Welcome to GCP This Month! This month, we’re taking a look at some of the top takeaways from the Google Cloud Next ’21. For this post, I’m joined by fellow GCP Gurus Broadus Palmer and Joe Lowery to chat about what announcements we found most exciting and which sessions they found most memorable.
Gain new GCP skills
Get started with A Cloud Guru today and learn by doing. ACG can help you master new skills and transform your career with courses and real hands-on labs around Google Cloud, AWS, Azure, and beyond.
Mattias: I would love to hear what your favorite announcements were. What stood out to you guys?
1. Project Euphonia
Joe: Well, I have to admit something: there wasn’t really a favorite announcement. Like most conferences, a lot of times, I go for the demos. I wanted to see something hot and fresh and new and great, and they delivered for me.
At the opening of the developer keynote Urs Hölzle, senior VP of technical infrastructure for Google, started, by speaking in five different languages, one sentence right after another: English, French, Hindi (I think), Spanish, and then German. He was not fluent in any of these except probably English and maybe German, but his voice was used as the model, and that speech was reproduced so that he could speak.
It’s the result of Project Euphonia, a spinoff of Google Research and DeepMind, where they use their speech-to-text technology, trying to help people who have difficulty being understood, be more widely understood.
So at first the digital devices can understand them, then we can bridge that gap and have people understand the digital devices. I thought, “Wow, wouldn’t this be a really wild kind of down-the-road application where maybe we would see language dubbing in an actor’s own voice?”
2. Vertex AI workbench
Broadus: For me, it was also diving more into machine learning and AI, which is something that I’m trying to understand more about. One thing I liked was the Vertex AI workbench. Not only learning about that, but Google Next did a quick start on it. They took a customer lifetime value model. They used BigQuery for analysis and then used TensorFlow with Vertex AI.
They analyzed how much a customer was spending, how often the customer was spending, and they created a baseline from that. Then they created a training model using Docker containers with TensorFlow and Vertex AI. This was to serve as an endpoint for the model to use what Vertex AI predicted. Going through that quick start really helped me understand more about Vertex AI and how it’s used from the managed notebooks that you can use with a TensorFlow enterprise, from a new notebook that you could create on your own as well. So it was a pretty amazing tool, and I’m glad I took that deep dive to learn more about it. I feel smarter already!
Joe: This whole emphasis on AI is really one of Google’s main thrusts straight across the business, almost everything that they do. They’re really trying to bring this in more and more, and you can see it begin to bubble up in different areas and different services across the board. That’s really exciting. So, Mattias, what was your biggest takeaway?
3. Google Distributed Cloud
Mattias: I’m always a sucker for all the serverless announcements, but I’m sure we’ll get to those in a minute as well. But one of the things that did stand out to me was the Google Distributed Cloud.
Google Distributed Cloud is a really interesting packaging of all of their different ways of managing stuff everywhere for you. I was going to mention how much I love managed services when you were talking about the Vertex AI Workbench, because then doing the work for us is really what I appreciate. So having them then reach out into so many more places with Google Distributed Cloud — I think that’s a really interesting thing. They had four categories of locations where they’re managing things for you. There’s the Google network edge with all of their points of presence, 140 of those around the world or whatever.
Mattias: There’s the operator edge where they interact with telecoms and manage responsiveness right at the edge of cell phone and other connected devices. There’s the customer edge — like on a factory floor or in a retail business that store or something like that where they need to do processing right there and have a responsiveness or offline capability or whatever, possibly. And then also the customer data centers. So not even just Google’s cloud but the customer data centers, colo facilities or whatever, and that’s not even getting into how they have now a new API for multicloud.
They demoed at one point deploying to both GCP and Azure through the multicloud API for Anthos and they’re just, they’re rolling it out everywhere, right? So if you want to have Google manage stuff for you, that’s the way to do it. I thought it was a lot of really interesting things that came together to to sort of put that, um, that package together for, for a lot of organizations that are looking at multi-cloud and wanting that sort of functionality, but also want the benefits of not having to manage stuff.
4. Google’s multi-cloud focus
Joe: You know, I think this really does reflect a lot of Google’s worldview and something that you and I have talked about often; they don’t look at it like “Hey everybody come onto our platform, and we’ll just teach you our platform.” No, what they want is to have people know how to work in the real world. They realize that not everybody is on the cloud, but people are moving there. And so they want to make that easier. They’re basically going to where the customers are. Whether you’re dealing with a hybrid cloud situation, working with your on-prem and hopefully Google Cloud (but maybe not) or trying to go multi-cloud and take advantage of some services that Azure has that are better or AWS that are better. And combine those with the Google Cloud services that are top-notch. So, this whole approach, this distributed cloud way of working, I think to me this is the Google worldview. This is how they see reality.
