Google Cloud Platform (GCP) is the clear number three player in public cloud behind Amazon Web Services (AWS) and Microsoft Azure. But it’s standing on a tripod of data, machine learning, and cost advantages, and is counting on those three areas to help them build market share and scale.
At the annual Google Cloud Next ’17 conference, Google did an impressive job of combining those three areas into a message around a fully managed platform that makes each one more accessible and easier to manage than its competitors. The conference was a blend of new product announcements (more than 100 of them), education, networking, customer testimonials, and sales pitches.
Here are the three things that stood out the most at Google Next.
1. Committed use discount
Google fired a shot across AWS’ bow with its new Google Cloud committed use discount program. Under this program, organizations can commit to consume bundles of compute in the form of vCPUs and memory, which are not tied to an instance type, just a region. This is significantly simpler than AWS Reserved Instances– a point Urs Hölzle drove home with reference to the amount of work AWS customers must devote to managing reserved instances.
AWS must have noticed because only four days later it responded with changes to increase the flexibility of its Reserved Instance plans–changes that still left them with a more complicated model than GCP.
With flexible instance types and per-minute billing, Google’s model appears simpler to navigate and less expensive to operate than AWS. The committed use discounts certainly seem easier to understand and use than the various permutations of reserved instances and committed consumption programs at AWS. Of course, committed use is a new offering, so we’ll have to wait to see how the cost models play out in practice with real workloads.
2. Cloud Spanner
Google’s Cloud Spanner database impressed me more than anything else I heard about at Google Next. It promises a relational database, running standard SQL, which supports high volumes of activity and is accessible around the world, with fault tolerance.
It’s a potential game-changer in the world of relational databases, bringing the scalability and performance of NoSQL databases to the relational database structures many enterprises need. NoSQL databases, while powerful and high-performing, require a fundamental change to the way data is structured and accessed.
If Google can actually deliver a relational database using standard SQL that delivers performance and fault tolerance without developers needing to think about replicas, log shipping, always-on availability groups, and other arcane tools, it should have a winner in Spanner.
Spanner’s impressive capabilities are delivered by leveraging Google’s massive networking, atomic clocks in each data center, GPS devices on individual racks, and some rather sophisticated engineering. Spanner, in some form, has been used internally at Google for around 10 years, helping to power its AdWords and Analytics businesses. These are high-value, high-volume products, meaning that this Spanner is more battle tested than most first-generation products, and may give enterprises more reason to believe it’s ready for production workloads.
The best analysis of the platform, its capabilities, and limits I’ve seen thus far comes from San Francisco-based startup, Quizlet. In a blog post, Quizlet concluded that “Cloud Spanner is the most compelling cloud service we’ve seen for scaling a high-throughput relational workload, MySQL in our case. It has some rough edges as a production system, but its latency and scalability are unique.” The blogger points out that Spanner isn’t ideal for all use cases: While it can be an expensive way to run small databases, it shines at scale.
3. Machine learning everywhere
Machine learning was evident in many places at Google Next, and this is clearly where Google feels it can distinguish itself from AWS and Azure.
Fei-Fei Li, Google’s Chief Scientist of AI and Machine Learning, offered an impressive demo of the new Cloud Video Intelligence API, which can scan and index video, and let you do things like “find where in this collection of videos there’s a dog, or someone running.” She also announced the general availability of the Google Machine Learning Engine, which fits well into Google’s message of democratizing access to these powerful tools and techniques, and making it more practical for smaller groups to leverage them at scale.
Machine learning also can be found in several new G-Suite features, such as its ability to predict what file you’ll need out of Google drive (saving you from a manual search). The new Data Loss Prevention API appears to leverage some of the same image processing and scanning tools. In one demo, the presenters successfully recognized and redacted the digits of a credit card number in a picture of the credit card sent over a chat session.
The potential for Google to flex its machine learning muscle and integrate it behind the scenes with its other solutions can give them an attractive advantage. AWS and Azure are apparently trying to move in similar directions.
That’s a wrap
Google’s cloud platform is trying to move further up the stack than AWS, and includes a stronger focus on fully managed tools. While the dashboard contains significantly fewer tools, Google Cloud will manage many of them for you.
It’s going to be interesting to see how these different approaches pan out over time.