internet2 - emerging opportunities in hpc cloud & co-location … · 2019-12-02 · emerging...
TRANSCRIPT
Emerging Opportunities in HPC Cloud & Co-location Services
at the University of Nevada - Las Vegas
Joseph Lombardo
Executive Director, UNLV National Supercomputing Institute
Jim Donovan
VP of Product, Wasabi Technologies, Inc.
2019 Technology Exchange in New Orleans, December 2019
Collaborations with the Cleveland Clinic Lou Ruvo Center for Brain Health and the Nevada Institute of Personalized Medicine are currently funded by 2
Institutional Development Awards (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health: #P20GM109025 &
#P20GM121325. The content is solely the responsibility of the author(s) and does not necessarily represent the official views of the National Institutes of Health.
Agenda
• About the National Supercomputing Institute (NSI)
• About Switch
• About Altair
• About Wasabi
About the NSI
Full-service supercomputing facility
Mission for excellence in education and research in
supercomputing and its applications
Provides supercomputing training and services to academic
and research institutions, government and private industry
Facilitates high-technology economic diversification in
Nevada by providing services not available in the private-
sector and by promoting partnerships between university
faculty and external
NSI @ Switch
• 2014 - UNLV moved its NSI
facilities to the Switch facility in
Las Vegas
• Hosted on Cherry Creek
system – large Intel system for
scientific and economic R&D
• > 30,000 compute cores
• Intel Xeon E5-2697v2 12C
2.700GHz, Intel Truscale, Intel
Xeon Phi 7120P
• Dedicated Research Network
(DMZ) with 200Gb/s potential
Numerous and complex workloads
• Hundreds of projects worldwide
• Highly compute-intensive research
Massive data needs
• Users must access massive data remotely to do their work
Time-sensitive projects
• Many NSI projects have critical governmental and environmental
significance, so timely and reliable performance is a key requirement
NSI Computing Challenges
New Data
Excel & Exome
Data(NeuroPsych data saved as
comma-separated values - CSV
SAM & BAM)
FreeSurfer(MRI data, which is converted by an
NSI application to CSV format)
Double Entry Lou Ruvo(Screening & Demographic Interviews –
Double Entry is supported by OpenClinica )
NSI(National Supercomputing Institute )
CSV Data(comma-separated values)
1. Reformat data + study meta-data(NSI’s multiple data parsing applications automates the
conversion of data from other applications (Excel) to a format
that OpenClinica’s data import application understands)
2. OpenClinica (OC) data import application(produces XML)
3. XML - Extensible Markup Language)
(this data gets imported into OC)
OpenClinica
OpenClinica Output(non-exhaustive)
XML
SPSS
CSV
Note: these formats can be
exported to programs that
store data in tables, such
as Microsoft Excel
CNTN
researcher
1. CNTN Defined Archive(XML, SPSS, CSV & Data Dictionary
files)
2. Subscriber Defined Download(same as CNTN Defined Archive with the exception that the
remote user selects specific fields in the record that they need
… CNTN will use the selected choices to determine which data
is most valuable to the remote research community)
and/or
• FreeSurfer Software Suite: An open source software suite for processing and analyzing (human) brain MRI images.
• Statistical Package for the Social Sciences (SPSS): SPSS is a software package used for logical batched and non-batched statistical analysis.
• Operational Data Model (ODM)-XML: a vendor-neutral, platform-independent format for exchanging and archiving clinical and translational research data, along with their associated metadata,
administrative data, reference data, and audit information.
• ODM clinical data extraction application: NSI’s software that produces an extract of clinical data from an ODM file produced by OpenClinica and then writes MRI thickness data for left and right
hemispheres to a new file.
• NSI application: Software created by the National Supercomputing Institute (NSI) in support of CNTN.
• Secure Sockets Layer (SSL): is a protocol for encrypting data transferred between two computers.
Data & Software Archives(freestanding files for download – no access to the
OpenClinca environment is allowed)
NIH: Data Management
and Statistics Core(s) Data Flow Summary
Prepared by Joseph Lombardo
Exe. Director, National Supercomputing
Institute (NSI)
CNTN & CEPM Data Core PI
(NSI’s ODM clinical data extraction
application is applied here to
create the archive files )
3. Software Archive
Shared Archive
Firewall SSL Encryption Firewall SSL EncryptionFirewall SSL Encryption
Firewall SSL Encryption
Remote Researcher
Firewall SSL Encryption
Excel
(CEPM)
In collaboration with the Cleveland Clinic Lou Ruvo Center for Brain Health and the Nevada Institute of Personalized Medicine. Funded by two Institutional Development Awards (IDeA) from the National Institute of General Medical
Sciences of the National Institutes of Health: #P20GM109025 & #P20GM121325. The content is solely the responsibility of the author(s) and does not necessarily represent the official views of the National Institutes of Health.
