WAN Services /
(back to top)
(back to top)
(back to top)
How we did it
Quick Tips
Top ISPs, 09/25/00
Top ISPs, 08/28/00
Top ISPs, 07/24/00
Top ISPs, 06/26/00
Top ISPs, 05/29/00
Top ISPs, 04/17/00
Top ISPs, 01/31/00
Top ISPs Report
September 2000 report
Here's the latest quarterly report from Network World and Visual Networks as we look at the top ISPs in the market. Through statistical analysis of Visual's Internet BenchMark data for dial-up service (see How We Did It), we ranked the top ISPs in the business-to-business, national and regional markets, as they compare with other vendors within the same market. We've also included results for the past six months, to show whether an ISP's ranking this month has spiked or is consistently above average. The marketplace seems to be tightening up. In September, AT&T's WorldNet service stayed on top in the national retail market, performing above the industry average in three out of the nine categories tested. BellSouth and SBIS shared the top spot in the regional retail market, and AT&T's Global Network Service (GNS) and UUNET's GridNet service tied for the lead in the business-to-business market. As always, you can check out monthly reports on Network World Fusion.
September 2000's top ISPs, by category
Month-to-month comparison
Industry averages for September 2000
How we did it
Quick tips
Who we tested
Top ISP archive
Month-to-month comparison
Industry averages for September 2000
How we did it
Quick tips
Who we tested
Top ISP archive
| September 2000 Top ISPs, by category | ||
| ISP | Score* | Strengths | National retail |
| AT&T WorldNet | 15 | Low CFR %** (all three), initial modem connect speed, average time to log on, average Web throughput |
| MindSpring | 13 | Average time to log on |
| MSN | 13 | Low Web fail/timeout percentage |
| Verizon-West | 13 | Low Web fail/timeout percentage | Regional retail |
| BellSouth | 14 | Low CFR% (business), high initial modem connect speed |
| SBIS | 14 | Low CFR% (evening), low Web fail/timeout percentage |
| Ameritech | 13 | Low CFR% (24-hour, evening) | Business-to-business |
| AT&T (GNS) | 13 | Average time to log on |
| UUNET (Gridnet) | 13 | Average time to log on |
| *12 = aver. **CFR = call failure rate | ||
| Month-to-month comparison | ||||||
| ISP | April | May | June | July | Aug. | Sept. |
| AT&T WorldNet | 16 | 18 | 20 | 20 | 16 | 15 |
| BellSouth | 12 | 11 | 13 | 14 | 12 | 14 |
| SBIS | 6 | 11 | 12 | 11 | 11 | 14 |
| Ameritech | 15 | 14 | 11 | 12 | 15 | 13 |
| AT&T (GNS) | 13 | 11 | 15 | 15 | 14 | 13 |
| MindSpring | 9 | 13 | 13 | 13 | 13 | 13 |
| MSN | 11 | 10 | 4 | 12 | 11 | 13 |
| UUNET (GridNet) | 13 | 16 | 11 | 8 | 7 | 13 |
| Verizon-West | 12 | 11 | 12 | 11 | 12 | 13 |
| Industry averages for September 2000 | ||||
| Category | National retail | Regional retail | Business-to- business |
Top ISP in category |
| 24-hour CFR % | 3.30% | 2.70% | 9.70% | Ameritech |
| Evening-hour CFR % | 4.70% | 3.00% | 10.40% | Ameritech |
| Business-hour CFR % | 3.60% | 3.50% | 11.20% | Verizon-North |
| Initial modem speed (K bit/sec) | 47.76 | 48.09 | 47.94 | BellSouth |
| Avg. time to log on (sec) | 29.66 | 31.23 | 31.5 | AT&T WorldNet |
| Avg. DNS lookup (msec) | 468.92 | 657.94 | 584.73 | PSI |
| Avg. Web throughput (K bit/sec) | 4.49 | 4.82 | 4.64 | PacBell |
| Avg. download time (sec) | 23.01 | 23.2 | 24.89 | AOL |
| Avg. total Web fail % | 1.20% | 1.20% | 1.40% | SBIS |
How we did it
The data for this report comes from Visual Networks. Through its Internet Benchmark data, Visual Networks rates national, regional and business-to-business ISPs according to several factors, including call failure rate, modem connect speed and Web download performance. We took the raw data from Visual Networks and applied statistical analysis to rate the relative performance of each ISP. First, we derived the standard deviation of the numbers in each performance category. Standard deviation is a measure of how far the numbers in a series diverge from each other.
For each category in which an ISP performed better than one standard deviation from the industry mean, we award it one point. If the ISP did better than two standard deviations from the mean, we awarded it two points. Similarly, if an ISP did worse than the industry mean by more than one standard deviation, we took a point away from its score. If it did significantly worse, meaning two or more standard deviations, we took away two points. We started with a baseline of 12 points, so if an ISP scores 0 points it ends up with an adjusted score of 12. After we scored each category, we summed the results to get a number that indicated the reliability and performance of each ISP. A rating of 12 means an ISP was about even with its peers, or the industry average. In September, nine ISPs scored higher than 12 and six ISPs were average. The bad news, however, is that 12 ISPs scored less than 12 points. A list of all the ISPs tested, as well as an archive of previous reports, appears below. (back to top)
Quick Tips
DNS lookup
Throughput and connection speed aren't the end of the story when determining network performance. Domain Name System (DNS) performance adds additional overhead to almost any network activity. DNS translates host names like www.yahoo.com into IP addresses used by Internet routers. For example, when you open a Yahoo page, your browser requires at least one DNS lookup to complete before it can request the data for that page. A typical DNS request is about 80 bytes in size, taking about 15 msec to squeeze through a 56K bit/sec line to your ISP. Responses vary, ranging from about 100 to 400 bytes in size and taking 15 msec to 70 msec to return. On top of this, each ISP's performance will differ based on how its DNS infrastructure is deployed. The impact of DNS performance can be fairly complex. Local DNS lookup caches are maintained by Windows and other operating systems. These speed lookups when the same host name is requested repeatedly. DNS servers also maintain caches, ensuring that DNS queries for heavily used Web sites will often be answered quickly. In addition, each Web page may contain several objects from different servers - or all the objects could be from the same server, maximizing the effectiveness of a local cache. (back to top)
Top ISPs, 09/25/00
Top ISPs, 08/28/00
Top ISPs, 07/24/00
Top ISPs, 06/26/00
Top ISPs, 05/29/00
Top ISPs, 04/17/00
Top ISPs, 01/31/00
Subscribe to the ISP e-mail newsletter
Archive of Network World Fusion Focus on Internet Services newsletters
