
MCI
We also interviewed several of UUNET's customers, which rely on UUNET for monstrous bandwidth requirements. Another particularly large content provider was in the midst of working out a DS-3-to-OC-3 upgrade. It was interesting to note that its primary reason for using UUNET was simply because of the vast number of organizations attached to that backbone.
A few of the other bandwidth-conscious multihomed clients reported that, lately, UUNET's backbone has really felt the heat. They also mentioned that while connectivity exists during network outages, backup routes are often painfully slow. Most agree, however, that slow routes are better than no routes, which attests to UUNET's redundant design and multiple private peering points. Incidentally, UUNET and ANS were the only providers willing to share any information about private peering. We see this as being rather absurd considering you could, given enough time, figure out exactly who is peered with who. Fortunately, we have better things to do than plot route paths. We'll just have to wait until the other providers stop this silliness before more extensive information can be published.
Finally, one of UUNET's shortcomings is support. Customers agreed that once they reached the upper levels of support, expertise was found and issues were dealt with, but many felt it was a battle to get to that point. In addition, there were complaints of little to no communication concerning outages.
GTE Internetworking
At the heart of GTE Internetworking lies BBN Corp., once home to the organization that helped build the original ARPANET. The unification of core Internet expertise with a telecommunications giant gives the ISP formerly known as BBN Planet access to previously unobtainable facilities-based resources. Combined with the promise of a larger backbone and the recent acquisition of the midsized ISP Genuity, GTE offers a healthy array of expensive services.
GTE's current IP infrastructure is based on Sonet technology running over a combination of OC-3 and OC-12 links. Unlike MCI, which has deployed ATM, one of the advantages of this approach is it avoids some ATM overhead costs. In addition, running IP over Sonet can avoid latency issues caused by cell loss. Although this sacrifices any hope of Layer 2 QoS (quality of service) control, this level of service on the Internet is little more than a pipe dream and may be addressed at higher layers in the future. Many providers are following suit when deploying their next-generation backbones.
GTE's growth model is steering away from NAP connectivity. In fact, GTE network architects go so far as to state that they will do only what is necessary to maintain functional NAP attachments--all efforts have been moved to increase private peering relationships. We agree this is the best approach (see "Private Peering: Nothing But Net," on page 120).
While servicing some of the industry's larger names (Cisco Systems, Hewlett-Packard Co., Novell and Silicon Graphics to name a few), GTE is not immune to petty mistakes. Although there were no earth-shattering complaints with GTE's service, many clients felt that GTE was unresponsive to their nonemergency issues: DNS updates slipping through the cracks, crossed wires, lost contact information, unresolved incidents mysteriously disappearing, to name just a few. Like UUNET, GTE has a strong past with what looks to be a bright future--but its customer base seems to think it needs to concentrate on present-day issues.
Sprint Corp.
At the base of Sprint's global Internet services lies a massive OC-3/OC-12 Sonet-based backbone. Using IP over Sonet, Sprint boasts high performance in more than 70 countries. Offering clients SLAs on both dedicated and dial-up availability, Sprint has positioned itself as an attractive provider for international organizations. Unfortunately, like most carriers, Sprint's service and support offerings are a little behind the times.
Much like MCI, Sprint customers had similar complaints. Sprint's client base seemed pleased with base networking services: Occasional outages are tolerated, latency issues are negligible and technical issues are eventually resolved. A few organizations are experiencing congestion problems on Sprint's backbone, but these did not seem to be common throughout all of those interviewed. However, as one customer phrased it, "Sprint needs some work on its bedside manner."
To Sprint's credit, many customers found that when a problem was reported, Sprint engineers were already working on it. However, very rarely were customers informed beforehand. One of the complaints heard most from carrier-based clients was the inability to interact. Known problems often had to traverse internal chains, yet customers had already identified the solution.
Sprint's service features are heavy in some areas, light in others. For example, Sprint offers an intranet service that takes advantage of the private IP ranges for internal communications. It has also addressed security concerns by partnering with firewall vendors. The one area that is noticeably absent is that of Web hosting--Sprint has nothing in place at this time.
|