XML-DD SPSS-DD CSV-DD
XML-DD SPSS-DD CSV-DD
XML-DD SPSS-DD CSV-DD
. . .
. . .
. . .
XML-DD SPSS-DD CSV-DD
Cloud & Co-location
8
Innovation Intelligence®
UNLV & Altair Collaboration
The University of Nevada, Las Vegas (UNLV) is home to the
“Cherry Creek II” supercomputer, housed in Switch’s Las Vegas
SUPERNAP data center. The system is among the fastest and
most powerful in the world. It gives scientists around the globe
access to significant high-performance computing power.
9
Innovation Intelligence®
UNLV & Altair Collaboration
All that computing power doesn’t orchestrate itself — so UNLV
enlisted Altair to deploy the Altair PBS Works™ high-
performance computing (HPC) management suite to securely
manage Cherry Creek II’s compute workload, simplifying
access to and utilization of the supercomputer’s capabilities and
capacity. Users can easily create, access, and manage physical
and virtual appliances on Cherry Creek II and run Altair’s
HyperWorks® simulation software as well as third-party
applications.
10
Innovation Intelligence®
UNLV & Altair Collaboration
The collaboration “sets the stage for UNLV to become an even
greater supercomputing powerhouse for the Southern Nevada
community.” — Joseph Lombardo, Executive Director of the
UNLV National Supercomputing Institute
11
The industry-leading Altair PBS Works™ workload management solution includes all the tools
you need to schedule, tune, and accelerate your jobs, including:
• Altair Access™ – a simple, powerful, and consistent interface for submitting and monitoring
jobs on remote clusters, clouds, and other resources, allowing engineers and researchers to
focus on core activities.
• Altair Control™ – an easy-to-use web application for monitoring and managing jobs and
nodes in an HPC environment.
• Altair SAO – an advanced tool for software asset optimization, built so you can right-size your
organization’s software portfolio using real data to make informed business decisions.
Altair SolutionsAltair Solutions
• Altair PBS Professional™ – a fast, powerful workload manager designed to improve
productivity, optimize utilization and efficiency, and simplify administration for HPC clusters,
clouds, and supercomputers.
12
Innovation Intelligence®
• Powerful, flexible customization capabilities -- can be
easily extended by adding site-specific processing
plugins/hooks
• Improved system manageability and extensibility:
• Lightweight solution
• Very easy to manage
• Not dependent on any specific operating system
13
Wasabi Cloud Storage For
Emerging Opportunities in HPC Cloud & Co-location Services
12 December 2019
14
Wasabi Introduction
• Mission: Low Price, High Performance, Secure Object Storage
• Started in 2015 by Carbonite’s founding team (David Friend & Jeff Flowers)
• Privately held & well funded with $80M invested to date
• Available via Internet2 Cloud Exchange since 2018
• Product: Cloud Object Storage as a Service
• Comparable to: AWS S3, Microsoft Azure Blob Storage,
and Google Cloud Platform (GCP) Storage
• Thousands of customers & partners across all verticals
15
Wasabi’s Value PropositionOptimal price, performance, and protection for object storage
16
PriceLower cost than all other major object storage providers (more info @ wasabi.com/pricing)
• Wasabi’s flat fee of $.0059/GB/mo ($5.99/TB/mo) for storage is a disruptor relative to competitors
• No charge for egress (downloads)(vs. up to $.09/GB with AWS S3)
• No charge for API requests (unlike all other public cloud storage providers)
• No complex storage tiers (Wasabi is hot storage at cold storage prices)
Storage Costs For 1 PB of storage
With 20% Data Egress Per Year
17
PerformanceBuilt-for-speed file system
• Purpose-built file system leveraging hardware technology• Enables significant cost reduction & performance improvements
• Faster than AWS S3 & meaningful time-to-first-byte (TTFB) advantages
• Highly distributed architecture providing exabyte-scale storage
Sample test results for write performance with
1 MB objects across different compute thread counts
Wasabi-built software deployed on
leading-edge hardware in top-tier data centers
High performance
enabled by Wasabi’s
system architecture
100 GbE
Switching
User Servers
