Cornell University creating a graduate course on cloud computing, to educate the next generation

Guest Blog by Ken Birman, N. Rama Rao Professor of Computer Science, Cornell University



At Cornell University, we have a long history in the area of “cloud computing” – dating back to before the subject even had that name! 

I’m Ken Birman, and I’ve been involved in distributed computing and scalability from the early days.  Back in the 1990’s, I created the first widely impactful software for fault-tolerant, scalable, consistent computing and data management, and my work was used very broadly.  Even now, if you fly into Europe there is a good chance that the air traffic control system guiding your plane uses an architecture (and software) I created.  For a decade my software ran the New York Stock Exchange and Swiss Stock Exchange without a single unmanaged fault – the code was designed to self-heal – and quite a few companies (including Microsoft) use my technology in their data-center management layers.

My students did well too: Ranveer Chandra is Microsoft’s Chief Scientist for Azure Global (and also the lead researcher for networking, and for IoT computing on smart farms).  Werner Vogels is CTO at Amazon, and Dahlia Malkhi leads in the development of Facebook’s Diem digital currency.

So, it seemed natural to create a graduate course on cloud computing, to educate the next generation of developers, entrepreneurs, and leaders.  When planning CS5412, “Topics in Cloud Computing”, I knew that hands-on experience would be key, and was fortunate to be offered tremendous support by Microsoft’s academic liaison group, along with Azure computing resources for our class projects.  At the same time, I wanted to give the students a focus and concrete scenarios, which led me to partner with Cornell’s ANSC 3510, a course focused on technology in dairy farming.  Working with Professor Julio Giordano, who created that course, we’ve created a unique structure in which two completely different classes are able to team up (using Microsoft Teams, of course!) to jointly create projects that demonstrate the power of cutting edge IoT infrastructures backed by Azure cloud ML and AI.

A few examples will help set the context (all the student groups define individualized projects, so these are really two of many).  When cows enter a dairy to be milked, a procession of as many as 500 animals will move from outside in the field into the milking area of the barn through a series of side-by-side corridors.  By installing cameras, we can photograph the animals as they pass through.  With cloud computing tools, my students are able to identify the animal (RFID tags should make this easy, but in fact they often fall off or are difficult to ping in crowded settings), assess the health of each animal, decide if the cow needs to be routed into the “cow wash” before milking etc.  One project looks at this specific question, hosting the solution on Azure IoT Edge.

A second project focused on building a highly flexible dairy dashboard for the farmer, tracking a variety of milk quality, quantity and safety properties while also providing an integrated perspective on the herd as a whole, dairy product price trends, and even allowing the farmer to zoom in and look at health statistics for individual animals.   Automated notifications alert the farmer if a cow is close to calving or needs to see a vet for some other reason.    Yet a third project looks at integrating data from diverse sources, such as weather predictions, farm-soil moisture levels and imaging to predict risk factors such as outbreaks of pathogens.  A farmer warned of trouble three weeks from now might have time to bring the crop in a little early, before the pathogen gains a toehold.

What I find exciting about these projects is that they bring my cloud computing students into real-world settings.  In teams of two or three, students from my cloud computing class partner with Professor Giordano’s students to understand the kinds of data available and to appreciate the proper way to approach technical questions seen on dairy farms.  They work together to map these questions to ML, decide how to train the models, and even collaborate to design dashboards that visualize the outcomes of the analysis.  Meanwhile, Professor Giordano’s animal science students, who come to these partnerships with a lot of knowledge about modern dairies and modern AI, learn about the challenges of putting those tools to work to create useful new technology, and are exposed to the power (and limitations) of the Azure cloud.   As a side-effect, by working with Microsoft Teams, the groups also become comfortable with cutting-edge collaboration technology.

My approach in CS5412 has been to design to the course around three central goals.  First, I do want my students to appreciate the relevant computer science foundations: topics such as state machine replication (Paxos), leader election, and fault-tolerant consensus.  Second, it is important for them to learn to view the cloud as a platform with powerful, inherently scalable services such as CosmosDB, the Azure Blob Store and Azure Data Lake at their disposal, along with powerful computing frameworks like the Azure Intelligence Platform and Cognitive Search.  These in turn center on fundamental practical questions, such as understanding data and computation sharding for scalability, and appreciating the roles of the IoT Hub and its integration with Azure Functions. 

Last is the practical goal of becoming proficient with the actual tools: Any student who “survives” my course emerges as a capable cloud programmer with a valuable skill set coupled to an understand of how the Azure documentation and prebuilt “recipes” can lead to customizable code frameworks ideally fitted to their needs.

It has taken us many years to evolve CS5412 to the current state, but today I feel as if my students really learn the right material, that the partnership with Animal Sciences is operating smoothly and bringing really high value to the table, and that the course is simply working very well.  My textbook, written in 2010, probably needs a thorough revision at this point, but on the other hand, modern students don’t really read textbooks!  What they do read is online material, especially coupled with “how-to” videos and even hands-on opportunities, and on this, Microsoft has excelled. 

The documentation available is outstanding, but beyond that, Microsoft has participated in Cornell’s Digital Agriculture Hackathons (the CIDA Hackathon), and my students find this to be a fantastic resource.  During a February weekend students come together to learn how cloud computing can solve problems in agriculture, looking at technical but also social and economic/business perspectives, with hands-on instruction by Microsoft experts.

CS5412 has become a real treasure here at Cornell, and is helping position our students to hit the ground running as they enter the workforce.  We couldn’t have done it without Microsoft Azure and Microsoft Teams, and are tremendously grateful for the partnership opportunity!