Remember when standing up a new database instance involved modeling the application using the underlying database to figure out what hardware (server, storage, etc.) was needed to support the new instance? And, since most organizations did capacity planning on a spreadsheet using linear metrics, there was a lot of estimating involved, thus over-provisioning the hardware was standard procedure.
Virtualization made it easier to make use of resource allocation with virtualized machines, storage, and even networks. Driving servers to 90%-plus utilization made IT hardware dollars go farther but complicated the process for performance tuning, since the virtual CPU(s) and storage associated with a virtual machine running a database instance were fighting for resources with other virtual machines.
Then came infrastructure as a service, simply called IaaS, for which you didn’t have to worry about purchasing hardware but paid for compute resources. Amazon Web Services (AWS) is the 800-pound gorilla in this market with $25.7 billion in revenue for 2018. However, Microsoft Azure is experiencing double-digit growth in enterprise adoption and has been adopted by 45% of firms compared to 64% of firms that have adopted AWS, according to a 2018 RightScale report.
Database as a service—commonly referred to as DBaaS—is the newest, most competitive, and, honestly, most alluring service to use. DBaaS is also lumped into the platform as a service, or PaaS, category by some vendors. And, with Microsoft introducing Azure SQL Managed Instance (ASMI), a new deployment option of Azure SQL Database, the competitive landscape has gotten hotter. With ASMI and similar offerings from Oracle, IBM, and others, you only pay for the database compute resources you need, drastically reducing (but not eliminating) the overhead associated with managing the database.
Let’s look over some of the more common pros and cons for each implementation option for DBMSs.
Physical Server (on-premise)
Pros include unfettered resources, the ability to physically secure hardware, high-speed iSCSI (Internet Small Computer Systems Interface) or Fibre Attached Storage, and complete control of all associated resources.
Cons include CAPEX requirements; the need for physical space, cooling, and security; higher operation costs associated with IT management; and the constant need to upgrade and replace hardware.
Virtualized Server (on-premise)
Pros include higher utilization of hardware resources, easier provisioning and on-demand addition of resources, easier failover of mission-critical virtual machines with clustering, and software-defined storage and networking.
Cons include the increasing costs of the tools related to hypervisor management, the need for expensive, specialized IT resources, the fact that there are the same expenses associated with physical servers, and, according to some, slower performance compared to physical.
Infrastructure as a Service
Pros include access to the latest in compute resources, on-demand resources for short-term or seasonal workloads, optional high availability and disaster recovery, the ability to pay for what you use, OPEX spending versus CAPEX, and access to vendor-specific migration tools.
Cons include the fact that, depending on the level of service, costs can exceed on-premise deployments; on-demand resources equate to expensive on-demand pricing; IT personnel costs don’t go away because of the need to configure security and network access; and the fact that once you lock in to a specific cloud platform, it might not be easy to migrate off.
Database as a Service
Pros include guaranteed service levels, no need for IT resources to support infrastructure, the fact that all updates and upgrades are handled by the vendor, high availability and automated backups, and the ability for DBAs to focus less on infrastructure and more on applications and performance.
Cons include possible cost overruns based on usage and the need for specialized and expensive DBA resources.
Continuing Cloud Dominance
Organizations are challenged with justifying continued support of cost-intensive data centers and the associated impact on a company’s CAPEX. Even though a lot of rogue applications and databases spun up in the cloud are now being brought back into the IT fold, the seduction of OPEX costs versus CAPEX costs often determines where the applications will run.
That said, it looks as if cloud has won over all other options, according to a June 2019 Gartner study, which stated that database cloud services already represented $10.4 billion of the overall $46.1 billion DBMS spending in 2018. Gartner predicts that by 2022, 75% of all databases will be deployed on, or migrated to, a cloud platform. And, if that is correct, the DBMS market will be forever changed and we’ll nostalgically look back at on-premise database implementations.