Uptime Truth


#1

There are a lot of opinions posted about uptime, especially when there are problems. I decided to check out the facts by subscribing to a third party tracking service that checks my sites once every 30 minutes. I started in April, so I only have two months, but I think they’re revealing. And they ARE fact. May uptime: 99.74% June uptime 99.97% Total downtime for the 2 months: 36 minutes 8 seconds.
Those are the FACTS; my OPINION is DH is doing a pretty good job. I suggest you check it out for yourself.


#2

Sorry jvigliano, but check every 30 minutes is not really check.
1 month has 43,200 minutes. Checking every 30 minutes, you are checking only 1,440 minutes/month, which is just 4% of your time. The only shure you have, is that your site was 99.97% uptime on 4% of june.

In 30 minutes, your server can goes down many times.

Please, contract a service that check every minute and you´ll see that your uptime is not that good.

If you don´t want to pay, there many trials and free service out there. I suggest http://basicstate.com/


#3

While your point about the limited sampling that a check “every 30 minutes” produces has some validity, you have no way of knowing whether more frequent checking would result in his uptime being less or being even higher.

DreamHost has many servers, and users experience different uptime results, His numbers may not have been derived from as broad a sampling as you would like, but at least he has a statistical basis for his report - and you do not have any such numbers for his site(s).

Reporting your sites’ numbers is one thing; assuming you can predict his sites numbers is pretty ignorant. :wink:

–rlparker


#4

You´re right. Perhaps his site/server is better than mine.
But keep my suggestion about check every minute. You may be surprised, or not.


#5

ALL uptime monitoring services are useless because all they do is check to see if a site is up from their testing locations. If there is a problem with one of the dozen or so (on average) hops between the testing location and the server, it will be reported incorrectly.

The only thing such services are good for is monitoring the relative level of accessibility when measured against a large sample of other hosting services. Also, having such a service checking your uptime is going to have an enormous effect on your traffic, hideously skewing your reported statistics and eating into your bandwidth allocation.


si-blog | Keystone Websites
Save $97 on yearly plans with promo code [color=#CC0000]SCJESSEY97[/color]


#6

Mathematically, checking more frequently will just decrease the amount of clock time necessary before you’ve gathered enough samples for your result to be statistically significant. In other words, the thing that’s important isn’t how frequently you check, but the absolute number of times that you’ve checked. So in a given day, you will be able to gather 1440 samples while the OP will only have been able to gather 48 samples. If you’re interested, we could go through an exercise of calculating the margin for error on the uptime statistics over various time periods using both 1-minute and 30-minute sampling at various uptime levels.

Again, sampling every 30 minutes just means that it will take the OP 30 times as long to get as good a baseline as you, given that you’re sampling every minute.

(Though actually, this is a worst case because it actually assumes that outage events are instantaneous. If outage events last an hour, sampling every minute and sampling every 30 minutes will actually require a similar number of calendar days before reaching statistical significance.)

Free unique IP and $67 off with promo code [color=#CC0000]FLENSFREEIP67[/color] or use [color=#CC0000]FLENS97[/color] for $97 off. Click here for more options


#7

Thanks, I´m really interested in see how that works. Cause IMHO checking in a 1 minute or 2 minutes resolution, will give you more change to catch a reboot, or a short server load.

Let´s say that you check your site every 30 minutes like this:
0:00, 0:30, 1:00, 1:30. …
And suppose that your server goes down during 20 minutes, from 0:05 to 0:25, 0:35 to 0:55…
This way, you will have a false 100% uptime, don´t you agree?
And since the uptime guarantee is measured in %/month, check each 30 minutes is not enough resolution.


#8

What “uptime guarantee”?


Chips N Cheese - Custom PHP installs and the like!


#9

As far I know, DreamHost does not have uptime guarantee. I just mean the 99,9% guarantee that we see over the net.


