Hello,
I am using an application which only works by connecting to a Linux Cluster with PBS scheduler. Before I can buy cluster credits, I want to test few things.
Is it possible to install some tools on my Ubuntu laptop, which can convert it into a temporary Linux cluster on localhost? Is there any tool available or a way to do it? What do I need?
Thank you for your time!
Dear Jitesh Doshi ,
Look the link, maybe useful.
https://support.ceci-hpc.be/doc/_contents/QuickStart/ConnectingToTheClusters/LaptopConfiguration.html
https://hpcc.usc.edu/getting-help/faq/cluster/
Regards,
Shafagat
I suggest to create two virtual machines, e. g. by dint of Oracle Virtualbox, and interconnect them with Corosync.
Thank you Shafagat Mahmudova and Aleksandar S. Sokolovski for valuable suggestions! Thanks Günter Fahrnberger , this is helpful. I will try that.
I suggest creating an environment with 3 virtual machines, Debian or Ubuntu, and install Kubernetes to set up a cluster in this architecture, one machine for the master node and the others for the secondary nodes. The virtual machine that will be the master node needs 1 Gb of RAM and 2 processors, the rest the configuration may be lower.
Peter Breuer , Thanks a lot for an insightful explanation, and sorry the question was not framed properly. Yes that was indeed what I was looking for - creating a cluster with just one machine.
This is what the application (Raccoon2) website says -
In this first release, Raccoon2 supports only Linux Cluster computational resources with PBS and SGE schedulers. The next version (currently in development) will provide support for other computational resources like the local workstation and Opal services.
Attached is a screenshot of application interface.
I tried with some basic server connections, but there was an error about incompatibility.
Before I could buy the cluster services for my calculations, I just wanted to try out the application locally, so virtual cluster is what I was looking for. Basically if I can emulate the cluster and scheduler environment on my own single machine, that was my question.
Your detailed answer gives me some idea, let me know what your thoughts are.
Hello,
if you are looking for a solution to test a Linux Cluster with PBS scheduler, I would recommend you to create it with Docker. Containers are lighters than VM and you can find a pre-configured container in Docker Hub like this one : https://hub.docker.com/r/pbspro/pbspro/
And you could lunch multiples instances of it in parallel to simulate a cluster.
Regards
While there were a few answers already (VirtualBox, Docker), I would suggest you also have a look at Multipass (https://multipass.run).
Its tagline is "A mini-cloud on your Mac or Windows workstation." But of course it can also be installed on Linux as far as I understand. I don't know about the PBS scheduler, but if that is a software that you install on the master of your linux cluster, it should probably just work on a virtual ubuntu machine.
Thanks a tonne Günter Fahrnberger and Luciano Aguiar Monteiro , I was able to do the job by creating two virtual machines (1 master and 1 node), and installing SGE scheduler on the master. Thank you so much for your time and valuable comments.
Jocelyn De Goër De Herve I did manage to get the work done with VM, but yes it was heavy. I'll explore the Docker option. Thank you for this option.
Jochen Ehnes Wow multipass is as simple as running any other command. Really appealing. I'll try my hands on that as well.
Thank you everyone in the thread for such support and valuable comments. Coming from bio background, I thought it would be very difficult, but all of your help is really useful and appreciable.
Hi,
basically, a cluster is made from a set of computers (physical/virtual). Researchers usually go for virtual node-based cluster unless otherwise there is no specific reason requiring physical node-based cluster (by considering the cost and compromising the performance & efficiency issues).
I am also a bit confused about "Linux" as a fundamental requirement to your cluster design. I assume that you have such a script that is written for Linux file system. Whatever the requirements you have for your cluster I would recommend you to have this with Docker Containerization technology. You could orchestrate your cluster(s) either with Docker Swarm or Kubernetes or anything else.
- Badges
- Science topic
- Similar topics
- Clinical Pharmacology
- Dosing