Modern Hadoop

- [Instructor] As we begin our journey looking at Advanced Hadoop, we're going to start with what I call Modern Hadoop. You might be surprised to know that both open source and commercial Hadoop is over 10 years old. What I'm finding though that's driving the adoption velocity are innovations, both in public cloud services around Hadoop such as those offered not only by Amazon but also by Microsoft and Google, and also innovations in other devices and domains. Once that I've had experience with over the last 12 to 18 months are IoT, or Internet of Things, and Genomics around bio and formatics for personalized cancer treatment.

Now we'll talk about these domains but I'll also generalize because the maturity of the Hadoop ecosystem has more and more applicability. Let's look at the core components, kind of as a review. So we have storage, compute, and management. In the area of storage we have Hadoop File System and then we have some vendor optimizations, one will be using this course is around the vendor, Data Bricks, which has a commercial version of Hadoop with the library Apache Spark, and that's the Data Bricks File System. We'll also look at using cloud based file systems, such as S3 from Amazon and the Google Cloud Storage File System.

These are called Data Legs. In the compute area, my expectation as I mentioned in an earlier movie is that you will be familiar with the core map reduced paradigm that Hadoop was originally built on. We're going to be focusing on some of the newer compute libraries that are available. In particular we're going to look at Apache Spark, which allows compute processes to be run in the memory of the worker nodes and significantly increases the processing speed of the Hadoop jobs. We'll also look at some of the other popular libraries out there, such as Storm, or now it's becoming Heron from Twitter.

In addition to this, we'll consider some of the management capabilities of Modern Hadoop. These include Yarn and Mesos.