skip to main content

Title: Demonstrating a Pre-Exascale, Cost-Effective Multi-Cloud Environment for Scientific Computing: Producing a fp32 ExaFLOP hour worth of IceCube simulation data in a single workday
Scientific computing needs are growing dramatically with time and are expanding in science domains that were previously not compute intensive. When compute workflows spike well in excess of the capacity of their local compute resource, capacity should be temporarily provisioned from somewhere else to both meet deadlines and to increase scientific output. Public Clouds have become an attractive option due to their ability to be provisioned with minimal advance notice. The available capacity of cost-effective instances is not well understood. This paper presents expanding the IceCube's production HTCondor pool using cost-effective GPU instances in preemptible mode gathered from the three major Cloud providers, namely Amazon Web Services, Microsoft Azure and the Google Cloud Platform. Using this setup, we sustained for a whole workday about 15k GPUs, corresponding to around 170 PFLOP32s, integrating over one EFLOP32 hour worth of science output for a price tag of about $60k. In this paper, we provide the reasoning behind Cloud instance selection, a description of the setup and an analysis of the provisioned resources, as well as a short description of the actual science output of the exercise.
Authors:
; ; ; ; ;
Award ID(s):
1841479 1826967 1941481 1841530 1148698 1730158
Publication Date:
NSF-PAR ID:
10211900
Journal Name:
PEARC '20: Practice and Experience in Advanced Research Computing
Page Range or eLocation-ID:
85 to 90
Sponsoring Org:
National Science Foundation
More Like this
  1. Sadayappan, Ponnuswamy ; Chamberlain, Bradford L. ; Juckeland, Guido ; Ltaief, Hatem (Ed.)
    As we approach the Exascale era, it is important to verify that the existing frameworks and tools will still work at that scale. Moreover, public Cloud computing has been emerging as a viable solution for both prototyping and urgent computing. Using the elasticity of the Cloud, we have thus put in place a pre-exascale HTCondor setup for running a scientific simulation in the Cloud, with the chosen application being IceCube's photon propagation simulation. I.e. this was not a purely demonstration run, but it was also used to produce valuable and much needed scientific results for the IceCube collaboration. In order to reach the desired scale, we aggregated GPU resources across 8 GPU models from many geographic regions across Amazon Web Services, Microsoft Azure, and the Google Cloud Platform. Using this setup, we reached a peak of over 51k GPUs corresponding to almost 380 PFLOP32s, for a total integrated compute of about 100k GPU hours. In this paper we provide the description of the setup, the problems that were discovered and overcome, as well as a short description of the actual science output of the exercise.
  2. The DeepLearningEpilepsyDetectionChallenge: design, implementation, andtestofanewcrowd-sourced AIchallengeecosystem Isabell Kiral*, Subhrajit Roy*, Todd Mummert*, Alan Braz*, Jason Tsay, Jianbin Tang, Umar Asif, Thomas Schaffter, Eren Mehmet, The IBM Epilepsy Consortium◊ , Joseph Picone, Iyad Obeid, Bruno De Assis Marques, Stefan Maetschke, Rania Khalaf†, Michal Rosen-Zvi† , Gustavo Stolovitzky† , Mahtab Mirmomeni† , Stefan Harrer† * These authors contributed equally to this work † Corresponding authors: rkhalaf@us.ibm.com, rosen@il.ibm.com, gustavo@us.ibm.com, mahtabm@au1.ibm.com, sharrer@au.ibm.com ◊ Members of the IBM Epilepsy Consortium are listed in the Acknowledgements section J. Picone and I. Obeid are with Temple University, USA. T. Schaffter is with Sage Bionetworks, USA. E. Mehmet is with the University of Illinois at Urbana-Champaign, USA. All other authors are with IBM Research in USA, Israel and Australia. Introduction This decade has seen an ever-growing number of scientific fields benefitting from the advances in machine learning technology and tooling. More recently, this trend reached the medical domain, with applications reaching from cancer diagnosis [1] to the development of brain-machine-interfaces [2]. While Kaggle has pioneered the crowd-sourcing of machine learning challenges to incentivise data scientists from around the world to advance algorithm and model design, the increasing complexity of