Dan Crankshaw
Dan Crankshaw
The [Clipper NSDI paper](https://www.usenix.org/conference/nsdi17/technical-sessions/presentation/crankshaw) and the original [design doc](https://docs.google.com/document/d/1Ghc-CAKXzzRshSa6FlonFa5ttmtHRAqFwMg7vhuJakw/edit?usp=sharing) included support for *selection policies* to support use cases ranging from A/B testing and label-based or query-based model to more complex...
This is fairly self-explanatory, but basically there’s no way to delete models and applications right now. This has a pretty clear use case I think. The hardest part here will...
Right now you have to grep log messages to see where they are coming from. It would be nice to have line numbers and file names printed as part of...
We should document all of our Redis table schemas and uses
It would be great to extend our monitoring infrastructure to measure more physical performance. To start with, @blackhat06 suggested tracking the following resource metrics: + [ ] Disk IO: %...
For Kubernetes, we can use the experimental GPU support feature: https://kubernetes.io/docs/tasks/manage-gpus/scheduling-gpus/ For Docker, we can use [nvidia-docker](https://github.com/NVIDIA/nvidia-docker).
The current query frontend accepts predictions requests with a REST-like interface using HTTP+JSON. This has the benefit of being easy to use and widely supported, but comes at a performance...
When we run PageRank for too many iterations (100 iterations consistently triggers it), we get a stack overflow that stems from reading deeply nested serial data using the default java...
When running PageRank on a cluster, sometimes I hit a NoSuchElementException that's caused somewhere in VertexSetRDD. Full stack trace and command below. The line numbers may be slightly off due...