Breaking

Monday, July 30, 2018

Google Cloud Next postmortem: The enterprise journey continues

Google is serious about convincing you that it's an enterprise technology company. Beyond the hiring of customer-facing engineers, it is upgrading its tooling and veering higher up the food chain. But this journey has just begun.

Google Cloud
Google Cloud Next postmortem: The enterprise journey continues
As the player that rounds out the top three global cloud platform providers, it is hard to ignore the sheer scale of resources that Google is investing to plant its beachhead in a market where the top two have already carved huge presence. It still trails distantly -- somewhere between $1 and $2 billion per quarter, which is a fraction of what AWS and Azure pull in.

The wild card, however, is that the game is hardly over, as there remains 80 - 90 percent of enterprise workloads running on premises that are still up for grabs.

It is playing the long game to change your impression of just being a technology company that in the past has primarily targeted developers. So, Google Cloud CEO Diane Greene reiterated the "enterprise" message this week at the annual NEXT conference that has just wrapped up in San Francisco. It points to growing traction among enterprises -- although digital upstarts like online gaming still steal the headlines. A key pillar of that message is its security and networking capabilities that have been singled out by some analyst firms.

But for now, Google looks more like a different type of "enterprise company" where customer-facing engineers outnumber salespeople. Maybe that's a blessing in disguise?

The blessing and curse of Google's reputation for advanced technology and AI is that implies a need for enterprises to transform if they are to truly benefit. Case in point? Recall our case study of Optiva, a telco billing solutions provider that is staking its turnaround on the Google Cloud Spanner globally distributed transaction database. Moving from Oracle, it had to refactor the code that had been sitting in database stored procedures back into the application because of Spanner's lighter weight architecture, and it had to change the way it laid out data. But in all fairness, if you embrace cloud-native platforms like Amazon Aurora or office suites like Microsoft 365, your organization will also have to embrace change.

But a subtle pivot could be found with Google Cloud's initial foray into what it terms "solutions." That's Google-speak, not for competing in the applications market or selling direct to enterprise customers, but for leveraging its AI to add smarts to partner applications. So, the first of the solutions to roll out is Contact Center AI which provides intelligence to hybrid chatbot/human call center processes by using natural conversation to streamline the "phone tree" process and then prompt live representatives with pertinent information -- all based on machine learning. Google has signed up a mix of call center and systems integrators as partners for its first solution venture. There will be more. On the show floor, we saw prototypes for a handful of other areas like recommendations systems that may or may not make the cut.

Part of the challenge of broadening out the enterprise focus is expanding the addressable audience. For instance, cloud rivals Amazon and Microsoft, and other analytics tool providers, Google is aiming to expand the audience for AI beyond data scientists. With the latest round of announcements, it still takes a different approach from Amazon and Microsoft in the breadth of its services for curated machine learning. But it has expanded Cloud AutoML from machine vision to natural language text and language translation services, leveraging the APIs that it already offers to developers.

An interesting announcement was the unveiling of BigQuery ML. It allows you to run machine learning routines (for now, only linear regression and binary logistic regression models are supported) inside BigQuery without having to move data.

But the messaging around BigQuery ML was a bit unclear. Yes, you can now run machine learning from a SQL program. That made it sound like any SQL developer could now invoke machine learning, but in reality, BigQuery ML is more of a subroutine facility that lets you insert machine learning models into your SQL code. So, this is not machine learning for the SQL crowd -- you'll still need to be a data scientist to take advantage of BigQuery ML. But what would be cool is if Google could marry AutoML with BigQuery ML -- now that would really open things up for the SQL folks.

Another part of Google's enterprise message is that it has made the biggest open source commitment of all the cloud players. That reflects the reality that many enterprises are embracing open-source first strategies where viable open source alternatives exist. Developers have voted in mass for open source, as it makes their skills far more portable.

Google's flouting of its open source cred is a bit ironic as the conversion is still recent -- it used to publish research papers, leaving it to others to conduct clean room development for open source. For instance, the basic building blocks of Hadoop may have been conceived at Google, but it took Yahoo, Facebook and others to perform the heavy lift to get them into the Apache community. But with Kubernetes and TensorFlow, both of which have become de facto standards, Google has sent unmistakable signals that it is now all in on the open source.

But as to being the most open source-friendly cloud? Well, that would get some strong pushback from Microsoft, especially in the wake of its GitHub acquisition.

Part of the process to becoming enterprise-grade is unifying and simplifying your tools, so devs and ops don't have to pluck raw standalone tools and constantly shift screens. The unveiling of Google Cloud Platform shows that Google is beginning the journey to integrate, unify, or in some cases, converge its tooling to a single pane of glass. The same goes for Google Kubernetes Engine, where the pane of glass can now be extended to managing Kubernetes environments on premises as well as the Google Cloud.

On the database side, Google is part way there in making the administration more refined. BigQuery recently introduced a new web UI to simplify creating and managing the database, although there are some operations like ingest where the flakiness of the stateless browser interface would not make advisable. Cloud Spanner has introduced new Avro import and export capabilities, but it still lacks the automated schema migration tool that would make standing up a new Spanner instance more seamless (Amazon and Microsoft already have such tools).

As we said, this is a journey.



No comments:

Post a Comment