problem statements demands of participants to be expert datamore »scientists, deeply knowledgeable in at least one other scientific domain, and competent software engineers with access to large compute resources. People who match this description are few and far between, unfortunately leading to a shrinking pool of possible participants and a loss of experts dedicating their time to solving important problems. Participation is even further restricted in the context of any challenge run on confidential use cases or with sensitive data. Recently, we designed and ran a deep learning challenge to crowd-source the development of an automated labelling system for brain recordings, aiming to advance epilepsy research. A focus of this challenge, run internally in IBM, was the development of a platform that lowers the barrier of entry and therefore mitigates the risk of excluding interested parties from participating. The challenge: enabling wide participation With the goal to run a challenge that mobilises the largest possible pool of participants from IBM (global), we designed a use case around previous work in epileptic seizure prediction [3]. In this “Deep Learning Epilepsy Detection Challenge”, participants were asked to develop an automatic labelling system to reduce the time a clinician would need to diagnose patients with epilepsy. Labelled training and blind validation data for the challenge were generously provided by Temple University Hospital (TUH) [4]. TUH also devised a novel scoring metric for the detection of seizures that was used as basis for algorithm evaluation [5]. In order to provide an experience with a low barrier of entry, we designed a generalisable challenge platform under the following principles: 1. No participant should need to have in-depth knowledge of the specific domain. (i.e. no participant should need to be a neuroscientist or epileptologist.) 2. No participant should need to be an expert data scientist. 3. No participant should need more than basic programming knowledge. (i.e. no participant should need to learn how to process fringe data formats and stream data efficiently.) 4. No participant should need to provide their own computing resources. In addition to the above, our platform should further • guide participants through the entire process from sign-up to model submission, • facilitate collaboration, and • provide instant feedback to the participants through data visualisation and intermediate online leaderboards. The platform The architecture of the platform that was designed and developed is shown in Figure 1. The entire system consists of a number of interacting components. (1) A web portal serves as the entry point to challenge participation, providing challenge information, such as timelines and challenge rules, and scientific background. The portal also facilitated the formation of teams and provided participants with an intermediate leaderboard of submitted results and a final leaderboard at the end of the challenge. (2) IBM Watson Studio [6] is the umbrella term for a number of services offered by IBM. Upon creation of a user account through the web portal, an IBM Watson Studio account was automatically created for each participant that allowed users access to IBM's Data Science Experience (DSX), the analytics engine Watson Machine Learning (WML), and IBM's Cloud Object Storage (COS) [7], all of which will be described in more detail in further sections. (3) The user interface and starter kit were hosted on IBM's Data Science Experience platform (DSX) and formed the main component for designing and testing models during the challenge. DSX allows for real-time collaboration on shared notebooks between team members. A starter kit in the form of a Python notebook, supporting the popular deep learning libraries TensorFLow [8] and PyTorch [9], was provided to all teams to guide them through the challenge process. Upon instantiation, the starter kit loaded necessary python libraries and custom functions for the invisible integration with COS and WML. In dedicated spots in the notebook, participants could write custom pre-processing code, machine learning models, and post-processing algorithms. The starter kit provided instant feedback about participants' custom routines through data visualisations. Using the notebook only, teams were able to run the code on WML, making