IEPM

Requirements for PingER Monitoring Sites

SLAC Home Page

Monitor host requirements

The PingER monitoring site provides monitoring host at the site. This avoids ownership, maintenance, security, and access concerns. The best place for the host running PingER is close to the border, e.g. outside the DMZ. However, this is not necessary since most of the measurements are over the WAN the extra delays added by the LAN are minimal. Many PingER monitoring machines sit inside the LAN at the site.

The host should run Linux. There are examples of hosts not running Linux, however, they are not supported. In addition you'll need a Linux web server (e.g. running Apache) with perl, make, dig, mail, lynx, ping, ping6, traceroute, traceroute6, wget, crontab and apache installed. Do a google search for example on 'intall wget on centos' to install missing items. Please disable SELinux if you're running it.

Also the host must be publicly accessible from the Internet, and registered in the domain name service (DNS).

Since only pings are sent and processed the cpu requirements are very minimal. A 500MHz or better PC with at least 128MB RAM, 2GB of disk space, a 10/100Mbps NIC should be quite adequate. Reliability is more important. For some sites UPS may be desirable.

The cheapest, smallest host we ever installed was one for the ePingER project. It has also been installed on a Virtual machine, see here.

Reverse Traceroute

It is also very valuable if you can run a reverse traceroute server at your site, for more on reverse traceroute servers see Reverse Traceroute Servers

Data Collection

You also need to make the data collected from the monitor machine available for the archive sites to gather. This is usually done via HTTP, if so you will need a web server with access to the data.

Security Implications

One security implication is that ICMP echo and reply must not be blocked or rate limited to this machine. There is no passive capturing of packets. Also the data collected by the monitor has to be made available, most sites do this by making the data available from a site web server. A few sites use FTP, but this is less desirable.

The software uses the standard ping facility by default. It runs at the user level.

You need to open TCP port 80 (for web server), UDP ports 33434 - 33465 (outbound) for traceroute, ICMP echo request and echo response (for ping).

Unix/Linux system files sometimes need to be added

The following are needed by perl, they are usually pre-installed, if not:

Effort Required

The effort is probably about 0.5 to 1.5 day to initially install the PingER Tools Following this there is the occasional request to follow up on some anomaly usually to do with being unable to collect the data. Also maybe once a year the list of beacon sites needs updating. Further effort will be needed to select appropriate hosts/sites to be monitored and to set up and maintain the local hosts configuration file. There is a list of "Beacon" sites that all monitoring hosts are supposed to monitor for consitency (each monitoring site is free to and expected to choose its own extra set of hosts that it is interested in monitoring). The amount of effort will depend on the number of hosts/sites monitored and the frequency with which changes are desired/needed. Typically this is like an hour a week after things are set up.

Alternatively, if you wish to give us an account with ssh access on an approriate machine, then we can install and maintain it for you.

PingER Diskspace

A typical PingER monitoring site can expect to see resource usage as described on our PingER Usage spreadsheet.

Network Overhead

Network overhead is about 100 Bits/sec/pair monitored.

Availability of Reduced Data

The reduced data is available publicly at: http://www-iepm.slac.stanford.edu/cgi-wrap/pingtable.pl

Revised August 22, 2010.
URL: http://www-iepm.slac.stanford.edu/pinger/mon-req.html
Comments to Les Cottrell.