Research overview
Articles

- Links to articles
- Cambashi in print

PLM debate
White papers
Reports
- Best practice case studies
Resources
Free stuff
 
Cambashi goes Broad(er)band

Much has been written about the opportunities that the new broadband era of ADSL and Cable Modems will bring to the small enterprise, but there are few reports about what has been achieved in practice. Cambashi is lucky enough to be based in Cambridge - a part of the UK where both ADSL and Cable Modems services are available. As well as conventional modems and ISDN, the company and its employees now have direct experience of cable modems (NTL) and ADSL (two flavours).

Effective communication is very important to Cambashi. Nowadays, pretty well everything of any importance comes over 'Net. The fax-server has been retired, the postman could retire, but loss of the e-mail server has become a show-stopper. Cambashi may be small, but its business is knowledge and its customers include many world-ranking computer corporations. Its network of employees and associates is far flung -- extending well beyond the company's HQ. Remote employees now use VPN links to gain secure access to the company computer resources. And as well as email and conventional web browsing, Cambashi runs a web spider and a web-server in support of its e-Xpert services.

Until recently, communications systems were reasonably conventional: a 128K leased line supporting inward access to the company's web and e-mail servers and providing staff access to the Web. Remote employees accessed company resources by means of a single auto-answering modem. It all worked, but it was both expensive and restrictive. With only one modem, tele-working wasn't a serious option.

ADSL promised a solution better fitted to the needs of a small company: a considerably cheaper and potentially faster service. The drawback was the lack of a service level agreement and poor reports of early ADSL installations.

Barring a few problems associated with the systems initial installation Cambashi's experience has been good. The leased line has gone; the DNS records have been changed and the company is now 100% ADSL-driven. Performance has, if anything improved: despite the fact that the service is 20:1 contended, download rates usually exceed the 128kbps maximum of the previous service - sometimes by a large factor. The 256kbps upload maximum is more restrictive, but in practice still provides a faster service.

At about the same time that Cambashi was installing its corporate 2Mbs/256kbps, 20: 1 contended, non-NAT ADSL service through Demon Internet, domestic broadband services were finally being launched in Cambridge. In parts of Cambridge both BT "Home 500" service and NTL's domestic-only cable modem service are available.

With adequate inward capacity, and with firewall protection and a VPN server in place at the office, these new domestic broadband services made teleworking a much more viable option for those members of the company lucky enough to be able to exploit them.

So it has proved. At present Cambashi has two users of domestic broadband services: one using an NTL cable modem and the other BT's "Home 500" service.

Those for whom domestic broadband isn't an option have still been able to benefit from "Surftime". By capping the cost of Internet Access, Surftime has made remote access a much more economically attractive option, even if it remains frustrating slow and sometimes of very variable quality.

Ping time consistency, analogue modem compared to cable modem

For a company whose expansion opportunities were originally constrained by the availability and cost of office space, broadband communication has provided valuable additional flexibility. Some of its benefits are surprising. It is widely observed, for example, that teleworkers miss-out on the social interaction that goes with a busy office. Cambashi has found that there are times when missing out on social interaction, whilst still being connected, is just what is wanted to finish an important project.

When Cambashi was trying to make comparisons between alternate services, there was very little real data to go on. Internet Newsgroups were full of postings from individuals who had problems and were soliciting solutions. For fairly obvious reasons, positive posts were rarer and less specific. Objective comparisons, of the "A is better than B with respect to following metrics" were rare. Since Cambashi is in a position to make a few comparisons, and in the hope that it might encourage others to do the same, we have created a table of observations derived from data we collected whilst trouble-shooting.

When looking at these data, please observe the obvious caveats: the sample data is limited, it relates to a particular time period, a particular ISP, a particular telecoms provider, particular equipment at both ends, … . We would like to collect better data but we would also like to learn from others experience.

Please send any comments/observations or suggestions to Ralph.Seeley@cambashi.com

First hop time (msecs)

Long ping incidence (%)
Incoming ADSL Transit (msecs)
Inter-site hops
Connection type
Identity
0.5
*
24.6
9
direct
D1
1.0
*
29.3
20
direct
Matrix IQ (Dallas)
1.8
*
27.7
13
direct
D2 (Janet)
14.3
0
24.5
10
cable modem
NTL
23.0
25?
15
ADSL
BT Home 500
46.8
5
22.6
5
ISDN
Demon
123.2
20
24.6
7

modem

Demon
141.8
30
25.1
11
modem
ISP2 (Loc1)
160.3
*
32.3
11
modem
ISP2 (Loc2)
average:
25.1

Notes : all times are from the perspective of a remote user / host.

First Hop Time - ping time to first off-site router. (This statistic is intended to reflect the speed of the Internet connection)

Long Ping Incidence - this statistics reflects the right hand tail of the ping time frequency distribution. (Set at times exceeding 2x median, but times up to 20x have been observed )

Incoming ADSL Transit - this is the average time taken to get to Cambashi's ADSL router from demon's incoming gateway. It should be similar for all connections

Inter-site hops - a measure of Internet "remoteness". The Cambashi ADSL service is provided by Demon, so dial-up connections from Demon may be expected to be shorter.

"*" & "25?" - very limited data

back to top