Home > DevOps, VMware, VMworld > VMworld Europe Preview: Differences + the DevOps Dilemmas

VMworld Europe Preview: Differences + the DevOps Dilemmas

October 7th, 2015

I previewed VMworld San Francisco in my post: VMworld: Juggling and jousting, VMware’s battles for the future where I initially highlighted the recent people changes at VMware as well as the rough and tumble world VMware lives in while marketing and selling storage and networking. I also spent some time going though the changing face of application development which I’ll expand on in this post.

It’s different

VMworld Europe is the smaller sibling to the exuberant big ‘ol US VMworld that always seems to pave the way. 8000 attendees rather than 23000 certainly makes a difference although calling it more intimate would be taking it a bit far. All the big announcements and strategy crystal ball gazing is mostly done by VMworld US so Europe has historically been left with the hand me downs in terms of announcements. The focus for VMware is to re-use its US content in Europe 2 months later without it feeling stale and introduce just enough new information and provide the networking opportunities to make people spend the money to go.

Culturally VMworld Europe is different from the US version. Europe by its nature is far more multi-cultural with different languages, currencies, food and sensibilities. Although people travel far and wide to VMworld US, locally the European melting pot is more varied along with closer flights from the Middle East, Africa and some of Asia. No-one could ever say Europe is homogenous, in fact even within Spain, the Spanish state, Catalonia, where Barcelona is the capital recently elected a coalition who want to break away from the rest of Spain. This means VMware and its vendor partners have a far bigger marketing job to do to reach people from so many varied geographies and languages. VMworld Europe thus becomes comparatively more of a networking and community event where talking to people in many different ways is the goal rather than getting everyone on the same page. This means so many more meetings for the execs as they have to sweet talk so many more partners, VARs, customers, system integrators etc. from each country.

My preview of storage, networking and cloud for my VMworld US post stands but there’s now more to talk about for the future of VMware which is directly aligned with the future of applications.

Let’s all DevOps

DevOps is the buzzword that in my opinion defines the future of applications and the stuff they all sit on, infrastructure. Virtualisation became a utility which allowed cloud to evolve, cloud now is the utility that allow DevOps to evolve in more ways than you may think.

I’m not going to go into the definitions of DevOps as they are as varied as cloud but perhaps everyone agrees the point of DevOps is far quicker application deployment. This means changes to people and processes to get code shipped quicker which allows businesses to get ahead of their competitors. How this is actually done is varied and complicated. In order to ship code quicker, you need to break up your monolithic code base into smaller pieces or components. Each component can then be amended without affecting the rest of the application and testing is easier as you have less to test with each change. This is the essence of the term “cloud native applications”, applications born into a (private/hybrid/public) cloud development and deployment model. Another important point to remember is developers can generally code “cloud native applications” on their laptops in the same environment as they can deploy at unlimited cloud scale which makes it all that more accessible.

One way to separate components is to use containers, take an OS (currently Linux but Windows too next year) and slice it up further and be able to run multiple partially independent containers which can easily be deployed and managed. Make sure most of those containers are stateless and don’t rely on available infrastructure and you can pretty much run and move them anywhere on any cloud. VMs provide similar portability across physical infrastructure but if your apps are being rewritten to use stateless containers you don’t need vMotion, DRS, HA, SRM etc. to manage the VMs on your infrastructure, you just spread out the containers across your servers/clouds and if any die just respawn them. This is trouble for VMware as its whole existence is all about managing VMs and if you don’t need VMs…

VMware isn’t waiting around, first of all it says you still need VMs as the security boundary around containers isn’t as strong as with VMs. That’s currently correct but I wonder for how long, will more robust isolation be written into kernels further negating the need for VMs as a security boundary. Another point VMware has is to merge VM management and container management. VMware has announced the Photon Platform, a mini hypervisor they’re dubbing a microvisor without all the bells and whistles full VMs need to run with an embedded linux distribution called Photon running in effect on bare metal and able to run containers. The orchestration layer, transparently to the container management layer, very quickly spins up a VM per container for that secure isolation so your Devs and Ops people have a common view mapping VMs to containers. Is this actually required? How much benefit will it have over native containers on physical hardware? Do your Dev and Ops people need to see the same thing or is this just to smooth the transition using existing enterprise toolkits?

