We would have preferred to entitle this
blog, "Choosing a reliable source for uptime statistics, and
learning why you should avoid free monitoring services", but
that would have taken up too much space. The purpose of this article
is to discuss how working with a reputable uptime monitoring service
can benefit both a hosting company and its customers, and likewise
how inaccurately reported statistics can be equally damaging.
In 2006, it was estimated there were
more than 6,000 web hosting companies in existence hosting more than
45,000,000 active domains across the globe. As one of these
companies, Canvas Dreams LLC is always working to become a better
host and improve the quality and diversity of services for our
customers. One way we can do this is through consistently providing
reliable, stable, fast access to all of the web sites we host.
What Is "Uptime"?
The single most important concern that
potential customers raise when considering a switch to us has to do
with our "uptime", which is the accessibility of our
servers and networks (and hence, the web sites we host) as measured
over the course of a given month.
When a web server drops offline and
becomes inaccessible, so do the web sites and email services
supported on that server. Any amount of this "downtime" can
be detrimental to a company whose business depends on its web site
and email services, so it is crucial that the hosting provider offer
the most stable and reliable hosting environment possible.
There are times when a short amount of
downtime is acceptable, such as for the occasional reboot of a web
server. Downtime can also be necessary to perform physical
maintenance on the server, such as installing new RAM or hard drives,
or even new hosting software that will improve the overall level of
service provided by that server. In any case, downtime is something
every hosting provider aims to avoid.
To help ensure the satisfaction of
their customers, hosting providers aim to offer the greatest uptime
possible. As an industry standard, 99.9% or better uptime, which
amounts to less than 44 total minutes of downtime per calendar month,
is a common goal among many providers. Over the past year, Canvas
Dreams' network uptime has averaged better than 99.991%, which is a total of 48 minutes of downtime, and includes times when we were performing
routine maintenance on a router, switch, or web server (and which is
typically done in the middle of the night so the least number of
customers possible are affected).
We're quite pleased to be able to offer
our customers this level of service, because it shows that the
hardware, staff, and knowledge we have used in building our network
is performing better than the industry standard. The result is
simple: the web sites and email services we host are accessible
99.991% of the time, meaning our customers' online businesses can
perform at their best and with almost no interruption.
How We Measure Uptime
The most reputable means to measure the
uptime of our network is to do so from outside our network, in a
"real world" setting such as what one of our customers
might experience in trying to access our systems. This can involve
using a third-party commercial monitoring service such as
WebSitePulse, or using a standalone tool such as Nagios hosted from a
secure location on the corporate extranet.
For most every case, if the monitoring
service has a stable Internet connection and uses standard methods
for determining network accessibility (such as pinging various ports
or checking the status of a web page), the monitoring service will
provide reliable results. Depending on the amount of time it takes to
receive a response, and the severity of any perceived slowdown, the
service can automatically alert a member of our staff -- through
email and text messaging -- to any problems it has encountered during
a scan.
For a number of reasons, namely
security, we cannot describe exactly how we monitor the uptime of our
network. However, we can say that three servers at different
locations across the globe are used to constantly monitor every piece
of hardware within our networks in Portland and Edinburgh. The
monitoring systems are so sensitive that even a 30-second restart of
a server can register an outage alert. And yet, these same, sensitive
monitors have told us that Canvas Dreams' network has had an average
99.991% uptime over the past year.
The Downside Of Free Uptime Monitoring
Services
At the same time, certain "free"
monitoring services claim to offer accurate uptime reports for
hosting companies they independently monitor. One such provider is a
web site called WebHostingStuff. They claim to list the greatest
number of web hosting providers around, and provide uptime reports
for each -- even if the hosting provider didn't ask to be included on
their site. Their domain is registered to an individual located in
Singapore, with the service itself being hosted at a datacenter in
Dallas, Texas.
After learning that our company had
become listed on their web site, we agreed to trial their service,
and for a while, it did seem to work all right. Last fall, when we
experienced two minor outages from performing scheduled maintenance on
one of our servers, sure enough, WebHostingStuff reported the outages
almost to the exact minute. From this early test, we agreed to remain
listed on their site and even add certain features to our web site,
such as a dynamic image showing a live uptime report provided by
them.
However, at the beginning of March 2007
their monitoring service began reporting all sorts of outages for our
network that simply did not exist. Our own monitoring infrastructure
reported none of the outages for the server in question. Though our
equipment was operating just fine, WebHostingStuff continued
reporting sporadic outages. We sent repeated requests that they
correct this problem, or remove us from their web site entirely,
since these reports could provide the wrong impression to potential
customers that we were offline, when in fact we were up and running
just fine.
One of the immediate remedies offered
by their FAQ was to log into our account on their web site and click
on a "Reset" button that would zero out the uptime
statistics they had archived since they began monitoring us last
fall. This means any host can simply erase undesired statistics and
then begin reporting an immediate 100% uptime since the reset. To us,
this was the first sign that their service might not be as objective
as we once thought.
It then came to our attention that they
were not monitoring our systems in the way traditional services do.
None of the ordinary ports (which we constantly track from our other
monitoring services) are being scanned by WebHostingStuff to report
their uptime results. We further came to learn that WebHostingStuff's
monitors don't even check to see if core services are running on a
server (which can mean they could report a server being online when
it is in fact not serving web content). Finally, we learned their
service can be tricked into reporting a 100% uptime through an
automated "daemon" that simply responds to pings from
WebHostingStuff. We have noted the sources for all of these
discoveries below for your reference.
After multiple attempts at contacting
them, we received no response from a customer service or technical
support representative. Our profile is still listed on their site,
including erroneous uptime figures. In fact, since we tried using the
Reset feature, they have begun showing our statistics all the way
back to the beginning of last fall, but without a single minute of
downtime -- even including the two times we admittedly were working
on the server in question. It's as if the server was never offline
during that time period. More than anything else, their ability to
reset all of the data, even to inaccurately reflect uptime for our
network when it should have been listed as offline, indicated to us
that uptime reports from WebHostingStuff could not be trusted.
In a final attempt to make contact, we
looked up the official, publicly-listed WHOIS record for their
domain, and learned that the email address listed for Michael Low,
the domain's administrator, is trashcan@w3centric.com. It would
appear Mr. Low does not want to be contacted for any reason. It
further demonstrates to us why they would not return our emails to
them.
As of June 14, 2007, they have gone so
far as to begin blocking our office from accessing their web site. A
quick search on Google indicates that we are not alone, and that
other hosting companies have run into the same problem. Sending
emails and asking that data be corrected has resulted in
WebHostingStuff blocking those companies from accessing the web site.
As a result, we would ask customers and
potential customers alike to take this all into consideration if they
choose to view our profile on that web site. One more thing. We have
blocked their entire network from accessing ours. So if you note an
inordinately high amount of downtime for June 2007 and beyond, you
now know the reason why.
The following links contain more
information on WebHostingStuff:
Summary: The Value Of Uptime Monitoring
The vast majority of monitoring
services provide a real and valuable tool to help ensure the quality
of every web hosting in service today. Monitoring helps customers and
hosts alike. While they can inform customers about their host's
reliability, monitoring also helps web hosts identify and correct
weaknesses in their network. As a diagnostic tool, uptime monitors
can greatly assist a host improve their service to all of their
customers, as well as their customers' customers.
To learn more about monitoring
services, how they work, and what options are available, simply do a
search on Google for "uptime monitoring", or go to the
following web sites.
