Leveraging RDMA to Build a Real-Time Cloud for the Internet of Things
Ken Birman ( Cornell University )
-
11:00 11th October 2016 ( week 1, Michaelmas Term 2016 )Tony Hoare Room, Robert Hooke Building, Parks Road (NB unusual place and time)
We’ve all read the hype about the Internet of Things: a technology trend that seems endlessly about to happen, yet has been
stubbornly hard to integrate with cloud computing. In this talk I want to ask why this has been so, and how we can fix it.
I’ll start by describing work Cornell has done over the past few years on creating a cloud platform to host “smart power grid”
applications. This involves (1) new rack-scale management solutions aimed at applications that need to run 24x7, (2) replication
with ultra-fast updates for scalable real-time responsiveness, and (3) new real-time storage solutions, to enable a new kind
of big-data temporal computing. But new models also create huge performance puzzles and there are clashes between the properties
of the cloud as normally operated today and the needs of the IoT world. We created a series of new solutions at Cornell that
aim squarely at these demands. I'll focus on two examples. The first is a novel new file system that offers a very friendly
environment for running machine learning applications on IoT data (we call it the Freeze Frame File System). The second is
a bit more technically gnarly: IoT has forced us to build a whole bunch of services like the file system that share a set
of common needs. Rather than solve the same problem again and again, we created a new operating system component we call Derecho
that offers stunningly fast data movement, replication and coordination, literally thousands of times faster than anything
previously available. Derecho is a bit trickier to understand, yet is in some sense the most important thing we've done, and
it opens the door to doing all sorts of super-fast, super-responsive IoT solutions (and should have other uses in the cloud
as well, because so many things need to move data as fast as possible). This work is all open-source, and we intend to build
up an open development, collaborative community that would use our solutions and share new technologies in this space. We
hope some in the audience might be tempted to join the revolution!