Skytap CTO Brad Schick has been invited to speak at the 2015 Seattle Technology Leadership Summit and the OpenStack Summit Vancouver on the following day! We sat down with Brad to learn more about hybrid apps, enterprise cloud adoption, Bimodal IT, and some of the exciting benefits that OpenStack and Skytap can provide our growing communities.
Noel: The abstract for your upcoming session at the OpenStack Summit 2015 states, “the boundary between one multi-tier enterprise application or service and other is often unclear.” Can you go into why this is so, and maybe where you’ve seen this?
Brad: This issue is pervasive in big companies. The reason is pretty simple, particularly for on premise applications: There is no broadly adopted approach to encapsulate an application and all of its dependences. In the worst cases, enterprises have data centers with a bunch of physical resources, and atop those resources you have virtualization and operating systems. All of that is often just a flat pool of “stuff” without much in the way of namespaces or isolation.
Within that flat pool of stuff, businesses deploy dozens or hundreds of applications and services. Some applications depend on certain underlying physical resources. Some applications communicate with other applications. A few applications may depend on the side effects from other applications. Those are just some of the dependencies that build up, and very few of them are codified in a machine-readable manner. For environments with long histories, even the humans forget how it was all strung together.
Often the only way out of this problem is to just start experimenting. Try to deploy a specific application elsewhere to start discovering dependencies. If possible, try to define the boundaries and codify the dependences in a manner that can be reused for the next effort.
Noel: You also point out that “it’s not uncommon for those who manage these applications to be unfamiliar with cloud concepts.” Are we at a point where it’s really time for managers to have a better understanding of the technologies that are having such a positive impact for dev/test teams and enterprises as a whole?
Brad: First, there is a lack of awareness of the possibilities in the cloud due to “newness”. Most businesses are aware of what the large cloud providers offer in general, but often not the specifics. Smaller companies like Skytap that are helping enterprises with more traditional applications are even less well known.
Second, there is not always in-depth experience with cloud services. Some businesses are further ahead, but many are not yet running their most complex applications in the cloud. Cloud platforms often have different limitations than traditional infrastructure, not to mention new performance, availability, and durability characteristics. They aren’t always worse in the cloud, but they’re often different.
Interestingly, since Skytap was designed to work with more traditional applications, rather than new, “born in the cloud” apps, we are finding ourselves in the position of being many customers’ first significant experience with the cloud.
Noel: What makes OpenStack public and private clouds an attractive option for those looking to move pieces of their app to the cloud?
Brad: OpenStack is shaping up to be a viable alternative to large single vendor public and private clouds. Its openness reduces vendor lock-in and gives businesses more control over their destiny. While it is certainly true that there are going to be lots of flavors of OpenStack, the differences between them will be substantially smaller than the differences between one proprietary cloud vendor and another. Today, OpenStack is behind the large public cloud providers in terms of functionality, but I think that gap will decrease over time.
For more technically sophisticated businesses, OpenStack also provides the option to do it yourself. You can take full control of (and fix) your own deployment while still collaborating with and benefiting from other individuals, businesses, and commercial vendors through open source—the same reason many open source projects succeed.
You also have another benefit that isn’t about openness at all. I believe that we’re just at the start of the workloads that OpenStack will tackle. Some vendors are building solutions that don’t even expose OpenStack directly to their users, but instead wrap it with higher-level offerings. This will bring a greater number of interesting products to market faster and with lower cost. This is analogous to the way many networking appliances are running BSD or Linux inside without really exposing those details to the end user.
I think the cloud market will see benefits from both direct and indirect OpenStack adoption.
Noel: Part of your role here at Skytap leading our own OpenStack strategy; when did you and others start this initiative, and what prompted that?
Brad: First, a bit of background on Skytap… We offer a service that helps enterprises speed up the SDLC for complex enterprise applications. In particular, we help dev/test teams become more efficient.
While we do not consider ourselves to be an infrastructure or public cloud provider, we have always needed infrastructure under the covers to run those complex applications. Out of necessity, we built our own technology to provision and clone fully isolated virtual data centers on demand. Much of our technology was developed before OpenStack reached a reasonable level of maturity.
We have been keeping an eye on OpenStack since its inception. OpenStack’s growing maturity and acceptance as a standard API is really what triggered us to start “mingling”. That is what I like to call our OpenStack direction.
OpenStack is focusing primarily on infrastructure and lower level platform services while Skytap is focusing primarily on the higher level features that dev/test teams need along with enterprise specific infrastructure features. Over time, we are mingling the best of Skytap with the best of OpenStack. That includes adopting “core” OpenStack APIs and services, and contributing back some of Skytap’s unique infrastructure technology.
Ultimately Skytap’s customers will benefit from more open services and APIs and faster time to market for new features, and hopefully, OpenStack will benefit from what we can contribute back.
More information about the Seattle Technology Leadership Summit and the OpenStack Summit Vancouver can be found here!
Noel: You’ve spoken about the concept of hybrid apps before, could you define what those look like, and if the number of hybrid apps has increased since you first started taking a look at them?
Brad: A hybrid application spans multiple infrastructure and cloud locations, making use of resources and services from each. In particular, components of the application may run on-premises and in the cloud. I’ve used the term to differentiate from “hybrid clouds” which usually means infrastructure that spans on-prem and public clouds.
My belief is that a hybrid clouds don’t solve significant problems on their own beyond basics like backup. Most of the effort, and the benefits, from moving workloads to the cloud comes from updating applications to take advantage of new components and services available in the cloud. That could be anything from your build and CI service running in the cloud to core application logic moved to the cloud.
My experience is a bit biased since I work at a cloud service company, but yes, I think the rate of uplifting tools and applications to leverage the cloud is increasing significantly. A few years ago there was much more hesitation than we see today.
Noel: What are some of the highlights of the evolution of complex enterprise apps that are most exciting to you?
Brad: There are services and efficiencies that cloud providers can offer that are very hard to do efficiently on your own. Good examples are workloads that have highly variable resource needs.
Let’s say you want to perform computationally intensive data analysis a few times a week. Owning a fixed pool of resources for that periodic need is not efficient. Cloud providers, however, essentially “timeshare” capital resources across many customers whose needs don’t all coincide. That translates to higher average utilization of a given resource and lower hourly costs for each individual workload.
Other common examples are services that are not easy for enterprises to create themselves. Why build, deploy, and manage your own service if someone specializing in that area is likely to do it better? Even if the cost is neutral, businesses will often get better results by consuming a ready-made cloud service instead of deploying and managing their own. Skytap is a good example of this.
Many of our customers could build their own on-prem Environments as a Service capabilities to support dev/test teams, but it would take years to create something as feature-rich, scalable, and hardened as Skytap.
Noel: You’ve mentioned that the complexity of these apps and the boundaries between them often make them difficult to move entirely to a public cloud. Have you found that people think a “move to the cloud” must mean moving it all, and not just the components that make the most sense for a company’s specific needs?
Brad: People have started talking about “Bimodal IT”. The majority of public cloud adoption is still driven by new applications that are born in the cloud (mode 2). Then there are more traditional applications that have been around for years or decades and often provide the technology foundation for businesses. These applications were designed and built to run in standard data centers (mode 1).
I think that is a decent assessment of how many businesses are thinking about cloud adoption. Because there are different constraints, there are different approaches to adopting the cloud.
While I have certainly read case studies describing wholesale migration of complex enterprise apps (mode 1) to the cloud, I have not personally seen it done. The challenges and opportunities that I’ll cover in my session derive from *gradually* moving these traditional enterprise apps into hosted cloud services. For most complex apps, I think that is the only viable option and many people are starting to agree.
