Hardware configuration of PROOF

ROOT
1

Hi

I am new to PROOF and parallel computing. Currently, I am trying to setup a small proof-cluster with several computers. May I ask for some advice and some guideline on the hardware configuration ?

Thanks !

2

Hello,

PROOF was designed to adapt to the available hardware; depending on what you plan to do, some aspects may be more relevant.
If you plan to do data mining, access of the worker machines to data is likely to be the bottleneck, so better to have the data locally or well network-connected.
RAM can be an issue on the master machine during merging operations.

Whatever it is, we recommend to operate the cluster with Proof-On-Demand (pod.gsi.de) with the SSH plug-in: you just need to install public keys on the machines and decide which one will be master. We can provide examples for that.

G Ganis

3

Thanks for suggestions.!

Here is the brief description of my hardware configuration.

I plan to connect several computers through ethernet cable and switches. They can access to my home Wi-Fi hotspot.

4

Hi,

Just try and let us know of issues you encounter.

Btw, the most relevant measures are RAM and I/O bandwidth of wherever your input files come from.

Cheers, Axel.

5

Hi

Thanks for your replies! Actually, I’ve not started constructing the clusters.

Below is a schematics of the hardware configuration that I have in my mind.

Does it work?

Thanks!
Note 26 Aug, 2015.pdf (93.9 KB)

6

I hope you realise that PROOF is NOT a “general purpose batch system”.
In order to understand what it is good for, see, for example (search for the string “PROOF” inside):
“ROOT User’s Guide” → “Trees” → “Using TTree::MakeSelector”
“ROOT Primer” → “File I/O and Parallel Analysis” → “N-tuples in ROOT” → “For the advanced user: Processing trees with a selector script”
TSelector