three data delivery cases for embl- ebi’s...
TRANSCRIPT
Three data delivery cases for EMBL-EBI’s Embassy
Guy Cochrane
www.ebi.ac.uk
EMBL European Bioinformatics Institute
Genes, genomes & variation • European Nucleotide Archive • 1000 Genomes • Ensembl • Ensembl Genomes • Ensembl Plants • European Genome-phenome Archive • Metagenomics portal • GWAS Catalog browser
Expression • ArrayExpress • Expression Atlas • Metabolights • PRIDE
Protein sequences • InterPro • Pfam • UniProt
Chemical biology • ChEMBL • ChEBI
Literature & ontology • Europe PubMed Central • Gene Ontology • Experimental Factor
Ontology
Molecular structures • Protein Data Bank in Europe • Electron Microscopy Data Bank
Reactions, interactions & pathways • IntAct • Reactome • MetaboLights
Systems • BioModels • Enzyme Portal • BioSamples
Read
Alignment
Annotation
Sample/method
Assembly
European Nucleotide Archive - Unrestricted data - Pan-species and application - http://www.ebi.ac.uk/ena/
Sequence data at EMBL-EBI
Read
Alignment
Sample/method
European Genome-phenome Archive - Controlled access data - Human data around molecular medicine - http://www.ebi.ac.uk/ega/
Read
Alignment
Annotation
Sample/method
Assembly
European Nucleotide Archive - Unrestricted data - Pan-species and application - http://www.ebi.ac.uk/ena/
Sequence data at EMBL-EBI
Read
Alignment
Sample/method
European Genome-phenome Archive - Controlled access data - Human data around molecular medicine - http://www.ebi.ac.uk/ega/
Infrastructure provision - BBSRC: RNAcentral, MG Portal - MRC: 100k Genomes data
implementation - EC: COMPARE, MicroB3, ESGI,
BASIS - etc.
Challenges
• Data have high volume and grow rapidly
• Data are dynamic (continuous feed) and their application has urgency
• Users require arbitrary and ad hoc access
Tara Oceans
Tara Oceans
Capacity
Infectious disease
• Opportunity: A methodological revolution in clinical and public health towards shotgun sequencing-based methods
• Scientific power: Sequence harbours rich information
• Diagnostic: identification, typing, resistance profiling, etc.
• Public health: outbreak detection, response strategy, vaccine development
• Mechanistic: host interactions, pathogencity, virulence, transmission, anti-
microbial resistance
• Informatics roles for EMBL:
• COMPARE: Rapid global sharing of surveillance and outbreak data, systematic integrated analysis, compute provision (Embassy)
• Standards for reporting, analysis and the communication of results
• New algorithms and analysis methods
• User interfaces for surveillance data reporting , across the domains
COMPARE: recently launched Horizon 2020 project in which EMBL-EBI is informatics provider
Global Microbial Identifier: Initiative with EMBL-EBI involvement supporting technologies, standards and data sharing for pathogen surveillance
COMPARE platform
Embassy infrastructure DTU infrastructure Embassy virtual domain
COMPARE workflow engine COMPARE Portal
‘Default’ tools
‘Hosted tools’ INSDC data exchange
Sources
COMPARE Registry
COMPARE Data Resource
Processes Portals and environments
Public data
Managed access data
Private data
API API
EBI infrastructure
Assembly & alignment
AnnotaHon
Typing
Food workflow
development
Clinical workflow
development
Outbreak workflow
development Workflow integraHon
API
INSDC data exchange
COMPARE platform
Embassy infrastructure DTU infrastructure Embassy virtual domain
COMPARE workflow engine COMPARE Portal
‘Default’ tools
‘Hosted tools’ INSDC data exchange
Sources
COMPARE Registry
COMPARE Data Resource
Processes Portals and environments
Public data
Managed access data
Private data
API API
EBI infrastructure
Assembly & alignment
AnnotaHon
Typing
Food workflow
development
Clinical workflow
development
Outbreak workflow
development Workflow integraHon
API
INSDC data exchange Urgency
Personalised medicine
• Motivation: Personalised studies of variation, cancer mutation, epigenetics, regulation, expression require references for comparison and interpretation
• As part of GA4GH, EMBL-EBI is working on
• Resources serving reference human genomic and transcriptomic data, including Google read API, variant ‘Beacons’, etc.
• CRAM compression supporting greater data fluidity and APIs to allow direct computational access
• Delivery and synchronisation of high volume datasets to local Embassy and remote cloud infrastructures
• Past and current FP7 projects include SLING, BASIS, ESGI
Personalised medicine
• Motivation: Personalised studies of variation, cancer mutation, epigenetics, regulation, expression require references for comparison and interpretation
• As part of GA4GH, EMBL-EBI is working on
• Resources serving reference human genomic and transcriptomic data, including Google read API, variant ‘Beacons’, etc.
