"Big Cloud" doubles down on Serverless

This article found on TheCloudIsServerless.com, you can view it here. Please subscribe for more news!

The past few weeks have been particularly interesting for “Big Cloud”. Now when I say “Big Cloud”, I am not simply referring to the three hyperscalers: AWS, Azure, and Google Cloud Platform. There are other major companies that are contributing to the larger cloud ecosystem that should be counted.

For example, CloudFlare stated in 2022 that 20% of the web and 30% of Fortune 1000 companies use their services. Last week we talked about Alibaba Cloud and IBM and Oracle both have cloud services. There are a myriad of cloud services that I would consider big players in the cloud aside from the “Big Three”.

The reason I bring this up is because I noticed a few interesting stories regarding serverless in Big Cloud. My employer just had their annual conference called Google Cloud Next in Las Vegas. There were a ton of announcements, many around AI but one in particular stood out to me in relation to serverless computing.

Google Cloud Run Application Canvas was announced as part of Next. “What does it do” you may be asking yourself. Picture this, I can tell an AI chat bot to “create a web frontend with a PostgreSQL backend” and it will build the integration for you. You will still need to provide your containerized code but the integration between your service and the backend will be simplified for you.

My thoughts on this is that it’s great. The whole idea with serverless is to ultimately simplify the experience for developers. This is a great way to utilize AI to further assist. In the past, you would have to do some manual tweaking. It wasn’t the end of the world but it was some extra YAML and/or clicking in the console. It also integrates with Vertex AI, giving your serverless applications access to models such as Google’s Gemini, partner models like Anthropic’s Claude and open source models like Llama or Mistral.

But enough about Google, you don’t want to hear me talk about my employer, after all, it will come across as biased.

CloudFlare made some news recently too. For one, they acquired the company BaseLime. Now admittedly, I am not too familiar with BaseLime so I had to do some research. They are a cloud-native observability company. Since the rise of Site Reliability Engineering (SRE), the idea of observability has been on the rise. In short, it’s the ability to look at your application in terms of historical data (logging) and in real time (monitoring) in order to ensure the best experience for your end-users. The idea originated in Google but has since taken a life of its own.

What made this company special is that it had a special niche with serverless applications. In many ways, CloudFlare is a serverless company so it only makes sense that they’d make this acquisition in order to better serve their customers. After all, if you are gonna build a global application on CloudFlare, you better see how it’s working.

Another interesting story out of CloudFlare are their new Workers AI going public. These workers will enable access to HuggingFace, an open source AI/ML platform. HuggingFace essentially makes it easier for your applications to interface with an LLM. People are building apps using HuggingFace and now those apps can be hosted on CloudFlare with Workers AI.

This was announced about seven months ago but was more or less in beta. Now it is GA and ready for production use. I have a strong stance about using beta products in production workloads. Usually this is because beta products offer no SLA and so it makes it difficult for you to provide any guarantee to your customers.

This news is important to me because, I don’t if you noticed but AI is a big deal today. But when we talk about AI, we talk about developing models and the cool stuff it can do. People often forget to talk about the “how”. “How do I build applications that can inference the AI models”.

Serverless is the perfect infrastructure as it simplifies the experience. Inference-as-a-Service is only going to grow and serverless architecture is the best way to implement and scale that.

Photo by Aleksandar Pasaric: https://www.pexels.com/photo/view-of-cityscape-325185/