UNIVERSITY
Northwestern
|LIBRARIES
aservation-First
Repository
Laura
Alagna,
Adam
Arling,
Carolyn
Caizzi,
Michael
Klein,
SIECTale
YN
OI0N
o
o
MEDE-1ViTo
ISTed
alo]
o[-
A
s
E-T-Ta
ST
A
F-X'VANI=T
o
I
(010
T
g
Repository
and
Digital
Curation
-
Northwestern
University
Libraries
-
Northwestern
University
A
New
Workflow:
Minimizing
the
Danger
Zone
ITEM
INGESTED
TO
PRESERVATION
STORAGE
s
—t
U
]
Using
a
modified
Ingest
Sheet,
the
files
are
ingested
and
organized
by
work
with
minimal
administrative
metadata
applied
en-masse.
After
works
are
inventoried,
they
are
digitized
at
scale
using
in-house
equipment
or
vendors
de-
pending
on
the
complexity
and
rarity
of
the
object.
Files
are
placed
in
preservation
storage
(S3/Glacier
synced
to
a
local
enterprise
solution)
and
critical
preser-
vation
tasks
are
executed.
Files
are
uploaded
to
an
ingest
staging
bucket
in
Amazon
S3.
Digitized
works
are
recorded
on
an
Ingest
Sheet
(a
spreadsheet
listing
every
file,
accession
number,
and
role)
and
QCd.
Audit
trail
created
for
preservation/file-level
actions.
De-
rivatives
such
as
pyramidal
TIFs
are
generated.
ITEM
INGESTED
o
New
()
DAYS
OR
WEEKS
workflow
ITEM
DIGITIZED
Previous
orkflow
()
MONTHS
OR
YEARS
e
OBJECT
IS
DESCRIBED
BEFORE
IT
IS
IN
PRESERVATION
STORAGE
Preservation
Danger
Zone
The
time
that
files
spend
in
temporary
storage
-
after
digitization
but
Q\:
before
ingestion
into
preservation
system.
\
Our
new
workflow
minimizes
the
time
objects
spend
in
the
danger
zone.
dws
METADATA
ADDED
TO
ITEMS
IN
PRESERVATION
STORAGE
b
clixir
ITEMS
DESCRIBED
VIA
BATCH
ACTIONS
ITEMS
PUBLISHED
2
[
Using
the
Ul
or
a
batch
export/edit/re-import
work-
flow,
works
are
described
by
metadata
and
organized
into
user-facing
collections.
Work
metadata
is
QC'd
and
items
are
published
(visibility
is
set).
Published
works
become
accessible
via
separate
Digital
Collections
(React)
front
end
and
the
ElasticsearchAPI.
Metadata
is
stored
in
Postgres
and
written
to
Elas-
ticsearch.
Files
remain
at
rest
in
preservation
storage,
all
ac-
tions
take
place
without
affecting
preserved
file.
ITEM
FULLY
DESCRIBED
'
ITEM
PUBLISHED
ITEM
INGESTED
ITEM
FULLY
DESCRIBED
Our
Tech
Stack
Backend
built
with
Elixir
using
the
Phoenix
framework.
Concurrent
and
multi-stage
ingest
pipeline
in
Elixir
with
Broadway.
Flexible
and
descriptive
GraphQL
API
layer
using
Absinthe.
React
front
end
with
Apollo
Client
for
state
mangement.
Leverages
best-in-class
AWS
solutions
like
S3,
Glacier,
Lambda
and
others.