Q&A: Enterprise Database Trends
Results of a recent survey about databases shed light on enterprise use, priorities, and trends.
- By James E. Powell
- December 13, 2018
Fauna, Inc., which offers the relational NoSQL database FaunaDB, conducted a survey in August 2018 that asked participants about their data stores, including what cloud platform(s) they use to store data and their concerns about scalability and high-availability. Respondents work for midsize companies averaging 1000 people and that manage hundreds of terabytes of data.
We spoke to Evan Weaver, the CEO of Fauna, about the survey results and database trends in general.
Upside: Given how enterprises always seem to be budget constrained (or at least budget conscious), I was surprised that security concerns were more important than cost. Why do you think that is?
Evan Weaver: One of the common threads we see through all the responses is that operational data management is about risk. Across the entire range of SMB to enterprise, the primary consideration for a database adopter is, how much can I trust this system and this vendor? Security capabilities are a critical part of that equation and one that has become more important with increased connectivity, increased data collection, and the move to the public cloud.
We read every day about databases that were compromised and firms that suffered financially and in the marketplace because of their inability to secure their data. Compared to the potential downside of data exfiltration, the cost of the database itself and even its hardware is marginal, and any savings gained from choosing a "cheaper" solution are outweighed by the additional investment that must be made in operational security practices by the team adopting the database.
Nevertheless, licensing and cloud costs for databases are relatively well-standardized across the market, although serverless delivery models (such as FaunaDB Cloud) are disrupting them. Database vendors differentiate their products based on their ability to manage risk -- consistency, availability, durability, and security -- as well as productivity by conforming to existing interfaces that teams already know (such as SQL) or by developing new interfaces that may have a learning curve but are more aligned with modern development practices and more productive, such as NoSQL.
At the end of the day, it's people that are expensive, not databases.
Half of your respondents said they preferred open source data stores. What's the top benefit you think they expect from this technology?
The ongoing shifts in the open source model are very intriguing. The Red Hat model of service and support for databases is effectively dead as adoption moves aggressively to the cloud and to systems designed or modified to benefit from cloud economics. Again, it's about risk: customers don't want a database that needs "support." They want a database that works all the time, in every environment. These self-managing systems (such as FaunaDB) are just coming online now and, especially in the cloud, can deliver not a devops experience but a "no-ops" experience.
Customers like open source databases when they don't trust the vendor or when no off-the-shelf product meets their requirements. That way they have a chance to manage it in-depth and potentially fix problems themselves and avoid lock-in in a theoretical way. Nobody except a database company or a cloud company really wants to be in the database business or the hosting business; it's very expensive!
However, especially in NoSQL, many vendors shipped very poor-quality systems too early and were not able to offer credible support or fixes. Looking for open source is rational under these conditions.
From the vendor side, traditionally open-source companies are adopting "commons clause" licensing models to prevent freeloading by third parties that run cloud services but don't do any of the core product R&D. These clauses are diluting the traditional open-source model and making it closer to the commercial shared source. There is nothing wrong with this, and in practice, it doesn't limit customer flexibility at all.
Even given these benefits, only a third of respondents said their enterprise is using open source data stores. What's getting in their way?
It speaks to my previous answer. Many of them are just not very good, especially when it comes to "true" enterprise-class integration and risk management, and the products that are mature are based on legacy architectures that still have heavy operational and development burden. Many of those designed for the modern multicloud world are too immature. For mission-critical workloads, customers want to work with a trustworthy vendor regardless of the licensing model of the technology.
One way to work around this problem is to use managed cloud services, which make operations the vendor's problem instead of the customer's. But those systems are effectively proprietary because public cloud vendors like AWS extensively modify the open-source products they build on, so the customer is doubly-locked in: to the infrastructure and to the code. The enterprise especially is not very happy with this state of affairs. Public cloud databases are effectively the mainframe model, and AWS is the new Oracle.
Your survey found a distinct set of "challenges, priorities, and use cases" between small and midsize companies and larger enterprises. For example, enterprises say cost is more important than performance but for SMBs it's just the reverse. What are some other differences? What factors due to a company's size do you think are responsible for these differences?
Performance and costs are partial substitutes: higher performance generally means lower costs. In our opinion, this speaks to the maturity level of the market segment. SMBs are less likely to include overhead in their cost calculations (partly because they simply have less of it (such as compliance requirements), so they focus on price/performance, whereas the enterprise is viewing the problem in terms of TCO.
We understand these tradeoffs at Fauna and have worked hard to deliver FaunaDB Cloud in the way that works best for each market segment: serverless cloud for individuals and startups, managed cloud for SMBs, and private cloud/on-premises for the enterprise. There is no feature differentiation to speak of--everybody gets the full benefit of FaunaDB's relational NoSQL model and has confidence that they can change how they operate the software in the future if needed. For example, a small financial startup that aspires to grow will eventually want to bring the database on-premises or to multi-cloud and will not adopt a system that is forever locked into a single region platform-as-a-service. Having the flexibility to change as you grow--that is paramount and very compelling.
Other differences typically come down to the risk vs. overhead tradeoff. SMBs are less interested in things that impact correctness (such as transactions and data recovery), and more interested in productivity. Enterprises are correspondingly willing to invest more upfront to get features that will reduce their risk at scale.
Respondents said tech maturity was more important than migrating data to the cloud when considering their next data store decisions in the next two years. If cloud is getting so much attention these days, why do you think "maturity" outpolls "migration"?
Risk, risk, risk. The future is hardened: modernized databases that are delivered in every cloud and deployable everywhere the cloud is not.
James E. Powell is the editorial director of TDWI, including research reports, the Business Intelligence Journal, and Upside newsletter. You can contact him
via email here.