Broadus: Yeah, I totally agree. Especially when it comes to data, too. Siloing data even on a cloud environment is error-prone as well. So understanding that to allow their customers to query data based on different cloud platforms is why BigQuery Omni is now in GA. So this is something that I think Google realized a long time ago, even with creating Kubernetes.
Get the Google Cloud Dictionary of Pain
Speaking cloud doesn’t have to be hard. We analyzed millions of responses to ID the top concepts that trip people up. Grab this cloud guide for succinct definitions of some of the most painful terms in Google Cloud.
5. BigQuery Omni
Mattias: Broadus, I’m really glad you mentioned BigQuery Omni, because I think that BigQuery Omni is really interesting in what it represents.
We’ve all heard of BigQuery. We’ve seen how much BigQuery is loved. Being able to query against all of your data and even use ESQL queries in a serverless way and being able to access BIgQuery through BigQuery Omni on AWS and Azure, just having that generally available is really awesome. I’m really excited that that is now available, but I’m also really excited because I think that it means that Google is taking a look at other technology that they’ve got. Those Google engineers are obviously really smart people, and we love what they’ve built. We love GCP because of that great stuff. I think that they’re looking at taking more of those things out to the other clouds as well. I’m definitely interested to see what they bring down the road.
Broadus: Google is now becoming a platform, and it’s not about, “Hey, come to our platform and use our products,” but what they’re doing is figuring out what the other cloud platforms need — what their customers need. So now, AWS and Azure would need their services to operate. So, using Google services, now they’re just collecting data and analyzing how it works, how it helps, how it integrates. Then innovation will come where they’re bringing a service that is going to be needed by those cloud platforms themselves. So, eliminate the competition, but collaborate. That creates a bigger cloud market. That’s where I see it going.
6. GCP Serverless
Joe: Oh, and they also have with BigQuery Omni taken an existing service and expanded it. You mentioned serverless before, and I know you’re kind of chomping at the bit to get to the serverless things. So I’m going to beat you there because I agree, and I took a look at what’s up with serverless.
That was fascinating because they said we went through all the big changes for Cloud Functions and Cloud Runs, to a lesser extent, Cloud Build. But they really are taking their strengths and building on those and making them better than ever.
So, for example, in Cloud Functions, they’ve added more runtimes. They have Ruby, .NET Core, PHP even, and they’ve updated their Python to more recent versions, like 3.9 and node.JS — they are even previewing version 16 of that. So they’re really moving forward with that, but they’ve also expanded what Cloud Functions can do, how you work with them for build customization. So you have private worker pools. Rather than having it completely out of your control how the Cloud Function is built, you can now specify a private worker pool and even build environment variables for that. One of the best things is that they’ve added minimum instances to the functionality, which will help people avoid the cold start of a function that spins up very quickly. But still, if you have a warm Cloud Function already set to go, you completely avoid that cold start timeframe.
Mattias: Well, as much as I’m a serverless purist, I was gonna mention that being able to set minimum instances for Cloud Functions is actually a good thing. Same with Cloud Run, actually, where you can pin it and have them running to do background processing and all that sort of stuff.
It really does fit circumstances that aren’t just the event-driven sort of a model, and I think what it does is help more organizations adopt these technologies in a way that works for them. I don’t know how many times that comes up in conversation in dev teams like, “Oh well, what are we going to do? Like, do we have to reimplement the whole thing?” And I mean, how often does that get approved, right? Unless it’s a super quick thing. It really doesn’t. So being able to get that much further forward by adopting what works for your circumstance and then taking a look at tweaking things, changing them down the road, incremental progress is always a great strategy to consider, right?
7. Cloud Run
Joe: You mentioned Cloud Run as part of some of the enhancements that are coming. I was really also taken with what was happening there.
Now they have what they call a second-generation execution environment that you can opt for, which offers increased network and CPU performance. They have full Linux capabilities and even network files system support. I thought that was kind of phenomenal.
As you said, they now allow you to develop locally your Cloud Run apps with a new emulator, and they can even deploy locally with a three-word command. I mean, you’ve got to love the three-word command! I’ve always been kind of jealous of App Engine. You go gcloud, app, deploy, and ZING! And all the other GCloud commands are 15 words with double dash flags and everything. So Cloud Run now has climbed on that ship, and now it can slide this flag GCloud run deploy. So they’re often running very quickly, and it’s taking a service that’s getting a lot of traction and improving on it and adding a lot more security to it as well. Bolstering security across the board also seemed to be a theme of the conference.
Mattias: Well, I definitely appreciate the short command lines because I pay per character. But the usability from a developer’s perspective is a really big deal. So not having really complicated things and being able to then deploy in debug locally in an emulator, that’s a really big thing. For a long time, the development story of serverless has been one of the harder parts of going in that direction and that strategy, and I love that these things are now getting so much more attention andsupport.
