Hi faithful QWK.Mon testers,

I'm posting in here just in case anything I write in here borders on advertising. :)

First, if you aren't familiar with QWK.Mon, go take a peek at http://www.qwkmon.com so you know what I'm talking about.

As many of you know, premium service was rolled out a few days ago. The response by people willing to whip out their credit cards to support the service has been, shall we say, lukewarm. Perhaps it's a case of people putting things off until the 1st, when the "grace period" is over. But in any case, it's becoming clear that I was optimistic in my expectations of what percentage of users would go for subscription-based service.

So, that leads me to consider the only other method of recouping costs from non-paying users: advertising. I had really wanted to avoid it, but it looks like I'm not going to be able to.

The trouble, as any of you who haven't been living under a rock know, is that the advertising market is taking a serious beating. So, I'm not sure that even enough revenue would be available there to pay for free users. As I figure it, I could get at most about $1.00 a month per user if one "performance report" e-mail is sent out every day. That might cover the cost of providing free-level monitoring, if I'm lucky.

(Edited in light of new advertising rates posted in another message.)

So, I'm soliciting ideas. Any suggestions anybody can offer on forms of advertising they'd be willing to endure in exchange for free-level service would be appreciated. I really don't want to have to go to a pay-only service - there are enough of those out there. Also, if you are willing to pay for service, and just haven't done so yet, please let me know - that will also help me gauge where to proceed from here.

Thanks,



[Edited by Travis on 03-24-2001 at 05:33 AM]

Hi Travis,

Just a thought, but what if you were to remove the graphs and all the other goodies, and just offer the 'email when down' feature to those who are not paying? Just a thought...

I understand your problem - because people don't need to visit every day, banner adverts would be wasted. The only time people get an email is when there is a problem with their site, and I think they'd be headed for their webpage rather than a banner advert.

Unfortunately I come bearing no answers :(. I guess the top paragraph applies most.

Rich

Well, the real issue with server load isn't the front-end stuff, it's the monitoring. There are a finite number of hosts that can be monitored per minute per server. An occasional visit, even once a day, by a user just doesn't generate much server load. I can see turning off some of the interface features as "encouragement", but I'd hate to do that.

As far as ads... precisely. People won't, as a rule, visit the site that often, so banners are out. I've seen some "other" monitoring services that have big, fat, ad-ridden HTML pages that pass as down notifications. Who is going to be looking at an advertiser's site when their server is down? It's amazing that they can sell the space in those down notices.

Thanks for your thoughts.

How much load does the monitoring generate exactly? It seems to me that it shouldn't be particularly high... the cost of requesting a page via HTTP should be about the same (in fact, probably less) as serving a page, which would mean you should be able to provide 5-minutely monitoring of thousands and thousands of sites from even a small server.

Or am I missing something?

Would you put thousands and thousands of sites on a web server?

I didn't think so. :)

Originally posted by Travis
Would you put thousands and thousands of sites on a web server?

I didn't think so. :)


No... but most websites receive more than one visitor every five minutes.

Any decent webserver should be able to handle over a million hits per day. If you accept that the cost of monitoring websites is about the same as the cost of serving them (and again I'd say it is probably much lower, since you simply have to check that the reply starts "HTTP/1.1 200" and ignore the rest), a single server should be able to query each of 3500 websites every five minutes.

There is more to it than just checking the status of the web sites. That is actually probably only 20% of where the time is spent.

First, every time a monitoring cycle starts (once a minute, currently) a complex database query must be made to collect all sites that are due for their next run.

Next, a database query must be made for each site in turn to collect the necessary details to run a check. A monitoring process must then be launched for each site to be checked.

Each monitoring process must do the following, depending on the result of the check:

- on a successful request: do one update query to log the status in the hosts table, and one insert query to log latency and other statistics in the history table

- on an unsuccessful request: all of the above, plus two queries for every notification method defined, plus an insert query if it's a page that needs to be billed for, plus send a page out by the chosen method.

Multiply all that out, and you'll see that the machine isn't exactly sleeping.

Of course, operating on a switched 100mbit ethernet segment is a must, as you will quickly outrun the packet rate on anything less. With full 1500 byte ethernet frames to carry <100 byte TCP packets (times 6 to 8 for an average HTTP check), you need equipment that can process a *lot* of frames during check runs.


Do I need to further elaborate? I've been through the numbers many times to figure fair pricing, and dare I say, a little bit of profit for the time spent developing this. If I seem to have a bad attitude, I'm sorry - I just don't like being second-guessed on a project I've spent months on.



[Edited by Travis on 03-23-2001 at 07:03 AM]

Obviously, I can't speak for everyone here...but I am of the "waiting until the first" variety.

The service is great, and we intend to migrate to the new paying service before the end of the month, but we have been swamped in dedicated server orders for some reason this month, and I have not had the time :(.

From a hosting perspective, I've always thought that it would be nice to advertise on a monitoring service, especially if our uptime is as good as it has been :). However, I am not sure we would be willing to pay $1.00 a user. Right now, we are seeing some really good deals in advertising because advertisors are desperate, so I think (and I may be wrong, so don't quote me) your $1.00 per user might be optimistic.

Thanks for your comments.

Just so you know: the $1.00 per user figure was actually figuring a reasonable CPM rate for daily e-mails, multiplied out by a month. So we're actually talking about 30 some ad impressions.

Sorry Travis, I didn't mean to sound like I was second-guessing you... as I wrote, I couldn't see what was so time consuming and I wondered if I had missed something.

The fact that checking the web site status is only about 20% of the work is obviously what I missed. ;)

Understood, and you have my apologies if I was in any way "snippy." There's just a lot going on right now, and I'll be damned if I'm going to accept anything but resounding success for QWK.Mon. :)

Hah! I signed up for the pay service, with one day to spare :)! How efficient am I :).

Hey, nothing wrong with getting the most out of your money. :) Thanks!

I will soon be using your paid service but I belive there are some definite was to improve it. I would like a way to have a 5 minute interval but only monitor 2 servers since I can't forsee having 15 servers anytime soon. The business I or II plan are perfect except for the interval being so long. If I am going to pay I don't want to wait 20 minutes after my server goes down to find out.

Also how many times in a row does the server have to be down before an email notification is sent. I have the email setup and I show the site being down one time when it was checked but I never recieved an email.

I agree. How about a custom plan? 5 min notifications and 2 servers/hosts would be ideal for me.

JeremyL:

Thanks for your feedback. Some new plan offerings will be rolled out soon to better represent the range of offerings people might want.

If a site doesn't respond, the system will schedule a re-test in three minutes. If it doesn't respond then, notifications are sent.

teck:

Same thing; working on some more options for plans. Custom plans are a good idea, but the system isn't currently set up for them, and it would take a substantial amount of work. Not to say they won't happen, just not immediately. In the meantime, I'l try to come up with an array of plans that fit most usage profiles.