KubeCon + CloudNativeCon North America 2017

GPU is becoming the new common, but at the moment, GPU resources are still hard to find for people who wants to have a taste. So how to build your GPU machine learning cloud?

Resource management & App templating
Even if your company or organization have purchased some GPU devices. Environment and resource isolation is always a problem. Also at the beginning the cloud is more used as a playground, so another consideration is to improve usage rate of resources. How we use Kubernetes to solve this problems.

How to use a wizard to generate machine learning, you can choose using tensorflow or theano, how many GPUs you need, etc.

Make the “customized changes” in immutable container be played back.
The features of container is immutable, which is a double-edged sword, which can ensure the environment can be unique/portable. On the other side, any changes inside the running container can be lost after recreation. How the customed env is saved and reuse?

Managing persistence storage in Kubernetes
How to turn our RBD served as hosted s3, to save models, training data, and so on. So The data scientist can access their data both as a volume and s3 standard api.
Support the running machine learning app,like tensorflow to do online resize.

App model & permission control
We'll talk about the app center , design of appcode and permission control.

GPUs are critical to some artificial intelligence workflows. In particular, workflows that utilize TensorFlow, or other deep learning frameworks, need GPUs to efficiently train models on image data. These same workflows typically also involve mutli-stage data pre-processing and post-processing. Thus, a unified framework is needed for scheduling multi-stage workflows, managing data, and offloading certain workloads to GPUs.

In this talk, we will introduce a stack of open source tooling, built around Kubernetes, that is powering these types of GPU-accelerated workflows in production. We will do a live demonstration of a GPU enabled pipeline, illustrating how easy it is to trigger, update, and manage multi-node, accelerated machine learning at scale. The pipeline will be fully containerized, will be deployed on Kubernetes via Pachyderm, and will utilize TensorFlow for model training and inference.

Kubernetes promises to be a multi workload platform. This talk will explore how Kubernetes can be easily leveraged to build a complete Deep Learning pipelines starting all the way from data ingestion/aggregation, pre-processing, ML training, and serving with the mighty Kubernetes APIs. This talk will use Tensorflow and other other ML frameworks to highlight the value that Kubernetes brings to Machine Learning. Along the way, key infrastructure features introduced to abstract and handle hardware accelerators which make Machine Learning possible will also be presented.

Database as a Service is one of the most interesting and challenging domains on the cloud industry. In eBay, we implemented a cloud-native geo-distributed document service based on the kubernetes. eBay extended the kubernetes to support local disk volume on bare metal machine, which enables the high performance DB can be deployed on the kubernetes as a Pod. On top of the kubernetes platform, we develop a control layer to orchestrate the databased pods and enable it can be distributed on multiple cluster, and expand the WISB model to use a workflow to auto manage the database cluster.

MySQL is the world's most popular open source database and there are a number of ways to run it on Kubernetes. This talk will cover each type of MySQL deployment strategy starting from a simple MySQL pod, to a asynchronous replicated master-slave, synchronous Galera cluster, and on to a Vitess clustering system which allows for horizontal scaling of MySQL and innately has built-in sharding, explaining how each is deployed, what features are available, and what type of application they lend themselves to.

How can UN humanitarian aid field experts use social media to gain insight, understand trends and track key humanitarian issues? Through a collaboration with Microsoft and UN OCHA, Project Fortis was created to accelerate the surveillance around humanitarian disasters and health epidemics around the world.

This talk discusses the architecture of a high-available native spark pipeline running across multiple Kubernetes clusters to support Fortis customers.

Big data used to be synonymous with Hadoop, but our ecosystem has evolved over time with new database, streaming and machine learning solutions which don’t necessarily benefit from the Hadoop deployment model of Map/Reduce, YARN and HDFS. These solutions require a generic cluster scheduling layer to host multiple workloads such as Kafka, Spark and TensorFlow, alongside databases such as Cassandra, Elasticsearch and cloud-based storage.

Eliran Bivas is a senior big data architect with years of hands-on experience working on both big data and cloud native solutions. Eliran will go over a common solution framework to create cloud native end-to-end analytics applications. It involves using Kubernetes as an alternative to Yarn, running Spark, Presto, machine learning frameworks (TensorFlow, Python and Spark ML kits) and serverless functions coupled with local and cloud-based storage. The session will showcase customer use-cases from IoT, automotive, cloud SaaS and finance. It will also include a live solution demo which demonstrates the benefits of using big data and analytics over a cloud native architecture, eliminating the existing challenges of complexity and moving towards a continuous integration and development architecture for big data.

In this talk we will demonstrate an approach to management of kafka clusters in kubernetes deployments. We will show how we can provision kafka clusters and configure it using kubernetes concepts and an operator process. The kafka and zookeeper cluster elements will be provisioned using StatefulSet. As these applications benefit from high performance storage, we will also show how we can use node selectors or persistent volume claims to schedule instances on correct hardware. In order for clients to use it, the necessary message topics have to be configured in kafka cluster. We will show how using an operator process, based on kubernetes custom resources or ConfigMaps we can manage this configuration in descriptive manner and ensure consistent configuration across different development and operations stages as well as cluster restarts. Finally we will discuss how all this ties in with service catalog.

Distributed databases can make so many things easier for a developer... but not always for DevOps.  OK, almost never for DevOps.  Kubernetes has come to the rescue with an easy application orchestration! 

It’s straightforward to do the orchestration leaning on relational databases as a data layer. However, it’s becoming a bit trickier to do the same when a distributed SQL database or other kind of distributed storage is used instead.

In this talk you will learn how Kubernetes can orchestrate distributed database like Apache Ignite, in particular:

• Cluster Assembling - database nodes auto-discovery in Kubernetes.
• Database Resilience - automated horizontal scalability.
• Database Availability - what’s the role of Kubernetes and the database.
• Utilizing both RAM and disk - set up Apache Ignite in a way to get in-memory performance with durability of disk.

One of the largest challenges facing the machine learning community today is understanding how to build a platform to run common open-source machine learning libraries such as Tensorflow. Both Joy and Lachie are both passionate about making machine learning accessible to the masses using Kubernetes. In this session they'll share how to deploy a distributed Tensorflow training cluster complete with GPU scheduling on Kubernetes. We’ll also share how distributed Tensorflow training works, various options for distributed training, and when to choose what option. We’ll also share some best practices on using distributed Tensorflow on top of Kubernetes, based on our latest performance tests performed on public cloud providers. All work presented in this session will be accessible via a public Github repository.

Database systems remain the last frontier for Kubernetes, and at the Patroni Project we're working on conquering it. Having fully automated PostgreSQL clusters using Patroni, the project is now working on making Patroni more "Kubernetes native", so that SQL databases can be seen simply as a PostgreSQL resource.

In this talk, we will explain and demonstrate the current projects integrating Patroni PostgreSQL with Kubernetes, including:

* Patroni Operator, using the CoreOS Operator pattern
* Kube-native Patroni, which uses the Kubernetes controller instead of its own management

These works in progress will both acquaint attendees with tools they can use for their own high-availability database architectures, and explore some areas where Kubernetes could improve to support database systems better.

Stateless, cloud-ready applications are the future for many enterprise users, but what do you do about legacy monoliths, and existing vendor applications? New StatefulSet features within Kubernetes allow developers and administrators to work with these types of applications, and still reap the many rewards of a containerized platform. This session will explore some of these features by deploying a full MongoDB cluster on-top of OpenShift.