Accelerating Time to Science:

Transforming Research in the Cloud

Jamie Kinney - @jamiekinney

Director of Scientific Computing, a.k.a. “SciCo” – Amazon Web Services
 Why Do Researchers Love AWS?

Time to Science Low Cost Elastic

Access research Pay-as-you-go Easily add or remove
infrastructure in pricing capacity
minutes

Globally Accessible Secure Scalable

Easily Collaborate with A collection of tools to Access to effectively
researchers around the world protect data and limitless capacity
privacy
Why does Amazon care about Scientific
Computing?
• In order to fundamentally accelerate the pace of scientific
discovery

• It is a great application of AWS with a broad customer base

• The scientific community helps us innovate on behalf of all

customers
– Streaming data processing & analytics
– Exabyte scale data management solutions and exaflop scale compute
– Collaborative research tools and techniques
– New AWS regions
– Significant advances in low-power compute, storage and data centers
– Identify efficiencies which will lower our costs and therefore
reduce pricing for all AWS customers
How is AWS Used for Scientific Computing?

• High Throughput Computing (HTC) for Data-Intensive

Analytics
• High Performance Computing (HPC) for Engineering
and Simulation
• Collaborative Research Environments
• Hybrid Supercomputing Centers
• Science-as-a-Service
• Citizen Science
Research Grants

AWS provides free usage

credits to help researchers:
• Teach advanced courses
• Explore new projects
• Create resources for the
scientific community

aws.amazon.com/grants
Amazon Public Data Sets
Public Data Sets
AWS hosts “gold standard” reference
data at our expense in order to
catalyze rapid innovation and
increased AWS adoption
A few examples:
1,000 Genomes ~250 TB
Common Crawl
OpenStreetMap
Actively Developing…
Cancer Genomics Data Sets ~2-6 PB
SKA Precursor Data 1PB+
 Nepal
Earthquake

Individuals
around the
world are
analyzing
before/after
imagery of
Kathmandu
in order to
more-
effectively
direct
emergency
response and
recovery
efforts
Peering with all global research networks

Image courtesy John Hover - Brookhaven National Lab

AWS Egress Waiver for Research & Education
Timeline:
• 2013: Initial trial in Australia for users connecting via AARNET and AAPT
• 2014: Extended the waiver to include ESNET and Internet2
• 2015: Extending support to other major NRENs

Terms:
• AWS waives egress fees up to 15% of total AWS bill, customers are responsible for anything
above this amount
• Majority of traffic must transit via NREN with no transit costs
• 15% waiver applies to aggregate usage when consolidated billing is used
• Does not apply to workloads for which egress is the service we are providing (e.g. live video
streaming, MOOCs, Web Hosting, etc…)
• Available regardless of AWS procurement method (i.e. direct purchase or Internet2 Net+)

Contact us if you would like to sign up!

Breaking news! Restricted-access genomics on
AWS

aws.amazon.com/genomics
Data-Intensive Computing
The Square Kilometer Array will link 250,000 radio
telescopes together, creating the world’s most
sensitive telescope. The SKA will generate zettabytes
of raw data, publishing exabytes annually over 30-40
years.

Researchers are using AWS to develop and test:

• Data processing pipelines
• Image visualization tools
• Exabyte-scale research data management
• Collaborative research environments
www.skatelescope.org/ska-aws-astrocompute-call-for-
Astrocompute in the Cloud Program
• AWS is adding 1PB of SKA pre-cursor data to the
Amazon Public Data Sets program
• We are also providing $500K in AWS Research
Grants for the SKA to direct towards projects
focused on:
– High-throughput data analysis
– Image analysis algorithms
– Data mining discoveries (i.e. ML, CV and data
compression)
– Exascale data management techniques
– Collaborative research enablement
https://www.skatelescope.org/ska-aws-astrocompute-call-
for-proposals/
 Schrodinger & Cycle Computing:
Computational Chemistry for Better Solar Power
Simulation by Mark Thompson
of the University of Southern
California to see which of Loosely
205,000 organic compounds Coupled

could be used for photovoltaic

cells for solar panel material. • 156,314 core cluster, 8
regions
Estimated computation time 264 years
completed in 18 hours. • 1.21 petaFLOPS (Rpeak)
• $33,000 or 16¢ per molecule
Some Core AWS Concepts
Region

Trans
AZ
it

• Geographic area
where AWS services
AZ AZ AZ are available
• Customers choose
region(s) for their
AWS resources
Trans • Eleven regions
AZ
it worldwide
Availability Zone (AZ)

• Each region has REGIO

multiple, isolated N
EC
locations known as 2

Availability Zones
• Low-latency links AVAILABILITY
ZONE 1

between AZs in a region

<2ms, usually <1ms EC
EC2
• When launching an EC2 EC2
2

instance, a customer AVAILABILITY AVAILABILITY

chooses an AZ ZONE 2 ZONE 3

• Private AWS fiber links

interconnect all major
Example AWS Availability Zone
Trans
AZ it

AZ AZ AZ

Trans
AZ it
Data Center Data Center

• 1 of 28 AZs world-wide
• All regions have 2 or more
Data Center Data Center AZs
• Each AZ is 1 or more DC
– No data center is in two AZs
– Some AZs have as many as 6
DCs
• DCs in AZ less than ¼ ms
Example AWS Data Center

• Single DC typically over

50,000 servers & often over
80,000
– Larger DCs undesirable (blast
radius)
• Up to 102Tbps provisioned
to a single DC
• AWS custom network
equipment:
– Multi-ODM sourced
Virtual Private Cloud (VPC)

VPC REGIO
• Logically isolated
EC
N
section of the AWS 2
cloud, virtual AVAILABILITY
ZONE 1
network defined by
the customer
• When launching AVAILABILITY
ZONE 2
instances and other EC
EC2
resources, customers EC2
2
place them in a VPC
AVAILABILITY
• All new customers ZONE 3

have a default VPC

Spot Fleet
What is Spot?

