<img height="1" width="1" style="display:none;" alt="" src="https://dc.ads.linkedin.com/collect/?pid=204641&amp;fmt=gif">

The role of the DBA in light of DevOps and Cloud Migration

Posted by Andreas Hope on Jul 27, 2018 8:53:00 AM
Andreas Hope
Find me on:

It's an old problem. The more results you deliver, the more that are expected. The faster you provide them, the faster they are expected. Next thing you know, old methods don't work as well as they used to.

New demands require new workflows. On top of it new technologies are appearing making it seem like your old tools are no longer needed. Next thing you know the environment in which you work has changed. It is barely recognisable, and you might be afraid that you are no longer relevant.

The above sentences could apply to just about any occupation. It is no less the case for Database Administrators. It's happening both in the areas of workflow and business culture, particularly with the growth of DevOps, and in the technological area of database servers migrating to the cloud.

The fears are real, and not unfounded, however perhaps they are a bit exaggerated. The need for DBAs in either environment is certainly not going away. Let's address these separately.


DevOps and Changing Workflows

As the software development cycle has by necessity sped up, the nature of how roles are defined are changing. Traditionally (if we can say traditionally about a field that's actually not that old; SQL itself was created less than a half a century ago), roles were pretty well defined. Developers wrote code. Sysadmins managed the servers. DBAs build and managed the databases, and handled all deployment from development, to staging, to testing, to production. QA s noticed the mistakes that everyone and made nobody happy (except the customers, by their reduction in complaints... oh who are we kidding? customers always complain). The workflow was pretty clear, however change was slow. At first there wasn't as much of a problem. Customers didn't have a lot of choice, and competition was relatively limited, so it was okay if any new development took months to come to


However, for better or worse that has changed. Changes in business models forced the need for a faster model for bringing products, data, and fixes to the customers. Next thing you know, it has become necessary to bring everyone on board throughout the whole process, and to make it far more iterative. If you're not open to change, it can be very disconcerting (especially with QA jumping into the fray throughout the process). Even worse, Developers are committing changes to the database schema. And it seems to be working (at least in some cases). What's a DBA to do?


Well, the first thing is to remember: “Don't Panic.” As a DBA, you know whether the database is running properly. Remember that nobody else knows the inner workings of how databases work, and how to manage performance. It's important to remember that DBAs are integral members of the “Ops” part of DevOps. DBAs are crucial to this role quite simply because they know how differences in one system can affect another. On top of this, some of the changes are requiring DBAs in different ways than they did in the past.


DBAs must change a little to fit the new machine. While it was possible in the past to get away with taking things slowly and making sure everything was running at its best before releasing a product, customer demands have increased. This is no longer a real option. As a response, it is crucial for DBAs to learn more about the Agile process wherein problems are broken into pieces and problems are addressed in the real world (I know this is a simplification, but this isn't an article about Agile).


Of course, it is true that some shops still take too long to deploy code. The DevOps folks are completely right about this. However if we are going to use these shorter cycles, the risk of bad code getting released is greater than it every has been before. Bad code can take down a database faster than you can say “ Bobby Tables .”

Of course you can't stop the pace of development, which means that the need for DBAs to be monitoring the databases (with good analytic tools) is greater now than ever before, quite simply because of these shorter cycles. Sure, the code worked in testing, but nobody can realistically tell how it will work with the volume that is seen in the real world.


For any smart business to provide a quality product, it's absolutely critical to be monitoring status and performance. With good tools that can effectively monitor and analyse the behaviour of this code in real life, DBA can keep track of any problems that need to be fixed, or database performance that needs to be tweaked.

Cloud Migration and Changing Technologies

Another major change affecting DBAs is the actual location of the database servers. In the past, our databases were all run on locally hosted servers. While this still is the case in many places (there will always be that dark freezing room in the basement where new forms of artificial intelligence are creating themselves and preparing for the singularity), more and more we are starting to host in the cloud.


Despite these advantages of cloud hosting, new threats emerge, hidden in the silky words of the cloud host's marketing language. There's a claim that their systems are now “fully automated.” All you need to do, they claim, is choose a few configurations, push a button and you're all set. This language getting into corporate managers' ears is fodder for any DBA's nightmare.


As we know (and anyone with any experience with the cloud will know), these promises are pretty much fantasies. Sure, there are many advantages to cloud hosting. In many ways it does run smoother; there's better replication and uniformity of access in the cloud. However, these systems still need to be monitored, maybe not entirely in the same way as your local servers, but also for other reasons.


The supposedly “self-managed” instances you are running in the cloud? Those are typically only subsets of what you are used to running locally. They may require less manual work quite simply because they have less functionality than your local systems. They are often virtual and shared. The hardware they are using? Opaque to you, but likely highly variable from one location to another. Instances may behave entirely different from one location than on another. Typically, the functionality is severely limited compared to your own servers which you can tinker with to your hearts delight.


On top of this, applications rushed into production as a result of the DevOps process will not suddenly perform better in the cloud. Sure you can add more resources to solve the problem, however it may not be the most efficient way, and for each bit of extra resources, such as new instances, more replication, this increases the expense.


This brings us to our next point: cost. Typically, most cloud hosts charge by usage. If you leave the management of these Databases to them, they have really no incentive to make these run well, and to identify if there are processes that are causing more processor time and/or bandwidth. In fact, it is in their interest not to fix these. So DBAs will need to monitor activity, keep good records, identify if there's a bad process, and be able to fix these. You need to be able to fine tune the performance, identify whether adding resources makes sense, or if in some cases clean up unused resources.


Overall, despite changes in workflow and technology, DBAs remain important for new reasons. With the right tools, their value relevance has actually increased as business needs increase.



Start your free trial today! download here