May 10-12, 2023 at Emory University in Atlanta, GA

Cloud Forum 2023 Presentation Abstracts

— Back to the Cloud Forum Schedule

May 11 Presentations

Research & Enterprise Cloud Venn Diagram

Joshua Stultz, National Institutes of Health
9:45 am | Slides

The NIH STRIDES team has established enterprise cloud infrastructure on each of the three major cloud platforms: AWS, GCP, and Azure. In order to support cloud adoption across the NIH, the STRIDES team needed to work with core IT functions in NIH’s Center for Information Technology (CIT) in order to extend on-premises services and infrastructure to the cloud. Examples of these functions include the following:

Often, conflicting interests can arise between the desire to help build innovative, cloud-enabled research and data science solutions and the requirements to provide secure and resilient enterprise cloud services. We have learned — often the hard way! — to identify common interests between research goals and enterprise cloud necessities, and also to raise awareness of the overarching benefits enterprise cloud services provide to the research community to help build better security, trust, and cooperation when developing and implementing cloud-enabled research solutions.

In this talk, we will explore the Venn diagram of Research Cloud and Enterprise Cloud and highlight empirical approaches for “minding the gap.”

Where’s my cloud instruction book?!

Joseph Rafferty, Texas A&M University
10:45 am | Slides
Cloud computing is often compared to being given buckets of mixed LEGO blocks with no instructions to use to make your ideas come to life. But instructions do exist! At Texas A&M, we’re using automation, infrastructure as code, and tools like GitHub and Kion to bring our own and our institution’s workloads to life in the cloud. Learn how we’re enabling our institution to get started building faster by giving them cloud accounts with billing, budgets, compliance, and security parts already assembled.

Lightning Talks

11:00 am | Slides

  1. Automating DNS Registration for Azure – Simon Fairey, Princeton University
    See how Princeton has automated Infoblox DNS registration for Azure resources using Service Now and protected ourselves against dangling DNS attacks in the process.
  2. SELECT * FROM CLOUD – Matthew Rich, Northwestern University
    Northwestern uses a tool called CloudQuery (https://cloudquery.io/) to create a daily inventory “snapshot” of all of our cloud environments which are saved to a SQL database. I’ll talk about how it works, what it does for us, and a few useful queries.
  3. Automating EC2 Patching – Ezequiel Gioia, University of Central Florida
    Enterprise servers are likely up to date in terms of security patches. But what about those unmanaged servers your researchers are spinning up in the cloud? A brief talk about enabling your users to automatically patch EC2 instances.
  4. Approach to Setting up Cloud-App Configs – Yu Li, Virginia Tech
    It is common for an application to require a custom set of configurations within different environments and development stages. For example, applications typically connect to other environment-specific resources, like databases or REST APIs. Setting up application configurations is a challenging problem, and it requires careful design to achieve a secure and efficient solution. In this talk, we will briefly discuss our best practice on leveraging AWS Secret Manager, GIT repo, Vault, and local environment variables to secure application configurations with convenience for both development and deployment.
  5. Cloud Insights: Who, How, Why, How Much? – Todd Reilly, National Institutes of Health
    The NIH STRIDES Initiative is one of many NIH-wide efforts to implement the NIH Strategic Plan for Data Science, which provides a roadmap for modernizing the NIH-funded biomedical data science ecosystem. Through STRIDES, NIH and NIH-funded researchers gain access to reduced pricing (among other benefits) on cloud services across three cloud platforms: AWS, GCP, and Azure. NIH leadership regularly evaluates the impact of STRIDES on enabling a more robust, interconnected NIH biomedical research ecosystem. To facilitate this, the STRIDES team must regularly compile, analyze, and concisely convey information across a range of interest areas about how cloud platforms in STRIDES are being used, by whom, and for what purposes. In this presentation, we will provide examples of some of the most common data requests that the STRIDES team receives, as well as the processes, business rules, and data visualizations used to keep NIH leadership informed about how cloud enablement is supporting biomedical data science. We will also preview our low-code business application that was designed to maintain our records and enable programmatic reporting as well as analytics to facilitate data-driven decision-making.

Tag, You’re It!