• Name your own price for EC2 Compute

– A market where price of compute changes based upon Supply and Demand
– When Bid Price exceeds Spot Market Price, instance is launched
– Instance is terminated (with 2 minute warning) if market price exceeds bid price
• Where does capacity come from?
– Unused EC2 Instances
m1.small
On Demand
Price:
$0.044/hr
c3.8xlarge
On Demand
Price:
$1.68/hr
cc2.8xlarg
e
On Demand
Price:
$2.00/hr
 Spot allows customers to run
workloads that they would likely
not run anywhere else..

But today, to be successful in

Spot requires a little bit of
additional effort
The Spot Experience today

• Build stateless, distributed, scalable applications

• Choose which instance types fit your workload the best
• Ingest price feed data for AZs and regions
• Make run time decisions on which Spot pools to launch in based on price and volatility
• Manage interruptions
• Monitor and manage market prices across Azs and instance types
• Manage the capacity footprint in the fleet

UNDIFFERENTIATED
• And all of this while you don’t know where the capacity is
• Serve your customers

HEAVY LIFTING
Making Spot Fleet Requests
• Simply specify:
– Target Capacity – The number of EC2 instances that you want in your fleet.
– Maximum Bid Price – The maximum bid price that you are willing to pay.
– Launch Specifications – # of and types of instances, AMI id, VPC, subnets or AZs,
etc.
– IAM Fleet Role – The name of an IAM role. It must allow EC2 to terminate instances on
your behalf.
Spot Fleet

• Will attempt to reach the desired target capacity given the choices that were given
• Manage the capacity even as Spot prices change
• Launch using launch specifications provided
Using Spot Fleet
• Create EC2 Spot Fleet IAM Role
• Requesting a fleet:
– aws ec2 request-spot-fleet --spot-fleet-request-config file://mySmallFleet.json
• Describe fleet:
– aws ec2 describe-spot-fleet-requests
– aws ec2 describe-spot-fleet-requests --spot-fleet-request-ids <sfr-………..>
• Describe instances within the fleet
– aws ec2 describe-spot-fleet-instances --spot-fleet-request-id <sfr-…………>
• Cancel Spot Fleet (with termination):
– aws ec2 cancel-spot-fleet-requests --spot-fleet-request-ids <sfr-…………..> -terminate-
instances
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/spot-fleet.html
mySpotFleet.json
{
    "SpotPrice": "0.50",
    "TargetCapacity": 20,
    "IamFleetRole": "arn:aws:iam::123456789012:role/my­spot­fleet­role",
    "LaunchSpecifications": [
        {
            "ImageId": "ami­1a2b3c4d",
            "InstanceType": "cc2.8xlarge",
            "SubnetId": "subnet­a61dafcf"
        },
        {
            "ImageId": "ami­1a2b3c4d",
            "InstanceType": "r3.8xlarge",
            "SubnetId": "subnet­a61dafcf"
        }
    ]
}
Elastic File System
The AWS storage portfolio

Amazon • Object storage: data presented as buckets of

S3 objects
Amazon • Data access via APIs over the Internet
Elastic • Block storage (analogous to SAN): data presented as
Block disk volumes
Store • Lowest-latency access from single Amazon EC2
Amazon • instances
Archival storage: data presented as vaults/archives
Glacier of objects
• Lowest-cost storage, infrequent access via APIs
Amazon • File storage
over (analogous to NAS): data presented as a
the Internet
EFS file system
• Shared low-latency access from multiple EC2
instances
Amazon Elastic File System

• Fully managed file system for EC2 instances

• Provides standard file system semantics
• Works with standard operating system APIs
• Sharable across thousands of instances
• Elastically grows to petabyte scale
• Delivers performance for a wide variety of
workloads
• Highly available and durable
• NFS v4–based
EFS is designed for a broad range of use cases,
such as…

• Content repositories
• Development environments
• Home directories
• Big data
Amazon Elastic Container Service

+
Key Components

Docker Daemon
Task Definitions
Containers
Clusters
Container Instances
Typical User Workflow

I have a Docker
image, and I want
to run the image
on a cluster
Typical User Workflow

Push
Image(s)
Typical User Workflow

Declare
resource
requirements

Create Task
Amazon
Definition ECS
Typical User Workflow
Use custom AMI
with Docker
support and ECS
Agent. Instances
will register with
default cluster.

Run EC2
Instances
Typical User Workflow

Get information
about cluster state
and available
resources

Describe
Amazon
Cluster ECS
Typical User Workflow

Using the task

definition created
above

Run
Amazon
Task ECS
Typical User Workflow

Get information
about cluster state
and running
containers

Describe
Amazon
Cluster ECS
Thank you!
Jamie Kinney
jkinney@amazon.com
@jamiekinney
 Additional resources…

• aws.amazon.com/big-data
• aws.amazon.com/compliance
• aws.amazon.com/datasets
• aws.amazon.com/grants
• aws.amazon.com/genomics
• aws.amazon.com/hpc
• aws.amazon.com/security