Easy Object Storage Import/Export Using the S3 Connector on Jetstream
1. Easy Object Storage Import/Export
Using the S3 Storage Connector
Lee Liming (LLiming@uchicago.edu)
University of Chicago
funded by the National Science Foundation
Award #ACI-1445604
2. funded by the National Science Foundation
Award #ACI-1445604
http://jetstream-cloud.org/
Cloud, specialized for the research community
Researchers browse VM images
saved by other researchers
– Web interface provided by
CyVerse
– Share fully-installed systems with
applications ready to go
– OpenStack API/CLI access for
advanced users
Images can be launched on VMs
scaled for research
– 1-44 vCPUs
– 2-120 GB RAM
– 8-480GB disk
Use is allocated by NSF/XSEDE
3. funded by the National Science Foundation
Award #ACI-1445604
http://jetstream-cloud.org/
Jetstream’s object storage (Swift/Ceph)
• 1.25 PB usable storage
(If all of JS was object store, no VMs,
volumes, etc. Of course, it’s not…)
• Access via Swift APIs
(and any tools that use them)
– Horizon web app
– OpenStack CLI
• AWS S3-compatible
• A place to learn &
evaluate before the
commercial cloud
4. funded by the National Science Foundation
Award #ACI-1445604
http://jetstream-cloud.org/
Research data starts & lives at home
• Campus systems or research facilities/labs
– Data often comes from
facilities/labs/observatories
– Results are often needed elsewhere
– Publishing requirements are becoming important
• Very often, available via Globus
• So…how do we get it from there into Jetstream’s
object store for use via Ceph, Swift, S3, etc?
– Must be: easy, fast, reliable
• (Jetstream to “next step location” is also
valuable…)
5. funded by the National Science Foundation
Award #ACI-1445604
http://jetstream-cloud.org/
Fire-and-forget transfers into Jetstream buckets
6. funded by the National Science Foundation
Award #ACI-1445604
http://jetstream-cloud.org/
So how does this work?
• Globus Connect Server (the
server version of Globus)
offers storage connectors to
subscribers, including an
AWS S3 connector.
• The AWS S3 connector works
equally well with OpenStack
Swift’s S3 dialect.
7. Setting it up
1. Create a VM w/floating IP.
2. Install GCSv5.2 w/S3 connector.
3. Create an S3 storage gateway
requiring an XSEDE ID.
4. Use Globus web app to create a
collection and supply S3-style
credentials.
5. Use the collection to transfer data
in or out.
6. Shelve VM until needed again.