Scaling SQL for cloud-native - Yugabyte CEO discusses the opportunity of a new class of database

Derek du Preez Profile picture for user ddpreez March 28, 2022
Summary:
Traditional relational databases have lost ground to the opportunity seen in NoSQL databases. But what if you can have the best of both worlds? That’s what Yugabyte aims to provide the market.

Image of Yugabyte CEO Bill Cook
(Image sourced via Yugabyte)

Yugabyte aims to give enterprise buyers the best of both worlds, when it comes to choosing their next generation of databases. In the on-premise world, SQL relational databases ruled the roost, but this thinking has since been challenged by a wave of NoSQL adoption in the cloud - led by the likes of MongoDB and DataStax (Cassandra). 

The general message over the past five years has been that relational models aren’t suited to scale in the cloud, whilst NoSQL databases are built for distributing unstructured data for internet applications. However, what if you didn’t have to compromise on things like consistency for transactions, whilst taking advantage of cloud characteristics, such as resiliency, scale and geo-distribution?  

The founding principle behind Yugabyte is that NoSQL databases will aim to become fully transactional, but also that a SQL database that is cloud-native and can handle multiple data models without compromising on transactions, performance and geo-distribution, could be superior. 

Yugabyte was founded by a team at Facebook, after they were able to scale the social media company’s platform from millions to billions of users in just a few years. Yugabyte has since chosen to build on PostgresSQL, an open source database that has been around for decades, and rebuild the ‘upper half’ to cater it to cloud trends. 

The ‘NewSQL’ company (a term often used to describe the market Yugabyte operates in), has since received over $290 million in funding and has been valued at more than $1.3 billion. Not only that, but it has customer names that include Kroger, Plume and Admiral. 

diginomica got the chance to sit down with Yugabyte CEO, Bill Cook, and VP of Strategy and Solutions, Suda Srinivasan, last week to discuss how the company is thinking about its go-to-market strategy and what the opportunity is for buyers. 

Cook explains that the success seen thus far is being driven by a number of cloud-based market dynamics. He said: 

The database tier, for the most part, had not been designed or built for this era - cloud native patterns - which means you can connect a database service per micro service, or app. And then, obviously, resiliency, redundancy and geo distribution, all those kinds of attributes as well, are important in today's world.

[The way the Facebook team] went about it was basically by saying ‘what the world doesn't need, or doesn't want, is yet another database that they have to deal with’. However, they do need the cloud native attributes of the underpinning database. And so the idea was, on the SQL side, to take the upper half of Postgres and put a more robust underpinning to it, for cloud native resilience and scale out attributes. 

The idea here is that Yugabyte is solving database challenges for cloud native environments, whilst making it easier for developers that have extensive experience using SQL tools to adapt. Yugabyte is built to be infrastructure independent - so the company is focusing on Google Cloud, AWS and Microsoft - but is also provided as a service, a managed service, or can be hosted in your own environment. 

Srinivasan reinforced Cook’s point, but added that companies that had ventured into the NoSQL world to find an alternative to relational databases are now knocking on Yugabyte’s door. He said: 

A lot of companies that are doing database modernization have been using traditional relational databases for the longest time, but they are looking to make a move into something that is resilient, scalable, geo distributed, cloud native, ready for the future.

And then we have other companies that have felt acute pain in terms of scaling or resilience with traditional SQL, have gone the NoSQL route, and then they realize that they have trade offs. 

For example, a large retailer now has their product catalogue running on Yugabyte. They had it on Cassandra, and they found that they did not get secondary indexes, they couldn’t do really fast lookups. Maintaining that stuff took a lot of a lot of effort. 

Cassandra is not a system that is built for operational simplicity at scale, in many ways. And they were losing money because of ACID transactions and the lack of transaction capabilities. So then they were looking for something that allowed him to combine these two worlds together.

Understanding the use case

Cook and Srinivasan said that the need for a relational database that can scale in the cloud is becoming more critical as customers explore multiple use cases that support edge computing, 5G, and IoT workloads. Doing these with traditional databases becomes cost prohibitive and complex, whilst the NoSQL vendors run into limitations, according to Cook. He said: 

There's trade-offs that you make when you move to a Mongo or a Cassandra, right? You give up on the core database transactional features that we've talked about with traditional. On the transactional side, you run into limits on MongoDB or Cassandra. That’s fine for certain applications, but the core applications and having transactional consistency…we would argue, why compromise either side of that equation if you don't have to?

Srinivasan said that the customers that Yugabyte is working with typically fall into two categories. He added: 

We've got the global 2000, large enterprise companies that are going through the process of database modernization. They spent the last 10 years on a digital transformation journey, and they've kind of been focused on application modernization, microservices and so on. And for them, they have database infrastructure that doesn't look like the rest of their infrastructure stack. And they're looking for modernization in the data layer, this is kind of a holistic approach: 2,000 databases, the next five years, we're going to be modernizing, where do we start as a workload? 

Then you have another group of companies that are building microservices to meet immediate needs. So they're essentially going cloud native in one part of their organization and they are looking for a database that is going to be a good fit for that world. And there the pain point is usually more acute. 

Finally, another key selling point for Yugabyte, Cook notes, is the opportunity to invest in talent where it’s necessary. He hopes that Yugabyte can prove its worth at scale - whilst making it easier for SQL developers to transition to a new platform. Cook said: 

The other nuance to me is, if you think about people, I've talked to many enterprises and people are their biggest pain point. How do I find skilled people that help me move the needle of my business? 

Database administration should not be a core competency that I want to invest in, candidly. I just want a database that works and scales automatically in a cloud environment and as a service to my developers. I want as many developers building features that I take to my customers as opposed to worrying about underlying infrastructure.

My take

It’s early, early days for Yugabyte. And we are yet to speak to any customers, although the names it is publicly talking about are impressive. Interestingly, Yugabyte has started at the top end of the enterprise market in terms of wide-ranging use cases for large global brands, whereas the NoSQL vendors typically started by fostering the developer community and then scaling up. How that will play out down the line is too soon to tell, but it’s just an interesting observation. 

I can see the appeal for a company that has traditionally used relational databases and is now being offered an alternative to MongoDB or Cassandra, with claims that it can do everything that they did before, but in the cloud. It’s a compelling proposition, but as I said, it’s early days and we need to speak to a broad range of customers to understand the proof points better. 

My initial assessment is that the Mongo and Cassandra customers I speak to are generally quite happy, but aren’t necessarily ditching relational databases altogether and are making use of NoSQL for targeted use cases. Whether or not they will swap out NoSQL tools for something like Yugabyte, I’m not sure. That being said, moving their older relational databases to something like Yugabyte could be compelling. I’ve got a feeling, like most things, companies will end up with a mixed bag of tools, targeted at what their needs are - but we will be observing this market closely to see how the dynamics play out. 

Loading
A grey colored placeholder image