Containers still need networking and storage so VMware is pushing VSAN and NSX as the solution to this but again, not necessarily required, just use local disk for stateless containers and shared/cloud storage for containers that keep persistent data. If Photon Platform is only for containers and can’t run normal VMs (my understanding) you then need to split your VMware environment anyway and although have a common view can’t use existing hardware and host native ESXi containers and VMs. Is the Photon OS the stopgap? This is a container specific Linux distro by VMware which can run as a VM but now you’re paying hypervisor prices to run VMs when you don’t need/want the advanced features.

Notable is VMworld + DevOps seems to be all about getting developers used to coding against vCloud Air, certainly important but a hard sell when Amazon offers so many more cloud native applications built in.

The Microsoft Effect

The other elephant in the room is Microsoft. VMware’s talk of cloud native applications is all about Linux which to be fair is what most cloud native applications are running on however Microsoft is throwing everything it has at cloud native applications by bringing Windows tech to the table in a big way. .Net is now open source, there are millions of app developers schooled in .Net application development. Sure, this may mean .Net ported to Linux but the bigger picture is developers choosing to use .Net to create any number of future platforms, think Internet of Things, where previously this would be unlikely.

Microsoft Azure is a truly impressive cloud platform, built from the ground up as cloud native, reinventing Microsoft and deliberately cannibalising their future Windows Server on-premises business. Microsoft’s vision for private cloud is all about bringing cloud native Azure into your data center rather than pushing traditional IT out from your data center into the cloud, an important difference from VMware vCloud Air. Managing the connection between the two is exactly where VMware is targeting with their hybrid cloud messaging of deployment to the middle between the two ends but at the moment this isn’t as simple, standardised and consistent as it should be.

Microsoft will not only have containers in Windows Server next year (it says its been running containers for Azure for years) but has been partnering with Docker, Mesos and Kubernetes to allow container management from traditional Linux only managers. A single container management system that looks after Linux as well as Microsoft containers is another unprecedented leap forward in interoperability unthinkable of the Microsoft of just a few years ago.

Identity

Still talking DevOps, one of the problems developers generally don’t realise they have is about identity, authentication and security. All these cloud native applications made up of many containers need to talk to each other and need to do so securely. Identity becomes a far bigger challenge than just logging into a server remotely to run a SQL query. One of Microsoft’s clever moves was to punt Office 365 first to customers. Customers were encouraged to federate or move their Active Directory into Azure so they could access Office 365. Email and PowerPoint are not the most business critical applications for many (of course, for some they are) but Microsoft allowed companies to get comfortable with their security in the cloud in a non threatening way. Once companies trusted the Microsoft security blanket then any other Azure service they tied into this authentication mechanism was so much easier to do.

Back to containers though, if each and every container needs to talk to a number of other containers they need to ensure they are talking to a trusted container, remember these remote containers could even be running in another cloud, so who provides this? This is where VMware’s Project Lightwave, another Open Source effort, comes to erm.. light. VMware wants to take Active Directory head on to be the cloud identity/authentication/security platform of choice, a tall order but someone has to do it. Many companies have tried to take on AD and no-one has come near unseating it as the preferred enterprise identity authority but things may be up for grabs in the clouds with containers needing something different from GPOs and user and computer logons. However, Novell NDS was a far far better directory service than Windows 2000 Active Directory and we know how that turned out!

After all that…

So, a lot to cover and learn at VMworld because there’s a hell of a lot going on in enterprise IT at the moment and DevOps is/will be the the major current/future driving force. Enterprises are spending a huge amount of time figuring out cloud but I believe DevOps is the catalyst to help create the killer apps that prove why they actually want to do cloud in the first place rather than just outsourcing infrastructure.

I’d love to catch up at VMworld if are you attending, please get in touch @julian_wood or bump in to me somewhere and we can continue the conversation.

Comments are closed.