use of a compute cluster of IBM's resources. The starter kit also enabled submission of the final code to a data storage to which only the challenge team had access. (4) Watson Machine Learning provided access to shared compute resources (GPUs). Code was bundled up automatically in the starter kit and deployed to and run on WML. WML in turn had access to shared storage from which it requested recorded data and to which it stored the participant's code and trained models. (5) IBM's Cloud Object Storage held the data for this challenge. Using the starter kit, participants could investigate their results as well as data samples in order to better design custom algorithms. (6) Utility Functions were loaded into the starter kit at instantiation. This set of functions included code to pre-process data into a more common format, to optimise streaming through the use of the NutsFlow and NutsML libraries [10], and to provide seamless access to the all IBM services used. Not captured in the diagram is the final code evaluation, which was conducted in an automated way as soon as code was submitted though the starter kit, minimising the burden on the challenge organising team. Figure 1: High-level architecture of the challenge platform Measuring success The competitive phase of the "Deep Learning Epilepsy Detection Challenge" ran for 6 months. Twenty-five teams, with a total number of 87 scientists and software engineers from 14 global locations participated. All participants made use of the starter kit we provided and ran algorithms on IBM's infrastructure WML. Seven teams persisted until the end of the challenge and submitted final solutions. The best performing solutions reached seizure detection performances which allow to reduce hundred-fold the time eliptologists need to annotate continuous EEG recordings. Thus, we expect the developed algorithms to aid in the diagnosis of epilepsy by significantly shortening manual labelling time. Detailed results are currently in preparation for publication. Equally important to solving the scientific challenge, however, was to understand whether we managed to encourage participation from non-expert data scientists. Figure 2: Primary occupation as reported by challenge participants Out of the 40 participants for whom we have occupational information, 23 reported Data Science or AI as their main job description, 11 reported being a Software Engineer, and 2 people had expertise in Neuroscience. Figure 2 shows that participants had a variety of specialisations, including some that are in no way related to data science, software engineering, or neuroscience. No participant had deep knowledge and experience in data science, software engineering and neuroscience. Conclusion Given the growing complexity of data science problems and increasing dataset sizes, in order to solve these problems, it is imperative to enable collaboration between people with differences in expertise with a focus on inclusiveness and having a low barrier of entry. We designed, implemented, and tested a challenge platform to address exactly this. Using our platform, we ran a deep-learning challenge for epileptic seizure detection. 87 IBM employees from several business units including but not limited to IBM Research with a variety of skills, including sales and design, participated in this highly technical challenge.« less
  3. In this paper we describe how high performance computing in the Google Cloud Platform can be utilized in an urgent and emergency situation to process large amounts of traffic data efficiently and on demand. Our approach provides a solution to an urgent need for disaster management using massive data processing and high performance computing. The traffic data used in this demonstration is collected from the public camera systems on Interstate highways in the Southeast United States. Our solution launches a parallel processing system that is the size of a Top 5 supercomputer using the Google Cloud Platform. Results show that the parallel processing system can be launched in a few hours, that it is effective at fast processing of high volume data, and can be de-provisioned in a few hours. We processed 211TB of video utilizing 6,227,593 core hours over the span of about eight hours with an average cost of around $0.008 per vCPU hour, which is less than the cost of many on-premise HPC systems.