• CRAM compression supporting greater data fluidity and APIs to allow direct computational access
• Delivery and synchronisation of high volume datasets to local Embassy and remote cloud infrastructures
• Past and current FP7 projects include SLING, BASIS, ESGI
Arbitrary access
ENA data (NFS)
FIRE1
ENA conventional read data delivery
ENA metadata
Conventional infrastructure (FTP, Aspera, GridFTP)
ENA data (Cleversafe)
FIRE2
ENA Embassy read data delivery
ENA metadata
Conventional infrastructure (FTP, Aspera, GridFTP)
FUS
E
HTTP
ENA data (Cleversafe)
FIRE2
ENA Embassy read data delivery
Embassy cloud infrastructure (VMWare -> OpenStack)
ENA metadata
Marine cache
Pathogen cache
CRAM cache
Tara Oceans Embassy
COMPARE Embassy
GA4GH Embassy
Conventional infrastructure (FTP, Aspera, GridFTP)
FUS
E
HTTP
ENA external read data delivery
…phase II
EMBL-EBI Embassy Cloud
Steven Newhouse
Head of Technical Services
The Challenge Facing EMBL-EBI
• Volume and variety of genomic data expanding
• EMBL-EBI data doubling every year - replication is challenging
• Infrastructure currently 50,000 CPUs & 60+PB
• Need to support complex analysis scenarios
• Web and programmatic access to services (3M unique users)
• Access to both public and managed access data sets
• Bespoke workflows and tools across a variety of domains
• Hard for users to replicate data sets for local analysis
• Use the ‘cloud’ to bring local analysis to EMBL-EBI data
18
EMBL-EBI Embassy Cloud
• Service hosted at EMBL-EBI data centres
• Direct network access to public and managed data sets
• Direct network to access public services
• Expect both academic and commercial users
• Technical Implementation
• Logically isolated outside EMBL-EBI’s LANs
• Secure flexible infrastructure for both tenant and host
• Resources exposed using VMware’s vCloud Director & OpenStack
• Provide isolated IaaS clouds to multiple users
19
Why ‘Embassy’ Cloud?
• An embassy is sovereign territory in a host country
• Host Country: EMBL-EBI Data Centre
• Sovereign Territory: Host Country not allowed to enter
• Virtualisation provides the protection for ‘tenant’ and ‘host’
• Host puts boundaries in place to protect it from the tenant
• Tenant has freedom and control within those boundaries
20
Embassy Cloud Concept
21
Virtualised EMBL-EBI Hardware
Pub
lic D
ata
Man
aged
Dat
a
Pub
lic S
ervi
ces
Em
bass
y C
loud
1
Em
bass
y C
loud
2
Em
bass
y C
loud
3
Priv
ate
Dat
a
PanCancer
User Benefits for the IaaS Model
• Tenant organisations get an empty virtual infrastructure
• They establish their own virtual machines and networks
• System administration performed by the tenant
• EMBL-EBI staff have no access to the VMs
• Added value from EMBL-EBI over other clouds
• Machines and data hosted in known jurisdiction
• Direct network data sets (public & managed access)
• Direct network access to public EMBL-EBI services
22
Benefits to EMBL-EBI of the IaaS Model
• A secure collaborative workspace
• Work does not contend with main EMBL-EBI resources
• Clearly define the committed IT resources and data
• Explore how to build more data focused analysis services
• Move the analysis to where the big data is located
• Learn from and inform other big data scientific communities
23
Embassy Cloud: Typical Uses
• Collaborative Environment
• Neutral ground outside internal network
• CTTV: Resources and VMs to host intranet, databases, …
• Data Staging
• Undertake submission from local machine (following data staging) rather from remote location
• BRAEMBL: Remote submission unreliable due to file upload
• Data Analysis
• Large scale management and analysis of data
• PanCancer: 1,000 cores, 2.5 TB RAM, 1.0 PB HDD
Issues
• Object Store Storage Infrastructure
• Essential for scalable high-performance storage
• Applications need to adapt to flat model
• Current caching strategy will have a limit
• Sharing resources between sites/communities/clouds
• Adopt a standards based model for federating resources
• Solutions for uploading and distributing VMs (+containers?)
• Replicating large data sets to ‘attract’ workloads to a cloud
25
Gaps à Activities à Solutions?
• Data Set Replication
• Strategic pre-positioning of data into clouds
• Leverage JANET/GEANT, GridFTP + Globus Transfers, …
• Cloud federation for mobile computing
• EGI has a federated cloud and VM distribution model
• ELIXIR plans to build on existing infrastructure where possible
• Wide-area file access needed for collaborative data analysis
• High performance wide-area object-store
• Need access control for human related data
• Coordinated investment in infrastructure
• Where is the UK coordination? What coordination is needed?
• Integrating commercial resources where they add value
• Integration with EU Infrastructure (ELIXIR)
26