My colleague in France, Bertrand Gugger, has been hard at work implementing SMS alert-sending capabilities for the Net_Monitor package. This functionality depends upon the nacient Net_SMS package, so we will be looking for ways to create end-to-end (from interaction with the gateway to display on the device) testing before release. It will be interesting to see how SMS providers respond, since this software could provide a boost to their offerings as system administrators adopt Net_Monitor for monitoring systems. Will the providers get it, and grant us test accounts? We shall see.
-
About
Robert Peake's poems have appeared in Iota, North American Review, Rattle, and others.
4 Comments
I’ve been doing testing with your net_monitor package since version 0.0.8 and am very interested in it. It is good to hear about this upcoming feature of SMS.
I have one question though, what is the estimate number of servers that this package can monitor per minute?
Thank you and keep up the good work
Hi Mans,
That is such a configuration-dependent question that it is very difficult to answer. For example, unlike many monitoring packages Net_Monitor actually makes a full HTTP/FTP/DNS connection and, in the case of HTTP, returns the status code. PHP prior to 4.2.0 doesn’t handle socket connections as well as later versions, so that is a big factor in performance. Also, the capabilities of the hardware and the network it is on play a role.
Net_Monitor has been optimized through many helpful suggestions from Pear developers. So, the footprint of the package is fairly small. It depends on other Pear packages that implement network service clients, which are also fairly lean. So, it is difficult to give a precise number but I can say the software itself should be fairly efficient.
The truth is that if you are launching a tremendous number of connections at once, it will simply take longer, since connections are currently not threaded. How much longer is acceptable depends on your polling interval. To create simultaneous connections, simply break down your polling into different groups run by separate instances of PHP on separate scripts. This approach should scale up close to whatever your hardware can handle in terms of simultaneous open sockets.
If you do get some kind of numeric benchmark for your hardware/software configuration, I’d be happy to hear about it.
Hi Robert,
I can understand you. I scanned the code and it cannot be more professional than what you have done. Actually, I did not run it on directly connected machine to the internet (all our company’s PCs are behind proxy!) So, it is difficult to give estimates from my part. But I have one suggestion, would not be faster if you use hosting system’s ping program through exec() or something similar?
This is a question of using PHP to call a system-native function, then string mangling the output to see if the result is OK, versus using the native PHP sockets mechanisms to make the same type of request. Net_Monitor does not deal with ping. Net_Monitor aims to tell you a lot more than what a simple ping tells you. Ping can tell you if a server is alive or dead, but not about the various services that are (or are not!) responding on that machine. So, it’s not really an apples-to-apples comparison.