Condor is a high throughput system, scheduling and providing large amounts of computational power over a long period of time. It provides the efficient use of a large variety of systems, from idle desktop workstations and dedicated clusters to grid systems all over the world, while its incredibly flexible configuration implements and maintains the machine owner's desired policy for the machine's availability.
LAM/MPI is an implementation of the Message Passing Interface (MPI) parallel standard that is especially friendly to clusters. It includes a persistent runtime environment for parallel programs, support for all of MPI-1, and a good chunk of MPI-2, such as all of the dynamic functions, one-way communication, C++ bindings, and MPI-IO.
Heartbeat is a full-function high-availability system for Linux and other POSIX-like OSes. It monitors services and restarts them on errors. When managing a cluster (more than 1 machine), it will also monitor the members of the cluster and begin recovery of lost services in less than a second. It runs over serial ports and UDP broadcast/multicast, as well as OpenAIS multicast. It is easily adapted to different interconnect media and protocols. When used in a cluster, it can operate using shared disks, data replication, or no data sharing. Versions starting with 2.0 are comparable to any commercial HA package, providing resource monitoring, larger clusters, and detailed dependency information.
distcc is a parallel build system that distributes compilation of C/C++/ObjC code across machines on a network. It can be set up in just a few minutes and makes builds up to ten times faster. It does not require machines to share a filesystem or have the same libraries or header files, and installation does not need superuser privileges.
radmind is a suite of Unix command-line tools and a server designed to remotely administer the file systems of multiple Unix machines. At its core, radmind operates as a tripwire. It is able to detect changes to any managed filesystem object, e.g. files, directories, links, etc. However, radmind goes further than just integrity checking: once a change is detected, radmind can optionally reverse the change. Each managed machine may have its own loadset composed of multiple, layered overloads. This allows, for example, the operating system to be described separately from applications. Loadsets are stored on a remote server. By updating a loadset on the server, changes can be pushed to managed machines.
BIE (Business Integration Engine) is an open source integration system that makes it easy for organizations to exchange data with external trading partners regardless of their native applications. It competes in the same space as applications like Microsoft BizTalk except that it is truly cross-platform since it written in Java.
Drqueue is a distributed rendering queue. It helps manage jobs on a render farm. It works under Linux, Mac OS X, FreeBSD, IRIX, and Windows, supports Maya, Blender, Lightwave, Mental Ray, Bmrt (Renderman), 3Delight, and Pixie out of the box, and can support any other renderer or computing task that can be handled via scripting.
GNU VPE creates a virtual ethernet by creating host-to-host tunnels between multiple endpoints. Unlike other virtual private "network" solutions which merely create a single tunnel, it creates a real network with multiple endpoints. It is designed to be conceptually simple and straightforward to setup (assuming prior IP routing knowledge) without sacrificing flexibility. It is designed to sit on the gateway machines of company branches to connect them, but it can also be used to tunnel into a VPN with a variety of protocols (RAW IP, ICMP, UDP, TCP, HTTPS-Proxy, DNS).