MOSIX is an on-line (operating system like) management system targeted
for High Performance Computing (HPC) on x86 Linux clusters,
multi-clusters and Clouds.
MOSIX supports both interactive concurrent processes and batch jobs.
It incorporates automatic resource discovery and dynamic workload
distribution, commonly found on single computers with multiple processors.
MOSIX is implemented as a software layer that allows applications
to run in remote nodes as if they run locally.
Users can start regular (sequential and parallel) applications on
one node, while MOSIX automatically seek resources and transparently
migrate processes to other nodes.
There is no need to modify or link applications with any library,
copy files or login to remote nodes, or even assign processes to
different nodes - it is all done automatically, just fork and forget.
Migrations are supervised by a comprehensive set of on-line algorithms
that monitor the state of the resources and attempt to improve the
overall performance by dynamic resource allocation, e.g. load-balancing.
A unique feature of MOSIX is that it operates on the process-level,
unlike other systems that operate on the job-level. This means that
the system adapts and redistributes the workload when the number
of processes of a job (and/or their demands) changes (using "fork" and
"exit"). This is especially useful for parallel jobs.
The latest version of MOSIX for Linux-2.6
can manage clusters and multi-clusters.
Flexible management allows owners of
clusters to share their computational resources, while still
preserving their autonomy to disconnect their clusters
at any time, without disrupting already running programs.
A MOSIX multi-cluster can extend indefinitely as long as there is trust
between the owners of its clusters. This must include guarantees
that guest applications will not be tampered with while running in
remote clusters and that no hostile computers can be connected
to the local network.
Since nowadays these requirements are standard within clusters and
intra-organizational multi-clusters, we recommend the use of MOSIX
in such cases.
MOSIX can run in native mode or in a Virtual Machine (VM).
In native mode, performance is better, but it requires
modifications to the base Linux kernel, whereas a VM can
run on top of any unmodified operating system that supports
virtualization, including Windows, Linux and OS-X.
MOSIX is most suitable for running HPC applications
with low to moderate amount of I/O. Tests of MOSIX show that the
performance of several such applications over a 1Gb/s campus multi-cluster
is nearly identical to that of a single cluster.
It is particularly suitable for:
Efficient utilization of system-wide resources- by
automatic resource discovery and load-balancing.
Running applications with unpredictable resource
requirements or run times.
Running long processes - which are automatically sent
to nodes in remote clusters and are migrated back when these nodes are
disconnected.
Combining nodes of different speeds - by migrating
processes among nodes based on their respective speeds, current
load and available memory.
Few examples:
Scientific applications - genomic, protein sequences, molecular
dynamics, quantum dynamics, nano-technology and other parallel HPC
applications.
Engineering applications - CFD, weather forecasting,
crash simulations, oil industry, ASIC design, pharmaceutical and other HPC
applications.
Financial modeling, rendering farms, compilation farms.
For further information see the related
papers
and the FAQ.