Launched in June, Mutiny's Network Monitor is a rack-mounted 1U appliance designed to monitor server hardware and network infrastructure.

The system is easy to configure, and can be set up in a few hours. In its unclustered standalone mode, a single appliance can monitor up to 1,000 nodes – a node could be a server, router or switch – and can store more than 12 months of associated historical data.

The cost of licensing has the advantage of being based on the number of nodes, rather than number of ports. Mutiny quoted a figure of about £11,000 + VAT for monitoring a 500-node enterprise network. So the cost for monitoring even a medium-sized firm's infrastructure is relatively modest.

Network Monitor is an evolution of earlier Mutiny kit, first released in 2002 and predominantly intended to monitor servers running under Windows 2000 Server and Windows Server 2003, mainly on Dell and HP kit. Now, with Network Monitor, other network equipment can also be monitored.

The appliance we reviewed was based on a standard Supermicro hardware package similar to that used for other vendors' systems. For instance, we were also reviewing hardware management vendor Raritan's CommandCenter NOC 250 appliance at the same time, and this device uses similar hardware.

Mutiny staff came into IT Week Labs to set up the system as they would for other customers, installing the latest version of the software, which runs under a hardened version of Red Hat Fedora Core 4 with version 2.6.15 of the Linux kernel.

The onboard hardware consists of a CD-ROM, two USB ports and system status LEDs plus reset and start buttons on the front bezel. On the back are four 10/100Mbit/s network ports, another two USB ports and PS/2 ports for a mouse and keyboard, VGA adaptor, and a serial port for local access to the hardware.

Connecting to the appliance locally on the same subnet or remotely, we found it easy to call up device details to check critical server parameters, such as CPU load.

We could also easily drill down further to change threshold values for parameters such as memory usage, and Network Monitor let us define a warning value as well as a critical one. We could pull down detailed information on the specific processes running on our servers, such as process names, and process IDs (PIDs) together with CPU usage in seconds, and memory used by the process.

The new network interface monitoring functions in the system let us check network port input and output statistics for data rates, percentage of port capacity used, and the number of errors per minute. We could also pick up SNMP information from our interfaces, such as IP address, interface type, MAC address and link speed. For instance, we could interrogate our HP ProCurve 4108gl switch through Network Monitor and graph input and output port usage data over both pre-defined and user-defined time intervals.

Network Monitor also lets network administrators connect directly to a switch and use that switch's own web interface for management. This meant we could directly connect to our switch and, for instance, perform more advanced configuration by starting a Telnet session to access the switch's command-line interface. Network Monitor could also allow IT staff to monitor trunked links remotely and receive alerts if usage exceeds warning and critical thresholds.

Similarly, for server hardware we could graph memory usage, disk usage and CPU load for periods varying from the last six hours to the last six months. For disk usage the system can generate information for the percentage of the disk used, disk status, absolute total used, and free space available.

To keep servers up and running, information on CPU load, disk space available and system memory usage is critical. Network Monitor allows IT staff to configure warning and critical thresholds for variables such as memory usage to trigger email alerts in the event of breaches to be sent to a pre-defined SMTP server. It also allows SMS alerts to be sent via USB analogue and GPRS modems.

With Network Monitor, the data collected is now stored in a Java-based JRobin database rather than the MySQL database previously used in Mutiny's tools.

The appliance can monitor up to 1,000 nodes, but firms needing to monitor more or needing a degree of redundancy can set up systems in master/slave configurations with up to 1,000 appliances "slaved" to a master appliance.

Our minor criticisms of the system concerned a slightly loud cooling fan and the fact that it was easy to clutter up the web-based GUI with pop-up windows.