Jonathan White, University of Central Florida
11:30 am | Slides
Has someone asked you what your university is spending in the cloud? If not, they will. One of the first steps to answering this question is tagging your cloud resources. UCF recently finished a major infrastructure tagging project and this presentation will highlight the steps we took to implement tagging across Azure, AWS, and eventually, Google Cloud. Using documentation from these cloud providers, we developed a multi-step project plan around discovery, design, remediation, and governance. We will share this plan, best practices, lessons learned, pain points, and provide step-by-step guidance so that you’ll be ready to implement tagging at your university.

Developing FinOps at UT Austin

Maggie Spangenberg, The University of Texas at Austin
1:30 pm | Slides & Recording
UT Austin will share their FinOps journey from initial cloud exploration to exponential growth. UT’s processes needed to evolve as cloud spend went from nothing to over $1M annually. UT is planning to show dashboard visualizations that allow departments across campus to see their current spend, cost optimization opportunities, and usage data. There are also internal processes for tagging cloud resources and billing back to departments for recovering spend.

Moving Faster With the Cloud Everywhere

Phil Fenstermacher, William & Mary
2 pm | Slides & Recording
Moving applications to the public cloud can be daunting. There are years of organic growth that need to be unwound, staff to train, project plans to write, and new architectures to design. To ease the pain William & Mary chose to adopt cloud concepts and best practices in our on-premise operations before any significant migrations. In this presentation, we’ll talk about what changes we made, how they prepared us for cloud migrations, and how it continues to make hybrid operations easier (because we still have a network on-campus to run). We’ll close with a brief case study of how the single mode of operation made moving our student information system to the cloud trivial and nearly transparent to our users.

Multicloud Cost & Usage Reporting at NIH

Nick Weber, National Institutes of Health
3 pm | Slides
Like most higher-ed institutions, the NIH is a complex organization. It comprises 27 Institutes, Centers, and Offices (ICO), each with its own separate budgets and administrative practices, and each comprising various hierarchical organizational structures to support many areas of biomedical research. The NIH STRIDES Initiative provides centralized billing and regular cost and utilization reporting for thousands of account holders for each of three major cloud service provider (CSP) platforms and across myriad projects and programs within the 27 NIH ICO. This ensures that NIH technical, administrative, and research/program staff have ready access to the information needed to facilitate research activities and manage budgets at the levels of individual users or investigators, groups of laboratories or offices, and multiple higher-level aggregations for divisions, ICOs, and NIH as a whole. For the past 5 years, the STRIDES team has integrated CSP-provided cost and usage data with other, internally derived records of account-holder and project information. The team has also implemented automation to generate hundreds of monthly spending and balance reports within NIH. However, scaling has become a major challenge, especially in response to a variety of stakeholder needs and wants. In this presentation, we will discuss the complexities and lessons learned in devising our current system for multi-cloud billing and cost reporting, as well as discuss ongoing and planned enhancements to our processes and tools to scale cost and usage reporting for STRIDES.

Introducing A Cloud-based Analytics Platform

Yu Li, Virginia Tech
3:30 pm | Slides
This presentation will introduce an analytics platform on the AWS Cloud we built from the ground up using cutting-edge open-source technologies. In the first part, we will present its data framework – how we organize and manage our PostgreSQL databases, and how we collect data from various and enormous data sources with the ETL tasks managed by an Airflow cluster. In the second part, we will discuss its backend web server framework – how we implement and secure our Django web servers with a bunch of AWS services, and how we achieve zero downtime deployment with GitLab ci/cd pipelines. In the third part, we will present its frontend website framework – how we implement a stable, secure, and efficient SPA website with Vue.js, Highcharts, and other open-source tools. The community will observe a practical way to build such a platform and gain knowledge that will be useful no matter what stage of development they are in.

Session Takeaways:

  1. Learn a proven architecture for building an analytics platform on cloud.
  2. Learn our innovations on improving availability, security, and performance.
  3. Learn best practices on DevOps for both development and deployment.

Meeting Traditional High-Performance Computing Demands in the Cloud

Circe Tsui, Emory University
Paul Petersen, Emory University

4:30 pm | Slides
In 2022, the Emory University Office of Information Technology successfully built a proof of concept high-performance computing cluster featuring premier GPUs, on the “AWS at Emory” platform. Several researchers completed their machine learning projects on the cluster and demonstrated the cluster’s value to the artificial intelligence research in the institution. In this session, we will discuss the motivation behind the POC, the design of the cluster and the POC results. In addition, we will share our lessons learned as well as the roadmap of a cloud HPC cluster offering at Emory.