Building the worlds fastest* CryoEM processing platform (*under $100K)

Building the worlds fastest* CryoEM processing platform (*under $100K)

Jason Andrade1, Simon Brown1, James Bouwer1, Joshua Silver1

1University of Wollongong

Abstract

Situation

Cryogenic Electron Microscopy (CryoEM) continues to generate truly vast amounts of data. These volumes continue to increase as the cameras get better and more microscopes are deployed

UOW identified that our data processing needs (and costs) were growing. A lot!

We needed to build a new infrastructure platform to allow our facility to continue to meet these increased needs and we needed to do this without the “luxury” of a Go8 university budget.

Task

Take the 5 years of lessons we have learned from processing CryoEM data across 3 generations of infrastructure. Lots of people do this work – what have we learned that lets us build a better, more efficient and hopefully a more effective platform?

Design and build something that will work “out of the box” and tuned to our main processing application but also deals with the environmental (cooling, power) limitations as well as building to a budget.

How can we do this in the simplest possible infrastructure footprint (No, you cannot have a cluster!)

Does this technology even exist yet?

Action

Our group worked with our chosen vendor and separately with Nvidia to come up with a design that we were confident would work

We will go through the details we focused on including:

-The specifics of what we needed for computing
-What were the other things we needed to prioritize
-How do we make this work within our environment

Result

UOW currently has the world’s fastest CryoEM processing platform under $100k

Biography

Jason has been working in the Higher Ed sector for over 30 years, primarily in building and supporting research computing infrastructure.

In 2000 he built and ran what was Australia’s largest public software archive – aka “The AARNet Mirror”. He always been interested infrastructure platforms to support research, particularly the conjunction of compute, data and networks.

He has previously worked for the University of Queensland, QCIF, NCI and ANSTO in a variety of roles to build and support research computing.

Since 2017 he’s been based in Wollongong and worked closely with the Molecular Horizons CryoEM team to build a world class data storage and processing platform to support UOW’s investment in cutting edge microscopy.

He works as an independent consultant within the sector to try to build simple architectures to solve complex compute infrastructure problems.

Categories