Best of the ISP-Lists

Fixed Wireless Business

DSL Shop Claims Real Reliability

It takes a lot of work, but you too can have only eight hours of downtime in seven years—and you cannot be blamed for the backhoe or the earthquake.

[June 27, 2005]

In a provocative post to the ISP-Wireless list in April, a DSL shop posted the following stats, challenging any WISP to match them:

8 hours of downtime in the last 7 years

1.5 megabits consistent download on our DSL. For those who wish top-of-the-line service, we offer a 3.5 Mbps connection for $59.95. And when I quote connection speeds, we provide that and more. The 3.5 connection is actually 3.8. The 1.5 is 1.8. And we deliver that.

No viruses ever. Not a single one has ever made it through our e-mail system in 11 years.

Virtually no spam (99.6 percent spam free)

Average length of time that we retain our customers is 3 years 7 months. This is in a college town where the average student is here for a little over two years. Take away the student clients and average length is a little over 6 years.

Challenged to explain these stats, the DSL shop owner explained:

...let's see if I can answer this all in one e-mail

Downtime stated was total system downtime. About 4.5 hours during the earthquake and about 3 hours when the main fiber bundle between our town and the rest of the world (yes there is really one fiber bundle for that) was cut. As far as user downtime because of CPE problem, that's a hard one. We've replaced 7 CPEs in the past 18 months. However, there are more problems than that. I think those problems are related to routing tables or some such in the CPE. When we have the customer reboot it, everything works fine. I've experienced this myself. I can usually go 4 or 5 weeks without a cpe reboot and then one day—wham. I have to reboot it. All the status lights say everything is ok but no routing takes place at all. So, this is a hard one to quantify. Mostly, we just tell the customer to reboot every week or two.

Now, once in a great while some of our outlying customers, those beyond the 4 copper-mile limit (using a DLC and ATM connections rather the PPOE) have definite problems. A DLC will go out or need to be rebooted. The last time that happened was on March 19th which was a pain because it took me an hour to get the phone company hopping and get things fixed. I mean, it was their customers being affected, too. We have maybe 1 problem every 6 months with our PPOE customers. In fact, with the exception of one fellow who has a grandson who "knows computers" and "helps" his grandfather with his DSL to make it faster, I can't remember the last time when we've had an issue with any PPOE customer beyond the CPE reboot.

Bandwidth—the local rural telephone company through whom our DSL runs has 3 DS3s. They provide me with bandwidth usage and availability stats. Unless they're lying through their teeth and unless my own tests on our DSL lines are somehow whacked, we've never not gotten less than 80 percent of our rated bandwidth on either our 1.5 or our 3.5. I have some random tests I run for file downloads, from 10meg to 4.5 gig files. Most of the time, I get the rated and greater.

Viruses—we don't let them through. We delete any attachments that are on Microsoft's suspect attachment list, except for .zip files. (I mean, we allow text files, pictures, pdfs, whatever). After that check, each e-mail is scanned by three separate servers—the incoming mail server, an intermediate server and the destination server. Each server has two virus programs on it—a combination of Trend, Clamwin and Norton plus one built into our destination MTA.

We used to advertise a "if you get an e-mail virus, you get your service free for that month" plan. Not a single person ever took us up on it. We make a big deal about "Your personal e-mail scanner, Norton, AVG, MCaffee or whatever you use, will never fire off when you check your e-mail." We got a nice scare in December when a customer called in to say that his e-mail scanner fired off when he checked his e-mail. After a little digging, he was still checking e-mail from his old account. Our record was still intact.

Again, we make a very, very big deal of this. Every customer is told before they sign up that they will never, ever see an e-mail virus. They don't. The only comments we get are "I can't believe it. I used to get e-mail viruse every day." BTW, I meet personally with 20 or so of my users every week and have personal contact with another hundred or so (the 20 are the same people, the hundred vary). Nearly everyone who walks into the store is asked if their Antivirus E-mail scanner every kicks off. The answer is always no. Deleting dangerous attachments and not relying on a single vendor for virus definitions is the key.

So, onto spam. I'm looking at the last 3 days of stats right now. 94 percent of the total e-mail message attempts were blocked outright. 3 percent of the attempts were whitelisted and 3 percent were allowed. Of the 6 percent that made it through the initial blacklist check, 50 percent, or 3 percent of the total, were flagged as possible spam using a content filter. The client has the option of quarantining those and creating their own rules to allow them through or simply just allowing them through. My spam quarantine is empty and I get more spam attempts than anyone except two of my users. Mine runs to 300 to 500 attempts per day. I've received 0 spam messages in the past 72 hours. I had one false positive in my quarantine (the content filter can be a little aggressive and it was an e-mail that mentioned pricing and something else...) and 0 false positives on the blacklists.

Do we get false positives? Yes. About 1 a week or so. We implemented Autosender Whitelisting recently and so that number keeps going down. I you send a mail message to someone, they're on your whitelist. We've implemented greylisting which allows me to analyze the requesting server to find out if it's an actual e-mail server or simply a zombie.

If it's a zombie, onto the blacklist it goes. If it's not, the second attempt by that mail server from that particular send to that particular user is allowed. We have our own DNSBL and DNSWL. So, when I claim 99.6 percent spam free, it really is. Actually, some users have made it 100 percent spam free. They've taken the content filter, told it to block everything except from those people on their whitelist and they're very happy. Not my choice but they can do with it as they will. Those messages are simply quarantined anyway so if they happen to block someone new, the message is quarantined and they can simply click a button that says "Never block messages from this person again."

Do I check with my customers about spam getting through? You bet. I have them call, e-mail, drop in...whatever it takes. The most common comment is "I get one or 2 a week." I think I can claim 99.6 percent rejection with no problem.

Back to false positives. When a message is rejected because of our blacklist protocol, the sender receives an NDR with an e-mail address and a web link. They can either e-mail that address or click on the link (which is what we prefer). We're notified that someone is requesting to be unblocked and we unblock them. The next iteration of this, due out this week, is to notify the recipient that someone wishes to be unblacklisted. If the recipient wishes to receive mail, they simply click on a link in that message and that user is added to their Autosender Whitelist.

We run a fairly extensive DNSWL so that yahoo groups and some other servers (AOL, MSN, Earthlink and some others) aren't blacklisted. They still have to go through the content/attachment filter though. And everything goes through the virus checking system.

I hope that's everything. My fingers hurt.

—End