#10

Sorry, I didn’t mean to get bogged down in a mathematical discussion.
Continuous monitoring would be the best assurance of uptime, but that’s not available. So we make compromises. And sure, the server could go down for 29 minutes every half hour, just missing the pings from the monitoring service. But that’s a bit unlikely to be consistent over the long term. I usually get emails from vistors when the site is unavailable.
I have not noticed huge bandwidth issues and my web page tracking service doesn’t record the pings as page loads, so it doesn’t skew my results.
Listen–this is inexpensive hosting. If I had “mission critical” needs, I hardly think I’d be relying on a $10/month solution. My desire for the monitoring service is to notify and record major outages–in MY determination, 30 minutes or more. Will I miss some shorter ones? Yes. But again, over the long term, not significant for my purposes. Will I get some “false positives”? Doesn’t appear so, since one link down on one path when several paths from several different locations are being used won’t result in a “site down” message. And again, if I get a false positive every now and again, I’m not rushing to my computer looking for another host.
As I said in the original post, I decided not to rely on the stated opinions of others, especially in the heat of a major downtime issue, to determine whether to stay or cut bait and find another host. I wanted factual information–relevant to my needs. And when I combine the monitoring service as I’ve set it up along with some common sense, I feel I’m getting value for what I’m paying. Just one man’s opinion based on one man’s experience.


#11

Ahhhh okay. I thought you were implying DreamHost had an uptime guarantee for a moment there, heh.

Personally, I’d say that 1min intervals are a bit extreme, though certainly not unheard of. I consider 5-10min intervals to better represent a realistic “uptime” percentage, as network errors from both sides do occur at times, and thus bogus results would be seen more frequently during a period of 1min checking. As Lensman mentioned, the difference between short and long interval checkings, is that they both require the same amount of samples to be even remotely accurate.


Chips N Cheese - Custom PHP installs and the like!


#12

“Uptime guarantee” An interesting term. Check out Hostgater’s “guarantee” for LOL. Like I said, if your uptime is that critical, you need to pay for service that does include performance objectives and penalties if not met. And if you find that for $10 a month, let us all know. :slight_smile:


#13

Here is what I have since April, checked every minute.

2007-04: 100%
2007-05: 99.79% (Downtime: 1 hour 27 min)
2007-06: 99.69% (Downtime: 1 hour 53 min)

Could be better, but it’s fine if you don’t run a business.


Promo-Code: [color=#CC0000] SAVEMONEY97 [/color] - Save [color=#CC0000]$97[/color] on your first year of hosting.


#14

Hey, it could almost always be better, but that is not bad, and is probably fine for most businesses (though, YMMV!) :wink:

–rlparker


#15

Every one minute checks can be pretty hard on a server if you’re going to a dynamic page. Keep that in mind when DH cuts you off.

And make sure that your monitoring isn’t more than 50% of your traffic…

Wholly - Use promo code WhollyMindless for full 97$ credit. Let me know if you want something else!


#16

If 1 hit/minute gets you cut off by DH, you have other problems. Really.


#17

One question regarding this topic, what is a good site to have our site monitored ?

Jonathan

Get 200GB[L1 Plan], $50, PLUS 1 additional Free Domain with [color=#CC0000]50AND1DOMAIN[/color] promo code. Sign Up NOW


#18
  1. Are you willing to pay for better service or do you want something free?
  2. Where are most of your users located?

BTW, next time, you should start your own thread rather than force everyone to read and navigate through the 16 other posts in the existing thread. In fact, you might start a new thread and tell everyone to stop posting in this one. :slight_smile:

Free unique IP and $67 off with promo code [color=#CC0000]FLENSFREEIP67[/color] or use [color=#CC0000]FLENS97[/color] for $97 off. Click here for more options


#19

[quote]1. Are you willing to pay for better service or do you want something free?
2. Where are most of your users located?[/quote]

  1. So, the free one is not a good service ? Ok, both of them if I may ?
  2. My targeted users are China, South East Asia, and US.

Thanks for your quick feedback.

Jonathan

Get 200GB[L1 Plan], $50, PLUS 1 additional Free Domain with [color=#CC0000]50AND1DOMAIN[/color] promo code. Sign Up NOW