I am very happy to kick off the Unpodcast blog post series with an interview with my Trexin colleague Arif Khan.
Lorenzo (L): You recently attended the AWS re:Invent conference. What do you think were the biggest highlights of the conference?
Arif (A): There were a couple of hot topics at the conference. AWS introduced a lot of Serverless features. Instead of focusing on dedicated servers, you can utilize the built-in services to deploy your business functions and you are not charged if you don’t use them. There were many sessions on Lambda, their Serverless platform, and all of them were jam-packed. A couple of key things were: the time limit Lambda functions were increased from 5 minutes to 15 minutes and AWS added support for more software development tools such as Microsoft Visual Studio Code.
The second hot topic was machine learning (ML). The big push from AWS was to offer some pre-built models that data engineers could leverage without requiring them to have deep knowledge of machine learning. What I am hoping to see at next year’s conference are examples of how companies have used these models.
L: I was at this year’s Google Cloud Summit and Google’s catch phrase was “the democratization of AI”. They are also offering pre-built ML models. We should compare and contrast these models.
A: I agree. AWS breaks down their ML into 3 different layers. The topmost layer consists of various AI Services such as Amazon Rekognition, Amazon Polly, Amazon Translate, etc. These services are tailored for end user consumption. The middle layer consists of ML Services such as Amazon SageMaker that allows users to train, build, and deploy ML models at scale. The lowest layer gives more power to the user and consists of ML frameworks and infrastructure such as Pytorch, TensorFlow, Keras, etc. These tools allow users to build models using cutting edge platforms. AWS has also launched a ML model marketplace where users can buy and sell pre-built models. It’s very plug and play. You can click on a model and it becomes available in your AWS platform.
L: That’s a very intriguing concept that makes a lot of sense. What are your thoughts on AWS’s announcement of their Graviton chips?
A: I didn’t spend a lot of time looking at that, but it seems that all the cloud vendors are all trying to match each other’s capabilities. If Google has custom chips, AWS will do that too. Another example is Azure OS. Microsoft made it available for on-prem use so customers could use it in their private cloud and then more easily move those workloads to Azure. Amazon now has AWS Outposts that allows customers to run AWS services on-prem.
L: You’ve worked with multiple cloud vendors. How would you compare them?
A: I have not had a lot of experience with Google Cloud Platform I have used Google Storage and found it to be very cost effective. As far as AWS, they have a far wider adoption. Their S3 platform is the platform for implementing Data Lakes. AWS has made it cheaper and easier to use and that has driven its wide adoption. They also have a comprehensive set of tools and depth within those tools. The more people use these tools, the better the tools become. That gives Amazon a big advantage, forcing the other vendors to play catchup.
L: If you had to summarize AWS’s greatest strength, what would it be?
A: It would probably be its ease of adoption and the availability of resources to learn and adopt AWS services as well as the community around all that. Also, their S3 platform. It’s one of the best things that AWS has done. All of the AWS services can leverage it fairly inexpensively, allowing customers to build very cool services on top of it.
L: What do you think is AWS’s biggest weakness?
A: I think Azure has stronger corporate adoption. Microsoft has caught up to AWS on the corporate front pretty quickly and it has an edge due to existing on-premise infrastructure such as Active Directory and Exchange.
L: Thanks for taking the time to chat. We should do it again soon.
A: Absolutely. It was fun.