Resilient IT Through DevOps
If you haven’t yet noticed that prioritization of non-functional requirements (NFRs) is changing amongst your user base, you will soon. For decades, we have held to the same familiar set of NFRs. Every team had its own definition and particular spin on NFRs, but the usual suspects are accessibility, availability, extensibility, interoperability, maintainability, performance, reliability, scalability, security, and usability.
But new priorities have surfaced, as IT has experienced a sea change over the past few years. Some organizations have even adopted completely new NFRs. The rise of DevOps has coincided with these changes, and the movement’s principles enable IT teams to more readily adapt to rapidly changing requirements.
Your grandfather’s mainframe was very reliable
Historically, IT system designs were praised for reliability. Robust and stable systems could “take a licking and keep on ticking.” As computing became more pervasive, scalability became the watchword. Systems should be able to grow and expand to meet increasing demands.
Scalability as an NFR priority represents just a slight shift from reliability as an NFR. Both operated off the mindset that the original system design was valid. Reliability ensures that the system continues to provide the stated functionality over time, and scalability ensures that you can do so for an increasing demand set.
Roughly 10 years ago, things began to shift as more and more organizations embraced movements like agile or XP, and architectural models like Service Oriented Architecture (SOA). These initiatives promoted adaptation and response to change as desirable system qualities. Next, cloud computing introduced us to the notion of elasticity, further promoting the values of flexibility and responsiveness to change.
A resilient system is a happy system
The state of the art for system design is always evolving, and we see noticeable leaps forward every few years. The current phase of evolution is toward resilient systems.
Legacy system designs relied upon expensive infrastructure with multiple-redundant-hot-swappable-live-backup-standby-continuity-generators (or whatever vendors are peddling lately). In contrast, resilient system designs embrace failure and promote the use of cheap, commodity hardware, coupled with distributed data management, parallel processing, eventual consistency, and self-healing operational nodes.
Some portion of your system is likely to go down at some point, and resilient systems are designed with that expectation. Resilient systems and resilient processes are able to continue operation (albeit at diminished capacity) in the face of failure.
The prioritization of resilience over reliability as an NFR can be seen within the DevOps movement, the development of the Netflix Simian Army, and the rise of NoSQL data management solutions.
DevOps and resiliency
DevOps is a multi-headed beast, more a movement guided by a set of principles than a tangible and well-defined construct. While organizations are free to adopt aspects of DevOps that suit their needs, one common thread is that of resilience. Failure is seen as an opportunity to improve processes and communication, rather than as a threat.
The principles of continuous integration and continuous delivery that are core to most DevOps practices exemplify a resilient mindset. Where the classic waterfall model relies upon detailed front-end design and planning with an all-or-nothing development phase and late-stage testing, DevOps teams are more agile, embracing a “fail early, fail often” model. This approach results in more resilient and adaptable applications.
Netflix Simian Army
Netflix gained world renown when the company broadcast details of its Simian Army work in 2010 and 2011. Through the automated efforts of Chaos Monkey, Chaos Gorilla, and a slew of other similar utilities, failure is simulated in order to develop more resilient processes, tools, and capabilities.
John Ciancutti of Netflix writes, “If we aren't constantly testing our ability to succeed despite failure, then it isn't likely to work when it matters most — in the event of an unexpected outage.”
A third illustration of the growing fascination with resilient, self-healing systems is the transformation now going on in the data realm. Data and metadata management have evolved considerably from the relational databases of yore. Modern data management strategies tend to be distributed, fault-tolerant, and in some cases even self-heal by spawning new nodes as needed. Examples include Google FS / Bigtable, in-memory datastores like Hazelcast or SAP’s HANA, and distributed data management solutions like Apache Cassandra.
Miko Matsumura of Hazelcast notes, “Virtualization and scale-out power new ways of thinking about system stability, including a shift away from ‘reliability,’ where giant expensive systems never fail (until they do, catastrophically), and towards ‘resiliency,’ where thousands of inexpensive systems constantly fail—but in ways that don’t materially impact running applications.”
Keeping pace with the cool kids
It’s often said that the only constant is change. The DevOps movement positions organizations to embrace change, rather than fear it. Continuous integration, continuous delivery, and continuous feedback loops between dev teams and ops teams facilitate an enhanced degree of agility and responsiveness.
As business and society evolve, our system design priorities must adapt in parallel. The cool kids will change the game again at some point, but for right now, “change” means designing systems and supporting processes that are responsive and adaptable by prioritizing resilience over reliability.