This post was spawned by a conversation John Furrier and I had with Google Cloud Spanner Product Manager Deepti Srivastava and Optiva’s CEO Danielle Royston, a customer of Cloud Spanner.
What you need to Know
Optiva, a company that sells software to communications providers, was in a turnaround situation an needed to throw a “long bomb” to get back in the game. It decided to completely change its technology approach and looked to Google Cloud Spanner to replace its Oracle database. The company’s CEO says they get a 10x performance advantage at 1/10th the cost from Spanner relative to Oracle.
Spanner is not going to put Oracle out of business. But it will compete for a new set of emerging probabilistic workloads that require greater scale and lower cost than Oracle can deliver. Read David Floyer’s excellent evaluation of Spanner v. Oracle with commentary on AWS Aurora, IBM DB2 and Microsoft SQL Server to get the architectural and business fit story. You’ll also see where Google Spanner falls short and some key questions with which customers should be pressing Google.
Cloud Spanner is the Real Deal
In our interview, Furrier stressed that Cloud Spanner is not the trial-run of new concept, but rather one that is hardened and well-founded from years of internal Google work and the “internal Googlers:” the customers themselves. Deepti Srivastava elaborated, “we’ve been using it internally for about seven-ish years…we’ve been using it for the most mission-critical systems within Google, from AdWords to Gmail to Google Photos and so, it’s definitely had a lot of use internally and…it’s actually been tried and tested and battle-tested, really, over the last many years.”
Danielle Royston from Optiva chimed in with the customer perspective, “we make software for Telcos…this is the monetization software for Telcos. So every time you are making a phone call, you know, downloading a video off of YouTube, the Telcos are trying to monetize that and track it to charge the different customers.” Up until this point, Optiva was running on Oracle, “to date, [it] was the world’s best database. An every Telco out there wants 10 times the scalability of Oracle,” she continued, “they want to get off the 10s and maybe 100s of millions of dollars they pay Oracle every year. And when we looked at Spanner, it was not only capable of replacing Oracle, but it gave us that ten times faster at the reduced cost, right? And so this is a game-changer for Telco.”
Spanner: The Price is Right According to the Customer
When Furrier pushed Royston about the cost, she readily clarified, “The Telcos have their data centers, they like to, I like to say, hug their servers, they like to see their servers. And every time they capacity plan, they have to plan nine months in advance. When you look at Google Cloud Spanner, and then Google Cloud,” she continued, “now we’re not having to pre-purchase that. It can literally auto-scale…and the Telcos do not have to pre-purchase all that compute power. So, that’s a lot of savings.” She added, “this isn’t marketing fluff. There’s really technical differentiation with Spanner. It is a different database, and like I said, we would have put our entire engineering budget to make this work.”
Google Claims: Spanner Makes the Switch to Cloud Easy
According to Google’s Srivastiva, “this is a great example of how the older industries are moving to the Cloud…Google Cloud Spanner provides that disruptive technology for them to really make the move.” Royston concurred, calling the switch to Spanner a no-brainer, “for us it was the low latency at the global scale. Sitting, tired and true, as we’ve been discussing, product.” She added, “my customers are running their business. This is their revenue, and so to get them off of Oracle, I need to have eye-popping numbers, right? Needs to be 10 times faster, and a 10th of the cost, for them to move, ‘cause if it’s two times faster, they’re going to stick with Oracle. If it’s just as expensive, they’re going to stay with Oracle. So [to] my engineering teams, I’m like, I don’t care what you need to do, figure it out, ten times faster, tenth of the cost, let’s go do it, and thankfully it was really easy.”
Royston claimed the switch to Spanner was easy. Skeptics will rightly point out that a rip and replace of Oracle isn’t trivial and the vast majority of workloads aren’t going anywhere for the time being. But in this case, the customer was able to pull it off.
On the question of data consistency, Srivastava responded, “It is the only database in the world that does strong consistency across the globe…there are products that can do strong consistency within a particular region, or within even a continent. But doing strong consistency with performance across the globe, using your database…as a Content Distribution Network for your transitional ACID-consistent data…that’s what we do.” She went on to discuss another issue, “you don’t have to have the big applications…there’s a lot of customers [that think], I have to have a petabyte-scale database or I have to have global data. And you don’t, you can start very small, you can have a few gigabytes. But the point is you get all of that infrastructure and even if you grow to all of that scale, you can just turn up the dial and get it, right?”
Not Your Grandfather’s Database
According to Google’s Srivastava, “I think we think Google Cloud Spanner is a future of databases, right? Because it gives you that best of all worlds, with scaled performance, manageability, and lower total cost of ownership in most cases,” “And we’re so excited,” Optiva’s CEO contributed, “I mean, I’m always testing my messages with my customers and I’m turning those heads and I know I’m onto something really, really great, and it’s thankfully to Spanner, Kubernetes, and Google Cloud.”
The Rest of the Story
Notably, Google’s Spanner is optimized for consistency and global scale. It’s not designed for ultra-high availability to run workloads that can never go down. Functionally, Spanner is new and as a general purpose database it has a ways to go. Specifically, Wikibon’s functionality assessment is that Spanner is limited, particularly with respect to OLAP capabilities and resource access controls. For mission critical workloads that can never go down, Oracle will remain the leader for decades and it’s unlikely that Spanner will eat into even five percent of Oracle’s installed base. But Spanner will compete for a much higher growth class of emerging workloads that are non-deterministic yet still require ACID capabilities.
Google, like AWS is investing heavily and over the long run will siphon more and more opportunities as cloud becomes the norm for running infrastructure as code. To compete in the enterprise, however, Google must do more than just deliver great tech. It needs to provide levels of enterprise support that heretofore have not been demonstrable in the company’s DNA.
Nonetheless, the tech behind Spanner is impressive and exciting. You can check out the interview with Google’s Srivastava and Optiva from Google Next 2018 where Google sponsored theCUBE at its big cloud event.
