This document summarizes testing of the LOCKSS system for preserving access to subscription journals. It describes: installing LOCKSS boxes at several institutions and connecting them to a test network; harvesting content from an Open Journal Systems journal management system; introducing errors like missing or corrupted files and verifying LOCKSS could recover from other copies; and concluding the software is robust but managing the central server requires more expertise. It questions whether license agreements could allow storing subscription journal content locally using LOCKSS and whether the Global LOCKSS network could support this.
2. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Agenda
• The problem description
• Installation of LOCKSS
• Harvesting from OJS
• Recovering from bit corruption
• Conclusions on the technical evaluation
• Thoughts about LOCKSS for persistent
access to subscription journals
4. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
The Problem Description
Test af Global LOCKSS
I dette hovedelement gennemføres test af Global LOCKSS med henblik på at
afklare, om Global LOCKSS er en brugbar platform for lagring af lokale kopier af
de videnskabelige publikationer, som universitetsbiblioteket har licens på. Det
første skridt bliver udarbejdelse af en liste over krav til systemet som danner
udgangspunkt for den efterfølgende test og evaluering. Her tages udgangspunkt i
Neil Beagries fjerde anbefaling og de krav, som han har registeret.
Test og evaluering af Global LOCKSS vil omfatte områder som etablering,
opsætning, daglig drift og ”trigger events”. Der testes for eksempel, om man kan
tilgængeliggøre en dansk publikation som endnu ikke er omfattet af en Global
LOCKSS-aftale og om man kan reetablere tabt indhold i universitetsbibliotekets
LOCKSS via indhold fra andre LOCKSS servere. Der installeres en LOCKSS-
BOX hos SB, KB og DTU Bibliotek for at teste blandt andet denne funktionalitet.
Metodemæssigt anvendes der en blanding af fælles drøftelser, analyser, installation
og test af prototyper samt lokale indsatser.
5. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
The Problem Description
Målet er, at etablere 3 fulde prototyper af
Global LOCKSS, hvor alle funktionelle krav
som integration med proxyserver, ingest og
logning af materiale, korrekt replikering
samt stabilitet, er aftestet.
Vurdering: Kravene i pkt. d har været
anvendt som udgangspunkt for en test af
prototyperne og der er gennemført test ved
projektets afslutning.
6. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
The Problem Description
Deltagere Opgaver Mande
måned
er
Test af LOCKSS
prototype:
Projektledelse KB Projektledelse og afrapportering 0,5
Krav og
opsætning af
testmiljø
KB, SB og
DTIC
Udarbejdelse af liste over krav.
Installation og konfiguration af
servere
1,5
Test og evaluering KB, SB og
DTIC
Implementering og gennemtestning
af indholdmæssige funktionaliteter
som Integration til EZ-proxy, Ingest af
danske og udenlandske (med
eksisterende LOCKSS aftale)
publikationer.
3
8. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Installation of LOCKSS
• Installing a LOCKSS box is a reasonably
well described process
• While LOCKSS is distributed it does need
access to a central server that configures
the LOCKSS network
• We asked Stanford to connect it to the
Global LOCKSS test network
• Support from Stanford was very helpful
9. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Installation of LOCKSS
• The central server distributes configuration
to the LOCKSS-boxes about:
– Participants in the network
– Journals that can be harvested and how
– Whether journals are live
• Standard procedure is to allow the
configuration server full access to the
LOCKSS box, including SSH
10. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Installation of LOCKSS
• At a later point it became clear the Global
LOCKSS would not host Danish journals
• Instead a Danish Private LOCKSS
network was defined
• Stanford was very helpful in setting this up
for us
11. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Installation of LOCKSS
• To run our own Danish Private LOCKSS
network, we would need to be able to run
the central LOCKSS server
• Installing and running a LOCKSS box is
fairly easy. Running a central server does
not seem to be so
13. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Harvesting our OJS material
• There are two things needed to harvest an
OJS journal
– The OJS server must present itself as ready
to be harvested
– The LOCKSS network must be configured to
harvest the journal
14. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Harvesting our OJS material
• Configuring the OJS server is easy
• A standard plugin will present the license
and structure of the journal for a LOCKSS
network to harvest it
• In newer versions of OJS the plugin
supports more granularity for private
LOCKSS networks
15. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Harvesting our OJS material
• Configuring the LOCKSS network to be able
to harvest the material turns out to be
something done on the central configuration
server
• Again Stanford was very helpful. If we want
to run out own configuration server we need
to make this configuration
• Once it has been set up, harvesting the
material is a matter of a single click
17. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Recovering from bit corruption
• We set up four test scenarios
– The file is missing, but available at the journal
server
– The file is missing, and not available at the
journal server
– The file is corrupted, but available at the
journal server
– The file is corrupted, and not available at the
journal server
19. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
Recovering from bit corruption
• The bit errors were all discovered
• The first set of errors introduced were
discovered without any intervention
• In the second set we did a manual check
(both to test this, and to reduce the time)
• The journals were reharvested from the
journal server when it was available
• The journals were harvested from other
LOCKSS boxes when it was not
22. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
LOCKSS technical evaluation
conclusions
• The LOCKSS box software seems robust
and well documented
• The support is excellent
• The end user interfaces are ”design by
engineers” - fully functional but sometimes
quirky
• The central server seems complex and
less documented
23. KÅRE FIEDLER CHRISTIANSEN
kfc@statsbiblioteket.dk
LOCKSS for persistent access
to subscription journals?
• Technical considerations
– The software is robust and functional
– Strong qualifications are needed to run the
central server
• Do we want to run it ourselves or can Stanford host it?
• Content considerations
– Can we get license agreements to put the
content in the Danish network LOCKSS boxes?
– Will the Global LOCKSS agreements cover this?