TSTN-029: The Engineering Facility Database Large File Object Infrastructure

  • Tiago Ribeiro

Latest Revision: 2022-01-05

Note

This technote is in rough draft.

TBD

1 Introduction

The Engineering Facility Database (EFD) has a requirement to support objects that are too large or unwieldy for regular SAL commands, events and telemetry. The Large File Annex (LFA) is the solution to this requirement. The LFA uses Amazon’s Simple Storage Service (S3) to store these objects.

2 Architecture

The LFA uses Amazon’s S3 to store objects. S3 uses the concept of buckets to organize objects. Conceptually for the LFA, buckets correspond to sites i.e. Tucson Test Stand, Summit, Base Test Stand. Bucket names have the following format rubinobs-{s3category}-{s3instance}. s3category refers to the category of bucket, currently there is only LFA defined. s3instance refers to the site name.

  • tuc - tucson

  • cp - summit

  • ls - base

Each object (file) uploaded to a bucket must have a key (name) that follows the Vera C. Rubin standard. The key format is {fullsalname}/{generator}/{yyyy}/{mm}/{dd}/{fullsalname}-{generator}-{other}{suffix}

fullsalname:

Refers to full CSC name including index name if exists {salname}:{salindexname}.

generator:

metadata that outlines where the object was generated i.e. FiberSpectrograph using different filters

yyyy-mm-dd:

Refers to the observing day, which does not turnover during the night observing.

other:

Extra metadata

suffix:

object extension i.e. fits

3 Client Support

Adding LFA support using salobj