DevOps in the cloud

There’s no question that DevOps delivers significant benefits to enterprise IT, including improved productivity and efficiency, shorter time-to-market, more reliable releases and higher levels of end-user satisfaction.

Meanwhile, enterprises, attracted by the flexibility and the lack of infrastructure overhead that Cloud solutions offer — advantages that are ever more appealing, given their massive scalability and elasticity — now enjoy heretofore unknown freedom for testing new business ideas and developing them rapidly should they prove viable.

This new, Cloud-enabled zeitgeist of open communication and collaboration between software engineering, quality assurance and operations is not without its challenges, however. 

A recent survey commissioned by Intel Data Center Manager (Intel DCM), “The State of Enterprise Cloud Services,” queried the opinions of over 200 US-based IT managers, directors, software engineers and DevOps professionals responsible for overseeing their enterprise Cloud strategy. 

The survey found that DevOps teams spend most their time monitoring their complex environments, consumed by such as tasks as tuning infrastructure, remediation analysis and Cloud application performance management.

The average time spent using, monitoring or evaluating tool and solutions is over 25 hours per week. Do the math, and that equates to nearly eight months per year that are not dedicated to more productive endeavors. In particular, DevOps pros in larger organizations require more time to tune and support their environment, demonstrating their tools don’t scale as necessary.

The survey also found that Infrastructure-as-a-Service (IaaS) is the predominant focus of enterprise IT environments, with 51 percent of companies that deploy cloud solutions using IaaS as compared to Platform-as-a-Service (PaaS) at 22 percent, or Software-as-a-Service (SaaS) at 20 percent. Meanwhile, 65 percent of system engineers claim they rely on PaaS.

Motivations vary behind a business’ decision to shift from one platform to another. One way to understand this is as a spectrum of migration, with wholly owned on-premise equipment at one end and total SaaS on the other. On this spectrum, IaaS represents a mid-point, providing a degree of control in terms of hardware selection and software environments, and a measure of security and customization as compared to a purely abstracted SaaS. It’s worth noting that any change has costs associated with it and is typically not undertaken without justification. 

Additionally, there may be underlying regulatory constraints that proscribe a definite geographic location for data, for example, stringent German data privacy laws, or industry-specific data security requirements, such as with the healthcare sector and the Health Insurance Portability and Accountability Act (HIPAA) and Health Information Technology for Economic and Clinical Health Act (HITECH).

The issue of the “Noisy Neighbor” is fraught with ironies and inconsistencies across the enterprise IT environment. A noisy neighbor describes a Cloud computing infrastructure co-tenant that monopolizes bandwidth, disk I/O, CPU and other resources, which can negatively affect other users' Cloud performance, including other virtual machines and applications that share the same infrastructure.

The noisy neighbor problem is a well-documented performance issue, but only half the DevOps population are aware of and actively addressing it. Among enterprise IT Directors and administrators, 69 percent of those surveyed said that they are aware of the noisy neighbor problem, and 82 percent of those who are aware claim that they have a strategy to address it. This indicates just how far DevOps is from a mature area of practice.

When technology evolves, most enterprise IT directors and administrators develop strategies to manage their environment. As the survey indicates, 84 percent of VMware users have a container deployment and management strategy, and approximately half of companies with a strategy either emphasize container use (55 percent) or implement on bare metal (50 percent).

So, what can IT managers and DevOps ninjas expect to happen in the near future, and what are the solutions on the horizon if not directly in front to them?

Clearly, the Cloud has been disruptive to the enterprise model and if the Total Cost of Ownership continues to favor the scale of Cloud providers, we can expect more business to move in that direction. Against this backdrop, DevOps requires an environment that is well instrumented at the software level so that a well-running environment ticks over without the need to constantly monitor it. Ideally, DevOps pros would receive proactive notification of issues, from Service Level Agreement (SLA) performance to server health, indicative of whether hardware is properly operational and functional.

As for the problem of noisy neighbors, measuring resource consumption in real-time is necessary, as is the ability to allocate resources to maintain an SLA, such as a latency response threshold, in order to ensure a certain performance level is maintained.

Over time, we can anticipate improvements in overall Power Usage Effectiveness (PUE) for data centers as they improve operational metrics. At the next level of granularity, we would expect to see further improvement for IT equipment by enhanced control of the equipment through access to dynamic and real-time metrics.

On this score, there are many ways in which deployment of a Data Center Infrastructure Management (DCIM) solution can be used to reduce operational and capital expenditures for data center operators through more efficient control of the facility. The ability to automate some of this orchestration represents the next frontier for IT equipment throughput and the next level of optimization at the application level with algorithmic control of workload allocation and load balancing.