EBI ResOps Course 3rd July 2017

Recently the European Bioinformatics Institute (EMBL-EBI) (an outstation of the European Molecular Biology Laboratory) based at the Wellcome Genome Campus, has been running a course on what they are calling ‘ResOps’. By this they mean essentially the adoption of DevOps-style practices and methods to research computing, notably pushing the latter in more of an infrastructure-as-code, cloud-native direction, thereby hoping to improve reproducibility and scalability.

The course was delivered by Dario Vianello and Erik van den Bergh, from the Technology and Science Integration team at the EBI. All materials from the course including presentations and code were available online, at https://bit.ly/resops_june. The team had also provided temporary accounts for course delegates on their Embassy Cloud, on which we could run practical exercises.

Dario contrasted two possible approaches to moving a scientific pipeline to the cloud – mimicking on-site infrastructure versus a cloud-native approach. The former is clearly easier, but does mean that you will miss out on many of the benefits of running on a cloud. The latter requires a lot more effort initially and may not be feasible, depending on things like application support for object storage, but it does offer great potential advantages, not least of all being lower costs.

As well as taking us through the process of porting a pipeline to a cloud, there were three lightning talks showing concrete examples of existing and ongoing ResOps work. These comprised Marine Metagenomics, RenSeq and the Phenomenal project. One common theme here is that the porting work provided a very valuable opportunity to profile and optimise the existing pipeline, deriving detailed performance characteristics that in some cases weren’t available before. In the Metagenomics case, it was necessary to remove site-specific code and assumptions that made sense when running on local compute resources. It is also important to work on the scalability of research codes, and to test scaling up on clouds, because in some cases it may be that beyond a certain point adding more cloud instances will not increase performance. If working on a public cloud, this would mean paying more for no benefit. The PhenoMeNal project was particularly impressive in its adoption of modern technologies for the virtual research environment portal, while the RenSeq work comprised an interesting example of containerising a workload, and hosting the containers within VMs. They found that it was useful to include a specific reference genome in the Docker image, even though that meant a very large 9GB image size.


PhenoMeNal architecture

In the afternoon we worked on our own test application, using Terraform to create the cloud resources, and Ansible for configuration, all pulling from publicly available GitHub repositories.

The EBI Cloud Portal

I was very impressed with the thoroughness of the course and its realism. While recognising the advantages of cloud-native computing, they were realistic about the challenges of moving scientific pipelines in that direction, and provided valuable real-world experiences that will help making decisions about how and where to host those pipelines. The fact that the materials are completely open made it even more valuable. I thoroughly recommend taking the course to anyone interested in moving science to the cloud.


Cloud Workshop

The Francis Crick Institute

It’s been a few months since our November workshop so there’s been some time to digest and reflect on some of the common themes emerging. Having attended a couple of other conferences and workshops from my community (AGU and AMS) it’s interesting to compare.

Firstly, it was great to see such a variety of application areas represented. For this our second annual workshop, we opened it for the submission of abstracts and this made such a difference.  There was a great response, Life sciences having the margin on other domain areas.  We had 160 register and 120 on the day.   It was fantastic to have the Crick as a venue.  It worked really well.

The first session looked at applications of hybrid and public cloud.   Two really interesting use cases (Edinburgh and NCAS, NERC) looked at trying out HPC workloads on public cloud.  This raised issues around comparative performance and costs between public cloud and on-prem HPC facilities.

On AWS, Placement Groups allow instances to put close to one another to improve inter-node communication for MPI-based workloads.   This showed comparable performance with Archer (UK national supercomputer) for smaller workloads but clearly there was some limit as this tailed off as the number of nodes increased whereas Archer performance continued linearly with increase in scale. This tallies with what I’ve seen anecdotally at the AMS conference where there seemed to be on the one hand increasing uptake of public cloud for use with Numerical Weather Prediction jobs (which need MPI). However, this seemed to be being done for smaller scale workloads where they can stay within the envelope of the node affinity features available.

Another theme was portability – what kind of approaches can be used engineer workloads so that they can be easily moved between providers. Andrew Lahiff from STFC, presented a very different use case showing how container technologies can be used to run for Particle Physics, cases where there the focus is high-throughput rather than HPC requirements and so much more amenable to cloud.  This work has been done a part of a pilot for the Cloud Working Group to specifically investigate how containers and container orchestration technology can be used to provide an abstraction layer for cloud interoperability.  A really nice slide showed Kubernetes clusters running on Azure and Google cloud managed from the same command line console app.  Dario Vianello’s talk (EMBL-EBI) showed how an alternative approach using a combination of Ansible and Terraform can be used to deploy workloads across multiple clouds.

Microsoft’s Kenji Takeda presents on recent Azure developments

It was great to have talks from hyper-scale cloud providers AWS, Azure and Google.  The scale in hyper-scale is as ever impressive as is the pace of change in technology:  very interesting to see Deep Learning applications driving the development of custom hardware – TPUs and FPGAs. Plans underway to host data centres in the UK will ease uptake. OpenStack Foundation and Andy McNab‘s talks showed examples of federation across OpenStack clouds.

In the private cloud session, Stig Tefler gave a nice illustration of network performance for VMs showing how a number of aspects of virtualisation can be changed or removed to progressively improve network performance towards line rate. Alongside talks on private cloud, the parallel session looked at Legal, Policy and Regulatory Issues a critical issue for adoption of public cloud.  Steven Newhouse gave some practical insights from a recent cloud tender for EMBL. There is clearly need for further work around these issues so that the community can better informed about choices. This is something that the working group will be taking forward.

For this workshop, we experimented with an interactive session – bringing together a group of around 20 delegates to work together on some technical themes agreed ahead of time including bulk data movement and cloud-hosting of Jupyter notebooks. There was plenty of useful interaction and discussion but we will need to look at the networking provision for the next time to ensure groups can get on with technical work on the day.

We discussed next steps in the final session. There is clearly interest in taking particular areas forward from the meeting: focus groups on technical areas like HTC and use of parallel file systems with cloud or organised around specific domains within the research community. Training figured also, in the form of a cloud carpentry course so that researchers can more readily get up and running using cloud. Looking forward, in each of these case we’re looking for discrete activities with an agreed set of goals and something to deliver at the end. Where possible we’re seeking to support relevant work that is already underway and initiate new work where there are perceived gaps. We will be looking at running smaller workshops targeted at specific themes in the coming months as a means to engage and disseminate some of this work.

Phil Kershaw, STFC & Cloud-WG chair