Customer has account with AWS and manages infrastructure
• Customer maintains traditional software vendor relationships
• Software vendor offers license flexibility (BYOL)
2) Vendor Managed Hosting to Augment On-Premise Application
• Client-Server model for acceleration of batch tasks
• Customer pays software vendor for AWS-hosted services
• Customer does not need to manage low-level infrastructure
3) Vendor Managed Software-as-a-Service
• Pay-per-use, fully web-based including GUI
Scenarios for Technical Software
Trusted by Enterprises Worldwide
HPC for Life Sciences
Customer Case-Studies
And a rich history in Life Sciences
AWS Public Data Sets • A centralized repository of public datasets
• Seamless integration with cloud based applications
• No charge to the community
• Some of the datasets available today:
– 1000 Genomes Project
– Ensembl
– GenBank
– Illumina – Jay Flateley Human Genome Dataset
– YRI Trio Dataset
– The Cannabis Sativa Genome
– UniGene
– Influenza Virrus
– PubChem
• Tell us what else you’d like for us to host …
Open Source ecosystem
• NCBI BLAST
• Crossbow
• CloudBurst
• Myrna
• Clovr
• BioPerl Max
• VIPDAC
• Superfamily
• Cloud-Coffee
• BioNimbus
• GMOD
• CloudAligner
• CRdata
• SeqWare
• Blend
• StormSeq
• BioConductor
Get links to AMIs at: https://github.com/mndoci/mndoci.github.com/wiki/Life-Science-Apps-on-AWS
MIT StarCluster Sun Grid Engine Condor
Torque Slurm Rocks
Chef Puppet
Number of Cluster nodes can be added depends on the computational
needs
Remove constraints Capex, operational skills,
processing limitations
Focus on the problem Not the technical challenges
of large compute clusters
Achieve more Perform bigger, more
complex jobs in a much
reduced time
Iterate around the
problem Do more and afford to take more
risks as cost of experimentation
reduced
Why
AWS?
Data Transfer
• AWS Import/Export
– Move large amounts of data into and outside AWS
– Data Migration, Content Distribution, DR, etc.
• AWS Direct Connect
– Secure private link to AWS
– 1Gbps, 10Gbps connectivity
– You can also co-locate hardware in AWS DX locations
• Bandwidth Optimization Solutions
– Commercial providers – Aspera, Riverbed, Attunity, etc.
– Open Source – Tsunami UDP, Globus Online
AWS Direct Connect
AWS Import/Export
Relational Database Service
Fully managed database
(MySQL, Oracle, MSSQL)
DynamoDB
NoSQL, Schemaless,
Provisioned throughput
database
S3
Object datastore up to 5TB
per object
99.999999999% durability
SimpleDB
NoSQL, Schemaless
Smaller datasets
Redshift
Petabyte scale
data warehousing service
Fully managed
Storage Options
1.3 Trillion
835k peak transactions per second
Objects in S3
Glacier
Long term cold storage
From $0.01 per GB/Month
99.999999999% durability
Archival
“Every day our genome sequencers produce terabytes of data. As our company moves into the clinical space, we face a legal requirement to archive patient data for years that would drastically raise the cost of storage. Thanks to Amazon Glacier’s secure and scalable solution, we will be able to provide cost-effective, long-term storage and thereby eliminate a barrier to providing whole genome sequencing for medical treatment of cancer and other genetic diseases.” - Keith Raffel, Senior Vice President and Chief Commercial Officer, Complete Genomics
Elastic MapReduce Managed, elastic Hadoop cluster
Integrates with S3 & DynamoDB
Leverage Hive & Pig analytics scripts
Integrates with instance types such as spot
Application Services
Feature Details
Scalable Use as many or as few compute instances running Hadoop as you want. Modify the number of instances while your job flow is running
Integrated with other services
Works seamlessly with S3 as origin and output. Integrates with DynamoDB
Comprehensive Supports languages such as Hive and Pig for defining analytics, and allows complex definitions in Cascading, Java, Ruby, Perl, Python, PHP, R, or C++
Cost effective Works with Spot instance types
Monitoring Monitor job flows from with the management console
Compute Storage
AWS Global Infrastructure
Database
App Services
Deployment & Administration
Networking
EMR Jobs
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
3,000,000
3,500,000
4,000,000
3.7 M clusters launched since May 2010
Crossbow
• Align billions of reads and find SNPs
– Reuse software components: Hadoop Streaming
h" p://bowI eAbio.sourceforge.net/crossbow2
• Map: Bowtie (Langmead et al., 2009)
– Find best alignment for each read
– Emit (chromosome region, alignment)
• Reduce: SOAPsnp (Li et al., 2009)
– Scan alignments for divergent columns
– Accounts for sequencing error, known SNPs
• Shuffle: Hadoop
– Group and sort alignments by region
…2
…2
Searching for SNPs with Cloud Computing.
Langmead B, Schatz MC, Lin J, Pop M, Salzberg SL (2009) Genome Biology. 10:R134
Worldwide research and
development
The Amazon Virtual Private Cloud was a unique option that offered an additional level of security and
an ability to integrate with other aspects of our infrastructure.
“AWS enables Pfizer’s WRD to explore specific difficult or deep
scientific questions in a timely, scalable manner and helps
Pfizer make better decisions more quickly” Dr. Michael Miller, Head of HPC for R&D, Pfizer
Spiral Genetics
• Alignment, Variant Calling, Annotation
• Turnaround time – Targeted : less than 40 minutes
– Exome : less than 2 hours
– Whole Genome : less than 5 hours
• Workflows can be easily defined
and automated with integrated Galaxy Platform capabilities
• Data movement is streamlined with integrated Globus file-transfer functionality
• Resources can be provisioned on-demand with Amazon Web Services cloud based infrastructure
Globus Genomics
Proprietary and Confidential. ©2013 Syapse
Syapse: Bringing Omics in Routine Medical Use
Laboratory Testing
Test Results Clinical Use
Syapse Semantic Data Platform
Syapse Omics Medical Record Application
Syapse Physician Portal Application
Syapse Discovery Application
Syapse
Leverage Spot instances in workflows 1 days worth of effort
resulted in 50% savings in cost
Harvard Medical School The Laboratory of Personal Medicine
Run EC2 clusters to analyze entire
genomes “The AWS solution is stable, robust, flexible, and low cost. It
has everything to recommend it.” Dr. Peter Tonellato, LPM, Center for Biomedical Informatics, Harvard Medical School
Illumina BaseSpace
• Data Analysis
– Alignment, Assembly, QC, Analysis
• Share data with colleagues
• Access high quality and diverse datasets
We are here to help
Enterprise support Trusted Advisor Professional Services Sales and
Solutions Architects