Mobile Data Labs & Microsoft • Sr. Applied Scientist/Software Engineer, MobileDataLabs (acquired team inside Microsoft) team building AI/Analytics platform • Spent couple of years with Microsoft Research • Spent couple of years with Azure HDInsight • Among other things, author of “Apache Mesos Essentials” • Opinions are mine and biased • You can find me as @dharmeshkakadia everywhere • Slides @ : http://dharmeshkakadia.github.io/talks/ whoami
Mobile Data Labs & Microsoft MileIQ: Mileage Tracking Made Easy Automatic Detection Easy Classification Personalized Experience Robust Reporting MileIQ.com
Mobile Data Labs & Microsoft • Already a widely adopted standard for deploying. We are just extending it to Data/AI • Operational benefits for free – monitoring, CI/CD, secrets, alerts, log management …. • No separate process and tools for data and other parts of engineering • Ability to leverage latest improvements faster • Better resource utilization • Helps avoiding data silos • Not have to worry about installing Nvidia drivers… • With a caveat, that you need a little more cross functional expertise Why build AI platform on k8s?
Mobile Data Labs & Microsoft • Drive quality measurement • Dexter: Signals and Metrics Framework • Marketing and Engagement analytics • Data management – GDPR, schema etc. • Data science and Experimentation • User segmentation • … • …Runs billions of tasks a day • Used by engineering, marketing, data science teams • SQL, Python, Spark, Pandas, Tensorflow, Jupyter AI/Analytics Platform : Our current use cases
AI/Analytics Platform : Process Mobile Data Labs & Microsoft Dev Work inner loop development inside Juypter notebook. Write docker + YAML file when ready for PR Build Takes Dockerfile and build and pushes image to container registry with build tags. Release Combines YAML file and secrets applies on k8s cluster Monitor and visualize Produces output data, models and results. That is used for further analysis/decision making. No special ops required.
AI/Analytics Platform : Guiding Principles and Tradeoffs • Optimize for agility & turn around time • Covers all the use cases – ETL, Streaming, ML, Visualization • Covers full life cycle – Dev, Deployment, Monitoring, Alerting, and so on • Full API access – connect to all the tools you wish • Get the cutting edge features (latest versions etc.) • Easy to use • Permission-less or self-serve • Somewhat future proof • Open Source and Linux friendly • Cloud Native/friendly • Best practice enforcing by default • Gall's Law – start with simple system that works and evolve Mobile Data Labs & Microsoft
Mobile Data Labs & Microsoft • Docker as a single build tool • Consistent deploys – even more useful in data experiments • Freedom to use any library/versions I want • k8s as a single deployment tool • Easier to think about for everyone on the team • East to remember (and optimize!) one pattern and workflow • Separation of concerns • Build/ops tools doesn’t need to understand how TF work • YAML for separating code and configs. • Secrets for code and secrets. • Blobfuse for code and data paths. Declarative ML deployments
Mobile Data Labs & Microsoft Volumes • Blobfuse • k8s volume plugin that makes blob data accessible as a mounted file system • Great for inner loop dev. Avoids additional IO to remote storage. • Allows read-only or read-write mounting • Not every tool needs to understand and integrate with blob • Easy when playing around with data rather than dealing with blob explorers • Configurable cache interval allows trading off fast access/freshness constraints. • Hostpath • Local SSD for temporary storage • Great for speed and intermediate results • Azure files • For permanent storage for fast non-blob data
Mobile Data Labs & Microsoft Example End-to-End pipeline • Tensorflow for model training and serving • Spark for feature engineering • Kubernetes & related tools for deployment • Data lives on blob and DW
FAQ : Serving • Early days • Currently we store and serve directly out of blob • Versioned though names • Want to validate and understand use cases to help us to choose the right tool • Need something that plays nice with other data tools as well i.e. spark etc. • We are considering onnx and tensor-serve
FAQ : Kubeflow? • We evaluated very early version 0.1.0 and had a lot of issues with it • Opinionated and bundles a lot of tools that we currently don’t need • Ksonnet L • End user simplicity is paramount for us • We like to start with simple tools and add tools as necessary vs big bang complex pieces • Gall’s law • Having said that, • Huge fan of the community. • We are keeping an eye on its direction • We do like some parts – especially TF job operator. We already use spark operator and realize the benefits.
dharmeshkakadia
codenote
chloe463
cyberagentdevelopers
naohachi89
asei
oracle4engineer
rynsuke
baseballyama
sansan_randd
kojira
kawagoi
mthomps
chriscoyier
eileencodes
panda_program
garrettdimon
kneath
marcelosomers
davidbonilla
zakiwarfel
hannesfritz
chrislema
skipperchong