Ganglia (Software)

Ganglia is a scalable, distributed system monitor tool for high- performance computing systems such as clusters and grids. It allows the user to remotely access live or historical statistics query (for example, average CPU load or network utilization ) for all machines that are monitored.

Ganglia

Ganglia emerged from the Millennium Project of the University of California, Berkeley, and is based on a hierarchical design of cluster associations. It supports to monitor to a multicast -based lists / announce protocol to the status of the cluster. It creates a tree of point-to -point connections between representative cluster nodes and uses widespread technologies such as XML for data representation, XDR for compact, portable data transport, and RRDtool for data storage and visualization. It uses carefully engineered data structures and algorithms at very low per -node cost in order to achieve high concurrency. The implementation is robust and has been ported to an extensive list of operating systems and processor architectures. Ganglia is currently on over 500 clusters around the world in action. It was used to connect clusters in universities and throughout the world and can be operated in clusters with 2000 nodes.

The ganglionic system consists of two unique services, a PHP-based web front -end, and a few other small utilities.

Gmond is a multi-threaded daemon that runs on each to be monitored cluster nodes. For the installation, neither a common NFS file system or database backend, nor the application of specific user accounts or maintain configuration files is required.

Gmond has four main tasks:

Each gmond transmits information in two different ways:

• The node status is sent in unicast or multicast method in XDR format using UDP messages.
• Sending XML data is done via a TCP connection.

The composite in Ganglia is via a tree of point-to -point connections between representative cluster nodes to aggregate the multiple cluster nodes. At each node in the tree, works a Ganglia Meta Daemon ( gmetad ) periodically asked this a collection of child data sources, parses the collected data and exports them as aggregated XML over a TCP socket to the client. Data sources can be either gmond daemons that have a particular cluster or other gmetad daemons, which sets of clusters, representiern. The data sources use source IP addresses for access control and can be used for failover using multiple IP addresses. The latter possibility is the aggregation of data from clusters since each gmond daemon contains the entire state of the cluster.

Ganglia PHP Web front -end

The Ganglia web frontend provides a look at the collected data about dynamic web pages in real time. The web frontend shows the Ganglia data in a meaningful way for system administrators and computer users. The web front-end of Ganglia began as a simple HTML view of the XML structure and has become a colorful system that includes all the collected data have been developed.

The Ganglia web frontend is intended for system administrators and users. It displays, for example, the CPU utilization during the last hours, days, weeks, month or year. The web frontend shows similar curves for memory usage, disk usage, network statistics, the number of running processes and all other Ganglia metrics.

The web front end is dependent on the existence of the gmetad, it provides data from several Ganglia sources. The web frontend opens local port 8651 (default) and expects there a Ganglia XML tree, which is displayed on a comprehensive Web page, but requires that the full XML tree is parsed on every page. Therefore, the Ganglia web frontend should run on a fairly strong, dedicated machine if it presents a large amount of data.

The Ganglia web frontend has been developed in the PHP scripting language and uses graphs generated gmetad to display the history. It has been tested on many systems, Unix (mainly Linux) with the Apache web server and PHP 4.1.