[Scholarly Communications Project]
Server Requirements for
To assist other libraries in establishing ETD collections, below are recommendations based on the current activities at the Scholarly Communications Project (SCP), University Libraries, Virginia Tech. When SCP began developing ETD process, access, and archiving systems in 1994, we used on-hand equipment and staff and free software available on the Internet. Our initial server was a NeXt 3.3 running HP, and it accomodated ETDs, ejournals, and online newspapers quite well. Winter 1996 the OpenText LiveLink search engine replaced freeWAIS and in the fall of 1997 the system described below became the primary server for ETDs and all other SCP resources.
Electronic Theses and Dissertations
Located in Newman Library, the SCP system is a Sun Netra Server with a 200 Mhz Ultrasparc processor and 128 Mb of RAM. Several gigabytes of disk space are allocated for the operating system and related tools. A 4 Gb partition currently houses the ETD collection. The Netscape Enterprise Server runs on Solaris 2.6 and Perl 5.004_01. For back-ups SCP has a Sun 8mm Ultra Wide SCSI tape drive that takes 170m tapes with a compressed storage capacity of 40Gb.
Scalability is important so that drives can be added as needed. A large drive (9 Gb) with a single partition is ideal. Redundant drives for backups or mirroring are a plus.
Have enough memory to process the work load, i.e. to handle dozens of cgi posts and thousands of hits each day. Memory is especially valuable if the same machine is used to index the collection. (128 Mb is good to begin with as wellas for a production setup running on a Sun).
If possible, the server should be as isolated from local network traffic as possible; i.e. it should not share bandwidth with workstations. A fast network connection and robust ethernet card are also plusses.
The server itself should run a mature and growing operating system, one for which there are commercial and public resources. A modern version of UNIX is strongly recommended.
Run a web server which is robust, fairly easy to configure, and for which there is a sizable knowledge-base among the online community. Consider Apache (available for most platforms without charge) and Netscape Enterprise Server. Any web server which is to be used should work well with scripts, and should have powerful, flexible, and easily implemented security. Netscape Enterprise Server in particular allows wildcards to be used for security, which we've found to be a bonus because we can restrict access to all URLs of the type /theses/delayed/*/materials instead of placing a .htaccess file in each materials subdirectory.
Run a relatively new version of perl and CGI.pm to use the existing ETD submission scripts. Perl is available for all modern operating systems, although the UNIX flavors are the best supported.
Perform a full backup, system and collection-wide on a weekly basis. Perform incremental backups nightly or even twice a day according to the load. Ideally, the tape drive should be able to backup the entire system in a few hours, and should be able to store not only the current contents of the collection but a significant amount of added material as the collection grows.
http://scholar.lib.vt.edu/about/server1.html Anthony Atkins and Larry Poltavtsev
Scholarly Communications Project, University Libraries, Virginia Tech.
Dec. 12, 1997; updated Sept. 18, 1998 (GMc)
[[ETD main page]]
[[SCP home page]]
| www.ndltd.org/ |
| ETD main page |
| SCP Home |