1 of 22

SRCNet Infrastructure updates for CHSRC

Swiss SKA Days - September 6th, 2023

Carolina Lindqvist and Pablo Llopis

SRCNet Coral update

2 of 22

SRCNet Objectives

3 of 22

SRCNet Roadmap

Focus on Data Management Solutions (Data and Metadata)

4 of 22

Coral Team - Collaborating Institutions

Only active collaborators shown. More institutions as observers and expected to join in the near future

5 of 22

Coral Team - Purpose and Goals

🔍 Identifying computing architectures and resources available in SRC sites

👩🏼‍💻 Prototyping

    • Data management testbed
    • Authentication and Authorisation Infrastructure
    • Science Platform
    • Assessing performance in different SRC sites and architectures

🚀 Impact: gather and share knowledge

🎓 Provide early feedback on the adoption of proposed solutions

6 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
  • Building blocks for science platforms
  • Collection of workflows

7 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
    • Supporting Rucio Data Lake operations
    • Storage Inventory and Metadata OpenCADC services
  • Building blocks for science platforms
  • Collection of workflows

8 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
  • Building blocks for science platforms
  • Collection of workflows

9 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
  • Building blocks for science platforms
    • Jupyterhub
    • Dask
    • Data lake integrations
  • Collection of workflows

10 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
  • Building blocks for science platforms
  • Collection of workflows

11 of 22

SRCNet high level topics for PI19 (Coral Team)

  • Data management solutions
  • Building blocks for science platforms
  • Collection of workflows
    • Offer a “shopping list” of reference workloads, organised by category.
    • Integrate with CI/CD to run workloads as a pipeline
      • SRCNet system validation
      • Performance and regression testing
    • https://gitlab.com/ska-telescope/src/src-workloads

12 of 22

SKA SDC JupyterHub prototype deployment

  • Based on a Helm chart provided by Magenta team
  • DNS (domain name) configured at external service provider
  • Appropriate configuration for SKACH infrastructure
    • All services running in a Kubernetes cluster
    • Separated by Kubernetes namespaces
    • Enabling HTTPS with Let’s encrypt certificate and Certmanager deployment
    • Ingress Nginx deployment and configuration

13 of 22

Nginx ingress

Certificate

Certmanager

Jupyterhub proxy

Jupyterhub

User notebooks

DNS configuration

User notebooks

Kubernetes cluster

hub.src.skach.org

HTTPS

14 of 22

OpenCADC Storage Inventory (SI)

Mini-SRCNet prototype: A bird’s eye view (All)

Storage Inventory Data

Storage Inventory Metadata

Global IVO Abstr.

Global Site

Local Site

SI Query service

(Luskan)

SI File service

(Minoc)

Metadata sync

(Fenwick)

Metadata validate

(Ratik)

File sync

(Critwall)

File validate

(Tantar)

IVO Lookups

(Reg)

Query service

(Luskan)

File locate

(Raven)

User exposed

Backend

Metadata sync

(Fenwick)

Permissions

(Baldur)

GMS

SKA IAM

SKA Global service

(External to OpenCADC)

Inventory +

Storage

CAOM

Query TAP

(Argus)

CAOM

Datalink

(Bifrost)

CAOM

Query TAP

(Argus)

CAOM metadata repository

(Torkeep)

Metadata sync

(Icewind)

15 of 22

OpenCADC Storage Inventory (SI)

Mini-SRCNet prototype: A bird’s eye view (SI only)

Storage Inventory Data

Storage Inventory Metadata

Global IVO Abstr.

Global Site

Local Site

SI Query service

(Luskan)

SI File service

(Minoc)

Metadata sync

(Fenwick)

Metadata validate

(Ratik)

File sync

(Critwall)

File validate

(Tantar)

IVO Lookups

(Reg)

Query service

(Luskan)

File locate

(Raven)

User exposed

Backend

Metadata sync

(Fenwick)

Permissions

(Baldur)

GMS

SKA IAM

SKA Global service

(External to OpenCADC)

File Storage

Inventory DB

16 of 22

Image credit: https://github.com/opencadc/storage-inventory

17 of 22

OpenCADC Storage Inventory (SI)

Mini-SRCNet prototype: A bird’s eye view (CAOM only)

Storage Inventory Data

Storage Inventory Metadata

Global IVO Abstr.

Global Site

Local Site

IVO Lookups

(Reg)

User exposed

Backend

Permissions

(Baldur)

GMS

SKA IAM

SKA Global service

(External to OpenCADC)

Inventory

CAOM

Query TAP

(Argus)

CAOM

Datalink

(Bifrost)

CAOM2

TAP Query

(Argus)

CAOM2 Metadata repository

(Torkeep)

Metadata sync

(Icewind)

File locate

(Raven)

18 of 22

Image credit: https://github.com/opencadc/storage-inventory

19 of 22

  1. User PUTs a file to the site1.minoc service, either directly or via negotiation with a global raven service.
  2. global.fenwick.site1 discovers the new inventory metadata for the file by querying site1.luskan.
  3. site2.fenwick.global discovers the new inventory metadata for the file by querying global.luskan.
  4. site2.critwall finds the locations of the new file via global.raven -- this returns a list of URLs from which the file can be downloaded.
  5. site2.critwall downloads the file from site1.minoc.

Image credit: https://www.opencadc.org/storage-inventory/ops/

OpenCADC Storage Inventory (SI)

File Synchronization example

20 of 22

DEMO 1

21 of 22

DEMO 2

22 of 22

PI20 features

  • Mini-SRCNet Science Platform deployment
  • Continuation of work for SRCNet workloads repository
  • Migration to new CSCS Kubernetes cluster
    • Rucio + Storage Inventory service migration
    • Data migration for Rucio + Storage Inventory
    • JupyterHub, Dask
    • Internal services (monitoring, Secret management system, Registry)
  • Explore Workload Management System solutions
  • Secure Authentication for SRCNet Nodes in Shared HPC Environments