  4. Light microscopy provides a window into another world that is not visible to the unaided eye. Because of this and its importance in biological discoveries, the light microscope is an essential tool for scientific studies. It can also be used with a variety of easily obtained specimens to provide dramatic demonstrations of previously unknown features of common plants and animals. Thus, one way to interest young people in science is to start with an introduction to light microscopy. This is an especially effective strategy for individuals who attend less advantaged or under-resourced schools, as they may not have been previously exposed to scientific concepts in their classes. However, introducing light microscopy lessons in the classroom can be challenging because of the high cost of light microscopes, even those that are relatively basic, in addition to their usual large size. Efforts are underway by our laboratory in collaboration with the Biophysical Society (BPS) to introduce young people to light microscopy using small, easy-to-assemble wooden microscopes developed by Echo Laboratories. The microscopes are available online as low-cost kits ($10 each with shipping), each consisting of 19 parts printed onto an 81⁄2 x 11 inch sheet of light-weight wood (Fig. 1). After punchingmore »out the pieces, they can be assembled into a microscope with a moveable stage and a low-power lens, also provided in the kit (Fig. 2). Photos taken with a cell phone through the microscope lens can give magnifications of ~16-18x, or higher. At these magnifications, features of specimens that are not visible to the unaided eye can be easily observed, e.g., small hairs on the margins of leaves or lichens [1]. As a member of the BPS Education Committee, one of us (SAE) wrote a Lesson Plan on Light Microscopy specifically for use with the wooden microscopes. SAE was also able to obtain a gift of 500 wooden microscope kits for the BPS from Echo Laboratories and Chroma Technology Corp in 2016. The wooden microscope kits, together with the lesson plan, have provided the materials for our present outreach efforts. Rather than giving out the wooden microscope kits to individuals, the BPS asked the Education Committee to maximize the impact of the gift by distributing the microscopes with the Lesson Plan on Light Microscopy to teachers, e.g., through teachers’ workshops or outreach sessions. This strategy was devised to enable the Society to reach a larger number of young people than by giving the microscopes to individuals. The Education Committee first evaluated the microscopes as a tool to introduce students to scientific concepts by providing microscopes to a BPS member at the National University of Colombia who conducted a workshop on Sept 19-24, 2016 in Tumaco, Columbia. During the workshop, which involved 120 high school girls and 80 minority students, including Afro-Colombian and older students, the students built the wooden microscopes and examined specimens, and compared the microscopes to a conventional light microscope. Assembling the wooden microscopes was found to be a useful procedure that was similar to a scientific protocol, and encouraged young girls and older students to participate in science. This was especially promising in Colombia, where there are few women in science and little effort to increase women in STEM fields. Another area of outreach emerged recently when one of us, USP, an undergraduate student at Duke University, who was taught by SAE how to assemble the wooden microscopes and how to use the lesson plan, took three wooden microscopes on a visit to her family in Bangalore, India in summer 2018 [2]. There she organized and led three sessions in state run, under-resourced government schools, involving classes of ~25-40 students each. This was very successful – the students enjoyed learning about the microscopes and building them, and the science teachers were interested in expanding the sessions to other government schools. USP taught the teachers how to assemble and use the microscopes and gave the teachers the microscopes and lesson plan, which is also available to the public at the BPS web site. She also met with a founder of the organization, Whitefield Rising, which is working to improve teaching in government schools, and taught her and several volunteers how to assemble the microscopes and conduct the sessions. The Whitefield Rising members have been able to conduct nine further sessions in Bangalore over the past ~18 months (Fig. 3), using microscope kits provided to them by the BPS. USP has continued to work with members of the Whitefield Rising group during her summer and winter breaks on visits to Bangalore. Recently she has been working with another volunteer group that has expanded the outreach efforts to New Delhi. The light microscopy outreach that our laboratory is conducting in India in collaboration with the BPS is having a positive impact because we have been able to develop a partnership with volunteers in Bangalore and New Delhi. The overall goal is to enhance science education globally, especially in less advantaged schools, by providing a low-cost microscope that can be used to introduce students to scientific concepts.« less
  5. Cloud providers offer instances with similar compute capabilities (for example, instances with different generations of GPUs like K80s, P100s, V100s) across many regions, availability zones, and on-demand and spot markets, with prices governed independently by individual supplies and demands. In this paper, using machine learning model training as an example application, we explore the potential cost reductions possible by leveraging this cross-cloud instance market. We present quantitative results on how the prices of cloud instances change with time, and how total costs can be decreased by considering this dynamic pricing market. Our preliminary experiments show that a) the optimal instance choice for a model is dependent on both the objective (e.g., cost, time, or combination) and the model’s performance characteristics, b) the cost of moving training jobs between instances is cheap, c) jobs do not need to be preempted more frequently than once a day to leverage the benefits from spot instance price variations, and d) the cost of training a model can be decreased by as much as 3.5× compared to a static policy. We also look at contexts where users specify higherlevel objectives over collections of jobs, show examples of policies for these contexts, and discuss additional challengesmore »involved in making these cost reductions viable.« less