Tuesday, July 22, 2014

Using DBaaS to Align Database Management with Your Application Delivery Processes [feedly]

Using DBaaS to Align Database Management with Your Application Delivery Processes
// Build AWS-compatible Private Clouds with CumuLogic

Today's application development teams are moving fast, but is your team being held back by slow database management practices and tools? For many organizations, this is a significant problem that deserves as much attention as getting application code changes flowing smoothly into production. In fact, they are tightly interdependent challenges that need to be solved together.

Perception vs. Reality

Your Database Environment: Perception vs. Reality

Think about the application development process for a moment. How long does it take you to stand up a new copy of your database? Are you sure you have configured everything correctly and consistently? Do your developers and testers have access to production data for realistic testing? For most organizations, it's unlikely that the answer to these questions are "minutes", "very certain", and "yes".

It's also critical to embrace the distributed and multi-instance nature of your organization's data landscape. Gone are the days of a single, large and controlled database that sits underneath all of your applications. Even in a single application team, it's a fairly universal truth that you will have multiple environments. From individual developer sandboxes, to integration and test infrastructure, and through to production clusters (and even DR copies), we aren't really working with a single monolithic database environment. Each one of these environments means a new deployment of your database configuration.

Mapping Database Management to Application Delivery Processes

Data is different from code. Data represents the persistent state of your application, yet dynamically changes and grows as the application is used. Code, on the other hand, is much more of a static entity that can be versioned cleanly and deployed through configuration management tools. Many advanced approaches and tools have been developed to manage configuration drift, and ensuring that the code that is expected to be deployed is, in fact, the code deployed is now easier than ever. It's the dynamic nature of data that makes managing it different from code deployments.

So what can you do about this difference? First, understand that there are aspects of your database environment that can be managed in a similar way as code and configuration. They fall into two buckets: the definition (configuration) of the database system itself and the definition of the schema / database code. Next, think about how the data being generated by your production application can be re-purposed efficiently for testing both new versions of your applications and the processes you will use to deploy your applications.

One great example of these concepts rolled into a consistent line of thinking is the work of Martin Fowler and Pramod Sadalage on Evolutionary Database Design. Evolutionary Database Design is a methodology for ensuring that your database environment is able to adapt and evolve along with your application. It advocates several specific approaches to ensuring that the database layer is an integral part of the entire application delivery pipeline.

While the Evolutionary Database Design pattern is certainly only one of many approaches to take, it's well thought through and worth reviewing to find some key themes:

  1. Database instances need to be (re)generated on-demand – Development, integration, test, staging, and production environments should be consistent.
  2. Database schemas should be defined as "refactorings" – Your schema should be created by series of evolutionary migrations from one version to the next.
  3. Actual data matters – Use test data for earlier on in the application's lifetime, but be prepared to use copies of production data later.

How does DBaaS Fit?

A Database-as-a-Service (DBaaS) platform can play a large role in how you advance your application delivery processes. Specifically, DBaaS can address two of the three themes from above: instance creation and data portability.

Instance creation is perhaps the easiest to comprehend, because that's what most DBaaS solutions are solving at their core. Think about the definition of your database environment as configuration metadata (or a blueprint), which when used to deploy a new database instance from a DBaaS platform should result in a consistent database engine deployment for each instantiation. It shouldn't matter if the environment's specification is as simple as a single node or as complex as a multi tier cluster configured for sharding, it should be just as easy to get a new environment online.

The second role that a DBaaS platform can play is in helping to move the data itself around in the application lifecycle process. It should support operations like replica creation and database cloning, again via a self service UI and with APIs. These two functions are often though about as "operational" considerations for a database environment, but if you think about the whole application development system, they can become surprisingly useful to the entire team.

First, let's define them:

Moving Data and DB Config through the Application Lifecycle

Moving Data and DB Config through the Application Lifecycle

  • Cloning is when a snapshot of the database system is created, and then transferred into a new "instance". It's a point in time copy created on-demand.
  • Replication is the synchronous or asynchronous replay of database changes against a "copy" of the database, which can then be detached from the "master" database to operate as a standalone system.

Both of these features offer a similar end result: you get a copy of the database that you can use for other purposes. Cloning takes a bit longer, but can easily happen off hours to minimize impact on the production environment. Replication can ensure a much faster "turn up" of the new environment with up-to-date data, but needs to be created ahead of the request for an instance. Regardless of the differences, they both provide a foundation with which each developer, tester and release manager can easily get a version of the production database when it's needed.

Think about that in the context of bug-fixing: All too frequently, developers are handed user reports of an issue that is hard to reproduce without having a copy of the production data in place to replicate the situation. If it was easy for a developer to instantiate their own copy of the production environment (on demand) from a clone created each night, all of a sudden that developer will be able to resolve the bug much more effectively. Not only that, but the testing process will be able to ensure that the proposed patch does, in fact, solve the problem with real-world data as input.


You can move the code easily, but data and databases are a much harder a challenge without the right tools. A well designed Database-as-a-Service (DBaaS) environment can play a key role in helping you be more nimble in your treatment of data services within your application development pipeline. A DBaaS should offers solutions to both the "course grained" configuration management / automation challenges of the database engines themselves and the need to have self service access to copies of the dynamic data that your applications generate.

The right DBaaS solution should give you:

  • Self service access new DB instances for your development and testing teams
  • Move database engine configurations through the application development pipeline
  • Easily pull data back into staging, test and dev environments to give your teams access to real-world data
  • Support production rollout of application releases through replication and / or database cloning
  • APIs to make these operations an integral part of your continual integration and deployment processes

If you would like to learn more about how the CumuLogic Database-as-a-Service Platform can help with your organization's database management challenges, get in touch or download the software!


Shared via my feedly reader

Sent from my iPhone

No comments:

Post a Comment