8. Google Workflows and Document AI
Broadus: Yeah, that’s an interesting point, Mattias, and also working with Google Workflows as well, where you have now implemented HTTP callbacks, GCP API connectors where at Google Next they were talking about working with Document AI, the new API from GCP that’s really going to be more of a competition with a service like Typeform. Where now you can now use artificial intelligence and typical machine learning to understand the type of documents you need. And look at where a document structure and how your documents should be placed as far as whether you have invoices, contracts, anything that you’re really presenting. Workflows is a great tool that’s really connected with that and more memory and concurrency as well.
So I’ve seen some exciting stuff with Google Workflows, especially with the Document AI which I think is worth checking out. They have some GitHub that you can pull from. You can even do a deep dive into how that works with creating the service accounts and deploying and see what that looks like in real-world scenarios.
9. First-class Spark support
Mattias: Another cool thing was the first class Spark support because now Google will manage Spark for you. I like when Google manages stuff for us. So one of the things is serverless, even, so devs can just ignore all the clusters and all that. You just submit a job, and then Google will take care of provisioning clusters and auto-scaling it all for you making sure that that job runs. So taking away all of that stuff you don’t need to care about is definitely a great thing. And there’s also first-class Spark support in other places. Now, these are in private preview, but you can get it in BigQuery, in Vertex AI, and DataPlex. So those are definitely some extra interesting goodness for people who use Spark.
10. Must-watch segments
Broadus: I want to talk about the must-watch segments on Google Next ’21.
For me, a must-watch segment was the interesting conversation with Vince Cerf and Jim Hogan on disability in tech. Jim Hogan talked about being autistic and not really being comfortable with putting that out there to his peers and his employers. But at Google, he found a home where not only did he feel accepted and nurtured but he felt that Google was actually making a statement by training other employees to be able to understand what it’s like for people with disabilities and how to communicate and connect with them. That was a pretty awesome session to look at and just to see Google take that initiative that many companies wouldn’t even dare to try to understand. Being able to have that inclusive ecosystem to help everyone thrive — I think that’s huge. Kudos for Google.
Mattias: I caught that session too. It was definitely a great one to watch, and it gives us the ability to take a look at some of our assumptions when we’re interacting with people. What are we assuming about that? And how can we do a better job?
Joe: I was going to go back and recommend that people watch the developer keynote if they hadn’t had an opportunity to do that, because not only does it have that great demo that I mentioned with the text-to-speech and different languages and a nice reference to Project Euphonia and a link to that is very interesting in and of itself, but they really hit the main key points of why they are a company for developers, and they went over a lot of the toolings, which you may be familiar with — like Cloud Shell — but went into it with some more depth, really explaining and revealing some things that I had not known or had not remembered or some of the capabilities of it. And that’s always expanding.
One other thing that I wanted to mention was security is really a big thread throughout the entire conference. It was specifically mentioned that Google is now adhering to SALSA, which is an industry-standard, not just a delicious dish, but they are using it for their Anthos Service Mesh and their Cloud Build hybrid systems, which allow full compliance and allow Google to implement binary authorization so that a policy can be set to make sure that what’s being deployed is something that your company wants to be deployed.
Mattias: Yeah, binary authorization is one of those things that doesn’t sound very exciting, but it actually protects the business and the development pipeline. I’m really glad to see also that they’re incorporating that more places. So that’s cool.
You mentioned security was one of the themes. I think we also talked about diversity and inclusion as a theme that came up. Another theme I saw a lot, actually, was sustainability and environmental impact. I saw that they have a number of new tools for people to use to identify their environmental impact with the workloads that they’re running.
Broadus: My final takeaway is not that Google was focusing on more new services this time, but more improvements to the existing services and the relationships they’re trying to build with other companies and other cloud services providers. Then I think in maybe a year or two to come innovate some more solutions and help other cloud service providers thrive in the ecosystem. So it was all about sustainability, security, more about artificial intelligence and machine learning where the future is going to be. And Google has it all in their hands.
Joe: One of the key innovations that none of us have talked about, and I think we really should focus on, is how Google has managed to just improve the efficiency of the conference. I mean, last year, what was it? Nine weeks of conferences? And now look at them. They’ve got it down to three days. I mean, that’s phenomenal. You’ve got to admire that efficiency.
Mattias: Well, I have to thank you both. I really have had a great time chatting with you about Google Cloud Next ’21. That’s all for this month, but we would love to continue the conversation with you. Join our Discord and let us know what you think about Google Cloud Next ’21. Take care, stay safe and keep being awesome Cloud Gurus!