OpenStack is gaining increasing industry attention and, while it can deliver huge advantages, some may say it is “hyped”. Although OpenStack has an ever growing range of enthusiastic practitioners and advocates, as you may be aware, OpenStack is not without its critics – including Gartner – who outline the challenges of OpenStack adoption. It’s therefore generally recommended that OpenStack adopters consider engaging professional services experts to help them avoid the pitfalls
With the November 2014 OpenStack Summit in Paris opening as I write this – you can find us at stand C3 along with our newest acquisition, Metacloud (Stand E37) if you are going – my thoughts turn to the issues and challenges facing our customers when they deploy OpenStack into production projects. And who better to ask than our Cisco Services consultants who are delivering OpenStack adoption services (which we launched this time last year at the Summit in Hong Kong).
These consultants are at the “coal face” (as we say in my part of the world, Scotland) of OpenStack– they are the experts digging deep in the IT equivalent of the mines working with real customers going live with real-world OpenStack. More than R&D investigations, these deployments are happening with customers who are betting their business dollars, pounds, yen and other currencies on OpenStack. However as the video (below) shows, OpenStack has its deployment complexities. Hence increasing numbers of our customers are engaging Cisco Services to help them on OpenStack.
To share our practical experiences with you, we sat down and came up with our “top 5” adoption challenges list which you may find useful if you are considering or embarking upon an OpenStack deployment:
- Cross-domain technical expertise is mandatory
- Going deep with Open Source
- “It’s just middleware”
- Don’t underestimate the learning curve
- Expect your OPEX to increase (sorry!)
Let’s explore these in more detail.
Note that our consultants are not “sales” staff. They are not telling you what an OpenStack deployment will or “should” be like. These are the guys and girls who get things working on site with you. So this “top 5” comes from their school of “hard knocks”. Their guidance is based on what it is and was like for them in real-life deployments, what surprises customers, what they don’t quite expect and so on– and in real-world deployments, not Proof of Concept tests (PoCs) nor pilots. I guess this is my way of telling you others may share a different “top 5” list with you. 🙂
Let me, then, explain these adoption challenges, one by one:
1. Cross-domain technical expertise is mandatory
OpenStack is a cloud middleware solution spanning compute, storage, networking, security and other domains. OpenStack installers in particular need to understand each area. The days of silos are going fast. This is also true to an extent for OpenStack developers. For example, if your software engineers are not networking savvy, it may be useful to start cross-training. Linux expertise is also critical. These considerations can be a surprise to those new to OpenStack.
2. Going deep with Open Source
Open Source is touted by some as the nirvana of cloud software. I’m not debating the advantages, neither am I in the “nirvana” camp. However to gain these advantages, you need to “go deep” into the (open source) software. You need to understand the various OpenStack components, Neutron, Nova, Cinder, and Swift. Unlike commercial-off-the-shelf (COTS) packages, we’ve found that OpenStack components can be inconsistent in places, with different “look and feel” from both behavioral and developer perspectives. And expect in cases to troubleshoot even down into the Open Source code base. And as I’m pretty sure you’ll already know, troubleshooting code written by someone else is never that easy (something I remember painfully well from my software developer days!)
3. “It’s just middleware”
I’ve worked with various “framework” or “abstraction layer” type middleware products over the years, and advocates will tell you they can do everything. Yes, with further development and integration, they can do lots, but out of the box, they just sit there and don’t do much! Precisely because OpenStack is “just” middleware, you have to invest in “connecting” or integrating your applications to OpenStack. The advantage once you’ve done this, of course, is that your apps are isolated from the underlying infrastructure, and you can leverage the proven, re-usable cloud operating environment components in OpenStack. That said, we’ve found customers surprised that OpenStack is not a fully complete cloud operating system, and such experience lends to the feelings of “hype” around OpenStack.
4. Don’t underestimate the learning curve
Yes, there is a steep learning curve to getting started with OpenStack (which is why we advise you to engaged Cisco Services to help accelerate your progress 🙂 ). Challenges including the fact that different components were built by different teams, the range of technologies covered, the rapid progress on the roadmap (with regular new releases, at approximately 6-month intervals), it’s an ever changing world. While customers are aware of some or all of these challenges, we find their reaction to the learning curve is still often one of surprise.
5. Expect your OPEX to increase (sorry!)
Many customers think: “Great! Open Source software will reduce our software licensing expenditure and reducing -your capital budget!” However, “every silver lining has its cloud”, as they say. For example, you may find yourself troubleshooting complex OpenStack subsystems, and while access to the code may help you, don’t forget that every hour you spend troubleshooting the Open Source, it’s one hour less you are investing in your value-add. Your boss may not appreciate that you’ve developed into an OpenStack guru if the business application you are supposed to deliver is late. Therefore don’t forget you will probably need a support contract with one of the OpenStack distributors (e.g., consider our good friends and partners at Red Hat). Most definitely this is money well-spent, because you can leverage the investment to deliver more applications at an accelerated pace and meet your project deadlines. That said, the net net is more than likely that your OPEX will increase since you’ll almost certainly need to upgrade the skill-sets in your operations teams.
Thanks and next steps…
The “top 5” list above came from some brainstorming with my esteemed OpenStack colleagues, the people at the forefront of helping Cisco’s customers adopt OpenStack through our professional services – so I’d like to thank (in alphabetical order!) Renato Fichmann, John Garrett, and Derek Hartung for their sage-like insights. You may get to meet these guys if you engage us to help you on OpenStack.
Of course you may have faced your own specific challenges – so the above list could easily be extended. Feel free to leave your comments and debate with us around other challenges. And if you like this topic, you may encourage us to write a follow-up on the next 5 challenges we come across most often – let us know via the comments field below.
PS: The OpenStack Paris Summit started today, November 3. In addition to the general speaking sessions, booth, demos and other events captured here, Cisco’s team members are also leading and participating in several design sessions to shape the Kilo OpenStack release. We look forward to having some great discussions at the Summit and hope to see you there as well! Please stop by Cisco (Booth C3) and our newest acquisition, Metacloud (Booth E37). Visit us at Cisco’s OpenStack webpage for more information.