Download OSCAR-Parallel Processing-Project Report and more Lecture notes Parallel Computing and Programming in PDF only on Docsity!
OSCAR
OSCAR (Open Source Cluster Application Resources) is a software package that is designed to simplify cluster installation. A collection of open source cluster software, OSCAR includes everything that is required for a dedicated, high-performance cluster.
Most of the software in OSCAR is available as standalone versions.
Installation
The installation of OSCAR must be done on an already working Linux machine.
Frontend Installation
- NIC must be configured
- Hard disk must have at least 2GB of free space on a partition with /tftpboot directory
- Another 2 GB for /var directory
- Copy all the RPMs that compose a standard Linux into /tftpboot
- These RPMs will be used to create the OS image that will be needed during the installation of cluster node
- Execute the program install_cluster
OSCAR
- This updates the server and configures by modifying file like /etc/hosts or /etc/exports
- It shows a graphical wizard that will help installation procedure
Node Installation i. It’s done in three Phases – Cluster definition, node installation and Cluster configuration ii. Cluster Definition – build the image with OS to install.
- Provide list of software packages to install
- Description of disk partitioning
- Node name, IP addresses netmask, subnet mask, default gateway etc.
- Collect all the Ethernet addresses and match them to their corresponding IP addresses – This is most time consuming part and GUI help us during this phase. iii. Node installation – Nodes are boot from network and are automatically installed and configured iv. Cluster Configuration – Frontend and node machines are configured to work together as cluster
OpenMosix is as turning a cluster of computers into a virtual SMP machine, with each node providing a CPU. OpenMosix is potentially much cheaper and scales much better than SMPs, but communication overhead is higher. (openMosix works with both single- processor systems and SMP systems.). OpenMosix is an example of single system image clustering ( SSI ) since each node in the cluster has a copy of a single operating system kernel.
To support process migration, OpenMosix divides processes into two parts or contexts: User context contains the program code, data, stack, etc and is the part that can migrate System context contains a description of resources the process is attached to and the kernel stack, does not migrate but remains on the UHN(Unique Home Node).
How is it used?
If a computer is part of an OpenMosix cluster: Firstly all processes will start running on a computer. With OpenMosix cluster, after a few seconds, processes will start to migrate from heavily loaded computer to other idle or less loaded computers in the cluster. If some of computers are slower than others, or some are otherwise loaded, OpenMosix will move the jobs around as best it can to balance the load. Once the cluster is set up, this is all done transparently by the system.
Not all processes migrate If a process only lasts a few seconds, it will not have the time to migrate.
Installation Overview
Basic steps: Download and install openMosix kernel patch Setup for new kernel startup on all nodes Cluster configuration setup Just setup a simple file or use the autodiscovery daemon
OpenMosix: possible configurations
Single-pool All the servers and workstations are used as a singlecluster: each machine is a part of the cluster and can migrateprocesses to each other existing node.
Server-pool Servers are a part of the cluster while workstations are no part of it: servers will accept other nodes’ tasks, while workstations will stay clean and no remote processes will migrate to it.
Adaptive-pool Servers are shared while workstations join or leave the cluster, depending on daytime
Physical Connection
Connect the main Head Node and the two Compute Nodes through Ethernet switch.
Installation Steps
Head Node
- Obtain a Fresh Copy of Latest Cluster software from official Rocks Website
- Boot from the DVD and type ‘frontend’ to begin the installation
- Configure the Ipv4 addresses
- Load the Packages e.g. we have installed ‘OS’ , ‘kernel’ and ‘base’
- Load the additional packages for monitoring the Cluster like Gandlia
- Choose Auto Partition and let the front end install
Compute Node
- Login the front node as root
- Open the Terminal and Enter ‘insert-ethers’
- Select Compute
- Head node will now wait for any compute node to send its DHCP request
- Turn On the Compute Node and
- Make sure Boot Priority is set as ‘Network Boot > CD/DVD > HDD’
- Boot it with the same DVD and just let it what it’s doing
- Main Node will soon receive a DHCP Request from the Compute Node
- Once this is done , we just wait for the compute node to install
- Once it reboots , we just logon to Ganglia and see the status of our Clusters
How does Rocks work?
Secret behind Rocks is the use of a mechanism called Rolls. So what is a Roll?
Rolls are the software components that make up a Rocks system. They provide mechanism for delivery of packages and configuration.They Define how we’re going to get our software onto a Rocks cluster.
A Roll consists of the following parts: Binaries - RPM format Configuration data Installation Map
Examples of a few Rolls are:
Sun Grid Engine Roll – Deploys SGE on all nodes and enabled job provisioning Bio Roll – Application Roll for Bio-informatics applications Xen Roll – Roll to deploy virtual machines Viz Roll – Roll that enabled building tile wall displays PVFS2 Roll – Installs and Configures the Parallel Virtual File System on your cluster
They are distributed as an ISO file and can be downloaded from the website.
Selection Criteria
This section will try to justify the selection of Rocks Cluster among OpenMosix and Oscar. We have based our selection on a few criteria’s which aret:
Freeware Versatile Customizable Well Documented User Friendly installation Available help across web Scalability
Taking all the above factors into consideration, the Team decided to use the Rocks Software. We hope to work on other tools as well in the summer.
