Website Monitoring in the E-Commerce Era
Intro
During the transition from the era of businesses running simple “brochure” websites to the emergence of full-fledged e-commerce storefronts, an interesting trend emerged. Web sites and web servers were going down, and the site owners were the last people to know about it. Customers, of course, discovered the problems right away.
As businesses became more and more reliant on their web sites to generate revenue and address customer service needs, the concept of web site and web server monitoring moved out of the realm of ultra techies-only and went mainstream. Unfortunately, many website monitoring solutions were only partially deployed because the people who were responsible for keeping the servers running had failed to consider certain essential web server “dependencies”.
What dependencies, you ask?
A web server is only part of the equation. There are many individual systems and processes that work together to keep a web site visible and running. If a web server monitoring system is not checking and monitoring all of these systems and subsystems, the customer could still experience a service failure and the site owners might not be aware until the complaints from users start coming in.
Let’s look at some of the most typical dependencies that can escape detection as the result of a partial deployment of a website and server monitoring strategy:
Most Internet servers run basic services including:
- HTTP Server
- DNS Server (Domain Name System)
- FTP Server (File Transfer Protocol)
- Database Server (Oracle, MS SQL, MySQL, Sybase, etc.)
- POP3 and SMTP Servers (Post Office Protocol or Simple Mail Transfer Protocol)
- Firewalls
- And more
If all your server monitoring software does is check to see if your home page is responding, it will not be aware of a service failure elsewhere on the server even though any of these failure points will most likely affect your overall service-level commitment.
In other words: Your server could still be in trouble even though it looks like your web site is “up”.
But even if you do configure your server monitoring software to test all of these services, you could still have a service failure and not know it. That’s because many of these services might appear to be functioning under one test scenario, but totally fail under another. And, if you are not running all possible test cases on your web server, then you can bet you will end up not knowing about a problem until it’s too late.
Your server may return an OK message when it retrieves your home page using an HTTP request, but a shortage of disk space, database problems, or some other situation might cause your shopping cart order form to fail when data is sent to it using the POST or GET command. And that’s just as bad as not having your server running at all.
The point is this: There are far too many critical points of failure between your customer’s web browser and your bottom line to run the risk of having web server problems and not be immediately informed.
That’s why the most robust and professional server monitoring software programs are configurable to test and respond to every dependency and any potential failure point including physical changes in the server’s actual environment.
Monitoring, of course, is just part of the solution. Notification and response strategies are just as important. Again, the best server and network monitoring software packages provide the ability to respond to some conditions and actually take steps to mitigate the problem without human interaction.
While network monitoring software that offers features like these are great for nearly any size company that relies on servers to operate their business, there is still a gap that exists for those companies who run specialized software and/or non-standard services. And for that group of businesses, the average network monitoring software program isn’t going to serve their needs.
So, if your organization fits that description, you’ll be best served by selecting a state-of-the-art web server monitoring software package that include “hooks” so your own IT people can integrate monitoring and reporting with your legacy and custom-designed software.
Even if your organization isn’t running those specialized programs which depend upon technologies like web services and XML, it’s just a matter of time until this new technology starts showing up on your network. This makes it all the more important to select a network and web server monitoring and reporting package that can handle everything you’ve got running now, and whatever is coming down the road in the future.
It takes more than a web server to run a company
As long as you are interested in making a server monitoring commitment, why stop at just monitoring your web server? Maybe you’re running Lotus Domino; maybe you have an assortment of Windows, *nix boxes running who-knows-what plus Netware. If you have the right monitoring and reporting package, none of that matters. It should be able to handle it all.
So, what are the most important features you should look for in a professional network monitoring package? There’s a laundry list full, but here are the most critical ones.
Your server monitoring program needs to have these monitoring capabilities:
- Support PING for ICMP/TCP and IPX
- Monitor TCP Services
- Monitor common UDP Services and provide the ability to define your own UDP-based checks.
- Perform DNS testing.
- Monitor Windows Servers for all common problems.
- Monitor web services and verify the status and content of web servers by testing all protocols.
- Be able to work through proxies and HTTP or NTLM password protection.
- Monitor database availability for Microsoft SQL server, Oracle, and MySQL.
- Monitor Netware availability.
- Provide support for SNMP checking and reporting including detail-level status checks.
- The ability to write your own checks and integrate them with your chosen monitoring package.
Monitoring is no good without a notification mechanism, and the ideal network server monitoring program should have a flexible array of options including:
- SMTP mail - With or without authentication and with fall-back to a 2nd mail server if needed.
- Message to any cell phone using a connected modem via SMS using the TAP or UCP protocol.
- Paging via the network using the SMPP or SNPP protocol a paging message.
- NotePager
- Instant Messaging alert via ICQ or MSN
- Popup message
And finally, your perfect network server monitoring program should be smart enough to mitigate the effects of certain threats or failures by taking automated or scripted actions including:
- (re)Start a service - Remote or locally
- Reboot a host
- Execute an application or batch command
- Wake On LAN - Send a magic packet to a remote system in order to wake it up.
- POST data to a web server
- SNMP trap - Send an SNMP trap to a remote management station
- Sound - Play a warning sound.
Of course, all of the monitoring and notification scenarios should be driven by a robust rules-based system that provides user-defined scheduling, alert thresholds, and responses.
Combine all of the above with a complete reporting package with exporting and graphical charting capabilities, and you’ve pretty much defined the dream network monitoring program.
Now all you’ve got to do is go find it.
Resources