University of Cheikh Anta Diop

The University of Cheikh Anta Diop in Dakar, Senegal, has a 64 kbps radio connection to Sonatel, the national telecommunications provider, for US$2,000 per month. Sonatel in turn has two Internet uplinks: a 128 kbps link to MCI in the United States and a 1 mbps to Teleglobe in Canada. UCAD shares its bandwidth by reselling it to four other institutions in the city through 64 kbps leased lines, helping to defray the university's Internet costs. The main Internet access center has nine terminals on an ethernet LAN, with a per-semester fee of 2,000 CFA (about US$3.25). The hours are 9 am to 6 pm, and there is a 25-minute limit on terminal time if there is anybody waiting for access. In addition, the network supports six dial-up connections and there are another 12 terminals on the LAN in a room usually reserved for classes.

Journals access: We had a great deal of difficulty accessing journal articles, particularly during the daytime hours. The HTML-based articles were accessible, but at a very slow transfer rate. Large PDF files consistently stalled completely on multiple attempts and failed to download, although we succeeded in obtaining a few very small files. During evening hours the transfer rate improved and even large PDF files became accessible, although still quite slowly. The onset of the improved results corresponded closely with the closing time of the main Internet access room and the end of the business day, when the other institutions sharing UCAD's bandwidth presumably logged off. Table 9 shows samples of sizes and transfer rates of files successfully obtained during normal business hours; Table 10 shows the same information for files obtained after 8 pm. These tables each include both PDF files and component files of HTML articles.

Table 9. Accessing journal articles from UCAD during business hours
File	Size in kilobytes	Transfer rate in bytes/second	Download time
1	21.1	112.0	3 min. 8 seconds
2	23.7	145.2	2 min. 43 seconds
3	30.6	105.4	4 min. 51 seconds
4	34.5	133.1	4 min. 19 seconds
5	52.4	82.8	10 min. 32 seconds
6	58.2	232.2	4 min. 11 seconds
7	62.0	91.6	11 min. 17 seconds
8	64.9	116.3	9 min. 18 seconds
9	112.6 (largest successful file)	89.2	21 min. 1 second
Total average data transfer rate: 99.9 bps

Table 10. Accessing journal articles from UCAD after 8:00 pm
File	Size in kilobytes	Transfer rate in bytes/second	Download time
1	36.5	323.8	1 min. 53 seconds
2	74.6	595.0	2 min. 5 seconds
3	86.4	641.0	2 min. 15 seconds
4	122.1	711.5	2 min. 52 seconds
5	144.3	691.0	3 min. 29 seconds
6	174.7	989.8	2 min. 57 seconds
7	224.1	481.4	7 min. 46 seconds
8	293.7	811.7	6 min. 2 seconds
9	461.9	554.4	13 min. 53 seconds
Total average data transfer rate: 554.0 bps

Technical modifications: Since UCAD has a well-established Linux operating system in place to which it remains committed, and its own proxy server, we did not install FreeBSD or Squid. However, the university's authentication server (TACACS) was crashing very frequently, and we were able to remedy the problem with a script that checks its status every 60 seconds and restarts it if necessary.

Assessment: Under present conditions it will be difficult for UCAD to make much use of online journals. The most likely explanation for the poor performance is that there are too many users for the available bandwidth. Reselling bandwidth to local institutions has certain economic benefits to the university, but these benefits must be weighed against the loss of Internet connection speed. The university is currently considering upgrading its connection to 128 kbps. An interesting experiment would be to see whether the performance improvement of such an upgrade might be matched at the current bandwidth level by eliminating the resale to third parties. If so, and if the cost of the extra bandwidth exceeds the revenue from the resale, then it would be more economical to forgo the revenue and the upgrade, and simply reserve the existing bandwidth for the university community. In the meantime, expanding the Internet access hours on campus might help to alleviate some of the congestion, but probably not to the extent that journal articles become easily accessible.

However, it is also interesting to note that there is some evidence that strategically placed mirror sites might go a long way toward improving access to online journals. During regular business hours at UCAD, i.e., the hours of worst results for journals, we were able successfully to download a one megabyte file from the Internet Africa server in Capetown. Further investigation revealed that Internet Africa and Sonatel both have direct connections to Teleglobe in Canada, which probably accounts for the speedy connection. If critical information resources were mirrored on servers that are reachable by similarly direct routes (common service providers), it is likely that access would be greatly improved.

AAAS > International > Africa