Database Servers Storage ServersCompute Threads
18
ProtectionBuilt for scale, durability, security and compliance
• Durability & Availability• 11 x 9s data durability with exabyte scale
• 99.99% availability SLA with multiple data centers & Wasabi Bucket Replication
• Data integrity checks at time of upload, download, and every 90 days
• Security• All data encrypted in transit and at rest
• Immutable buckets prevent accidental deletion/modification
• Strong identity & access management & multi-factor authentication
• Compliance with industry privacy, security, and data center standards
19
Multiple Public + Private Interconnect OptionsEnables high-speed exchange between Wasabi storage & customer compute resources
• Public internet (N x 10 Gb/s)
• Wasabi or AWS/Azure/GCP ‘Direct Connect’ (N x 1 or 10 Gb/s)
• Wasabi Ball Transfer Appliance (up to 100 TB per appliance)
Public Cloud or Private Data Center
Compute Wasabi us-west-1 region
(Oregon)
Wasabi eu-central-1 region(Amsterdam)
Wasabi us-east-1 &
us-east-2 regions(Northern Virginia)
Wasabi apac-1 region (Tokyo – Dec 2019)
Wasabi BallTransfer
Appliance
Key networking, compute & data center partners include:
20
AWS S3 CompatibilityWill my existing AWS S3 applications work with Wasabi?
• Wasabi fully supports the AWS S3 & IAM APIs• Wasabi looks just like an Amazon S3 implementation
• Same AWS API constructs for storage & identity management
• No need to change apps you may be currently using with AWS S3
• Any 3rd-party AWS S3-compatible app or platform should work with Wasabi
• 200+ apps listed at wasabi.com/interop
AWS S3 & IAM APIs
=
Interop Categories Include:
Backup &
Recovery
Archiving
Content
Delivery
AnalyticsApp Dev
ToolsIoT
Storage
Gateway
21
Wasabi Management ConsoleCommon look-and-feel with AWS Management Console
• Wasabi’s storage and identity access management console is modeled on AWS S3 (to make it simpler for new users to adopt)
• Same concepts of storage buckets, access keys, users, policies etc.
• Demo video @ wasabi.com/help
Wasabi UI
= same
look & feel
as AWS S3
mgmt
console
22
Cloud Strategies For Leveraging Wasabi (HPC and more…)More info @ wasabi.com/solutions
On-Prem
to Cloud
Move all on-prem
storage to the
cloud and
eliminate
maintenance fees
Hybrid
Storage
Extend on-prem or
private cloud
investments with
affordable public
cloud storage
Eliminate cloud
lock-in & choose
best-of-breed
providers for price
and performance
Multi-Cloud Data Lake
Eliminate tiers and
store everything in
active archives for
your data analytics
projects
High
Performance
& Edge
ComputingOptimize system
performance &
costs with cloud
storage
Migrate your tape
archive/backup to
leverage
affordable public
cloud storage
Tape to
Cloud
23
Use Cases For Leveraging WasabiMore info @ wasabi.com/solutions
Internet of ThingsStore more data coming from
billions of smart connected
sensors and devices
SurveillanceStore more video and photos to
improve security and law
enforcement
Data Analytics
Store more data to enable data
analytics and better business
intelligence
Backup and Recovery
Store more to enable business
continuity across hybrid or
multiple clouds
ArchivingStore more and evolve complex
archive tiers into a single simple
active archive
Store more to enable custom
app development integrated with
compute and cloud partners
Application
Development
Store more data to enable
artificial intelligence and machine
learning to transform the future
AI/MLStore more to accelerate audio
and video content and software
distribution
Content Delivery
24
Thank YouFor more information, please visit wasabi.com
Questions?
Thank you for your attention!
Joseph Lombardo
UNLV National Supercomputing Institute
Jim Donovan
Wasabi Technologies, Inc.
Victor Wright
Altair Engineering, Inc.
2019 Technology Exchange in New Orleans, December 2019
Collaborations with the Cleveland Clinic Lou Ruvo Center for Brain Health and the Nevada Institute of Personalized Medicine are currently funded by two
Institutional Development Awards (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health: #P20GM109025 &
#P20GM121325. The content is solely the responsibility of the author(s) and does not necessarily represent the official views of the National Institutes of Health.