Skip to main content

Unpacking a GPU computation server...Leviathan unleashed

What / why?
As you might know, EMBL-EBI has a very powerful cluster. Yet some time ago we were running into some limitations and were pondering on how great it would be if we had the ability to run more concurrent threads in a single machine (avoiding the bottleneck that inevitably appears on the network for some jobs).

It turns out there is an answer, namely in the form of a GPU (graphics processing unit). This is the same type of chip that creates 3D graphics for games in your home PC / laptop. While the capabilities of individual calculation cores are relatively limited on GPUs compared to CPUs, they can have a massive amount of them in order to generate 3D environments at the speeds of 60 frames per second. Schematically it looks like this (CPU left, GPU right):

As you can see, the CPU can handle 8 threads concurrently, whereas the GPU can handle 2880 (see also this great youtube video by the myth busters). We have all kinds of ideas of calculations we want to run on the GPUs (that have shown to work well in MD), but now first ... the geek tradition that is unboxing!

The guys at Nvidia were very generous and provided us with 5 GPUs (thanks to Mark Berger and Timothy Lanfaer). Tim was also very quick with technical questions concerning the hardware specs needed and software troubleshooting. Thanks again!

At the EMBL-EBI people typically work with laptops or thin clients, and the cluster consists of blades so there was no place to put our GPUs. Yet, after a quick investigation we had a list of hardware we wanted and a big box was delivered two weeks ago !

Time to unpack...

So after opening and removing the hardware, we had a tower / 4u rackmountable chasis

Next up, placement of the GPUs inside the chassis:

Some tinkering was in order:

And finally we could boot and install the OS. We choose Ubuntu 12.04 LTS because of the stability, and availability many packages (with source code). 

Just one question remains, why 'Leviathan'? 

Given the availability of python based cuda packages, we will probably start there. Hence our server we be a very powerful incarnation of python, and what's more awe-inspiring than the Leviathan?

CUDA running
After some trouble getting the drivers to work (we use Ubuntu 12.04 LTS), Michal got everything up and running!

Potential projects
Some of the projects we will be starting with are CUDA based random forests, similarity matrix calculations, and compound clustering. If you have a good idea and would like to collaborate and co-publish, please contact us via email!

The server contains the following hardware:
Case: Supermicro GPU tower/4U server 
PSU: 1,620W Redundant PSU
CPUs: 2*Intel Xeon E5-2603 1.8GHz 4core 
RAM:  8*8GB Reg ECC DDR3 1600MHz 
Disk: 1*2TB 3.5” SATA HDD
GPUs: 1*Tesla K40; 2*Tesla K20 (one extra to be added later)

Michal & Gerard


Chris said…
Why CUDA rather than OpenCL?
Unknown said…
At the moment we had to make a choice and CUDA seemed a better place to get started as newbies (motivated by the fact that we have some concrete applications which we can start quickly with using CUDA).

(Also CUDA exposes more hardware and runtime info)

However, we should also consider a potential switch (we have now installed 5.5 with 6.0 about to be released so in any case an upgrade / change will be due in a few months).

In the mean time we would like to get started.

Popular posts from this blog

UniChem 2.0

UniChem new beta interface and web services We are excited to announce that our UniChem beta site will become the default one on the 11th of May. The new system will allow us to better maintain UniChem and to bring new functionality in a more sustainable way. The current interface and web services will still be reachable for a period of time at . In addition to it, the most popular legacy REST endpoints will also remain implemented in the new web services: Some downtime is expected during the swap.  What's new? UniChem’s current API and web application is implemented with a framework version that’s not maintained and the cost of updating it surpasses the cost of rebuilding it. In order to improve stability, security, and support the implementation and fast delivery of new features, we have decided to revamp our user-facing systems using the latest version of widely used and maintained frameworks, i

A python client for accessing ChEMBL web services

Motivation The CheMBL Web Services provide simple reliable programmatic access to the data stored in ChEMBL database. RESTful API approaches are quite easy to master in most languages but still require writing a few lines of code. Additionally, it can be a challenging task to write a nontrivial application using REST without any examples. These factors were the motivation for us to write a small client library for accessing web services from Python. Why Python? We choose this language because Python has become extremely popular (and still growing in use) in scientific applications; there are several Open Source chemical toolkits available in this language, and so the wealth of ChEMBL resources and functionality of those toolkits can be easily combined. Moreover, Python is a very web-friendly language and we wanted to show how easy complex resource acquisition can be expressed in Python. Reinventing the wheel? There are already some libraries providing access to ChEMBL d

LSH-based similarity search in MongoDB is faster than postgres cartridge.

TL;DR: In his excellent blog post , Matt Swain described the implementation of compound similarity searches in MongoDB . Unfortunately, Matt's approach had suboptimal ( polynomial ) time complexity with respect to decreasing similarity thresholds, which renders unsuitable for production environments. In this article, we improve on the method by enhancing it with Locality Sensitive Hashing algorithm, which significantly reduces query time and outperforms RDKit PostgreSQL cartridge . myChEMBL 21 - NoSQL edition    Given that NoSQL technologies applied to computational chemistry and cheminformatics are gaining traction and popularity, we decided to include a taster in future myChEMBL releases. Two especially appealing technologies are Neo4j and MongoDB . The former is a graph database and the latter is a BSON document storage. We would like to provide IPython notebook -based tutorials explaining how to use this software to deal with common cheminformatics p

ChEMBL 30 released

  We are pleased to announce the release of ChEMBL 30. This version of the database, prepared on 22/02/2022 contains: 2,786,911 compound records 2,157,379 compounds (of which 2,136,187 have mol files) 19,286,751 activities 1,458,215 assays 14,855 targets 84,092 documents Data can be downloaded from the ChEMBL FTP site: Please see ChEMBL_30 release notes for full details of all changes in this release: New Deposited Datasets EUbOPEN Chemogenomic Library (src_id = 55, ChEMBL Document ID CHEMBL4689842):   The EUbOPEN consortium is an Innovative Medicines Initiative (IMI) funded project to enable and unlock biology in the open. The aims of the project are to assemble an open access chemogenomic library comprising about 5,000 well annotated compounds covering roughly 1,000 different proteins, to synthesize at least

Multi-task neural network on ChEMBL with PyTorch 1.0 and RDKit

  Update: KNIME protocol with the model available thanks to Greg Landrum. Update: New code to train the model and ONNX exported trained models available in github . The use and application of multi-task neural networks is growing rapidly in cheminformatics and drug discovery. Examples can be found in the following publications: - Deep Learning as an Opportunity in VirtualScreening - Massively Multitask Networks for Drug Discovery - Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set But what is a multi-task neural network? In short, it's a kind of neural network architecture that can optimise multiple classification/regression problems at the same time while taking advantage of their shared description. This blogpost gives a great overview of their architecture. All networks in references above implement the hard parameter sharing approach. So, having a set of activities relating targets and molecules we can tra