I’ve waited a week after VMworld 2015 ended in San Francisco before writing this post: there were too many blog posts from every blogger trying to cover any new announcement like a newspaper, and with the addition of news collected directly from the Expo Floor, sessions I’ve attended and other sources, the amount of info to digest required a bit of time. This post is not a recap of the event, just the things that I’ve seen and I found interesting. If something is not here, it may be that I’ve missed it rather than don’t liking it.
VMware VAIO libraries
VAIO stands for VMware API for IO Filtering. From an architecture perspective, this is going to be the new and VMware approved way to deal with IO happening into ESXi.
Let me give you a 30 seconds pitch on how it works: ESXi exposes the VAIO framework that runs in kernel. In User World (as a Linux user mys I would call it userland…) a 3rd party vendor can deploy a IO filter. Thanks to IO policies (I can imagine a policy like “send any write IO of this VM to this IO filter” for replication purposes) the VAIO framework sends IO streams to the filter. Then, depending on the use case, the IO filter may or not return a modified IO to the framework if it’s a write caching system for example. In this way, there’s no need for any 3rd party solution to go deep into the kernel and risking to create issues, and anything can be done using now this “official” IO stream.
Since VAIO was first announced, these libraries were at the top of my list of “cool things coming out one day in the future”, even more than VVOLs. The wait is not over yet, but a more detailed roadmap and info about their availability have been released at VMworld. VAIO libraries will require at least vSphere 6.0 Update 1, and obviously a software solution that uses VAIO, either for caching or replication, that are the two use cases at the moment, with other solutions like Encryption coming in the future.
Photon OS / JeVM
I’m not a devops kind of guy, and I’ve always worked at the infrastructure layer. One of the most difficult things is to have a proper conversations between IT guys and developers, even when things designed for devops like containers are involved. IT wants to see what’s happening at their environment, and be able to optimize resources consumption, but so far container were not easy to deploy and manage at the IT layer. A single linux VM running Docker for example would be able to host many containers, but from outside the only visible result would have been a single virtual machine. And speaking of virtual machines, VMware had its issues in trying to convince people that the hypervisor was still the best way to run containers rather than a linux OS running on bare metal. Thanks to project Photon things are going to change: VMware developed a new highly customized version of linux called Photon OS, which is already slim and fast. But when executed together with vSphere Integrated Containers, it uses a “pico” version of it called JeVM (Just Enough VM) that only consumes 24 MB on disk (yes, it’s that small) and can execute one container. It’s a container environment as seen from a developer point of view, but the beauty is that it’s seen as a VM in vCenter. Think about using HA, vMotion, DRS against these containers and in this way immediately offering that kind of High Availability that was missing natively in containers or usually solved at the application layer. For enterprise also, this means to leverage a proven and well known platform like vSphere to run containers rather than moving to some custom Linux platform that may be good, but not well known by admins. Also, thanks to another technology called Instant Clone a Photon OS machine can be cloned in hundreds of copy in less than a second! Again, I’m not a developer, but I can see the benefits of using such a platform leveraging a proven platform like ESXi/vSphere, the Photon Controller as the orchestator, and the JeVM as the final container (sorry for the pun) for containers.
One thing I’ve noticed: people that work in devops environments as the ops guys were super excited, so it should be cool.
VVOLs, VVOLs, VVOLs!
After being around for some time in the beta phase, and with the chance to see many demos of this new technology from different storage vendors, during the last VMworld 2014 VMware announced finally the general availability of this new storage solution, that then was available in March 2015 together with vSphere 6. However, with few remarkable exclusions available at day one like HP 3PAR and few others, it’s taking quite some time for the market to have different storage vendors supporting VVOLS, and even today after almost one year, the list of available solutions is still small.
I’ve got a couple of discussions around this topic at VMworld. First, I’ve heard from customers that some of them are already actively testing VVOLs. Maybe not yet in production environments, or maybe what they meant was that they were running low-profile virtual machines. As much as it’s an interesting solution, VVOLs is still a 1.0 technology, so I don’t blame end users waiting a bit to put them in production. Also, one of the problem I see and that I exposed in these discussions is a sort of a loop: customers are not using VVOLs because their actual storage (or the one they are planning to buy) do not support VVOLs, and on the other side storage vendors maybe are not rushing to support them because there are few requests.
But things are changing quickly, and I’ve seen many announcements of storage vendors going to support VVOLs. I’m not going to list all of them, also because I had not enough time to go and check with all of them if support was just announced, or if the update to support VVOLs was alredy available to users. One vendor tough needs a mention: I’m really happy that someone at EMC maybe read my ranting blog post I wrote some months ago (Are you a Software Defined Storage vendor? Become our hero!) and released a Virtual Storage Appliance to help people test VVOLs in their lab. This is not an improved version of the vVNX (Virtual VNX), but an appliance called VVols Tech Preview It’s free to be downloaded here. This appliance has only VVOLs capabilities, but nonetheless kudos to EMC for this and for sure you will see finally many new blog posts from me on VVOLs, now that I’m going to have them in my lab.
Cross-Cloud vMotion
vMotion is by far THE technology that better identifies VMware. When I first saw it live, I think just like all of you I’ve been amazed. It has now become so commonly used that we do not even consider it as an incredible solution, but we still should. Especially if we think for a moment that after vMotion between ESXi few years ago we got vMotion between vCenters, and now VMware has expanded it even further with Cross Cloud vMotion. What does it means is that a vSphere environment managed by vCenter can vmotion a VM without interruption to Vmware own vCloud Air platform. This is possible also thanks to NSX and its network extension capabilities.
The solution per se is extremely interesting, but I’ve also heard some service providers using vCloud Director a little bit disappointed because there has been no clear answer if this solution (or even SRM to vCloud Air is another example) will be avalable also to Service Providers to be offered to their customers.
vCenter HA
Ok, this is not something that will arrive in vSphere 6.0 U1, but it will come in 2016 together with the next major version. But finally there will be a form of HA thanks to a active/passive configuration, where two vCenter servers will be connected together (plus a witness) without having a shared storage for the quorum. Also the PSC component will have HA capabilities without requiring an external load balancer. It will not be a cloudnative-scale-out-active-active-nosql-html5 sort of solution, but nonetheless we will have a solution that is no more a single point of failure.
Final notes
I’ve read around comments from different bloggers (obviously not those working “for” VMware) that in the last couple of years there has not been that WOW effect in the announcements. This is probably true if we look at the “historical” vSphere components like ESXi or vCenter, but technologies like VAIO or PhotonOS are really great in my opinion. Simply put, VMware these days is many more things than just vSphere, and sometimes enhancements comes from different technologies than the simple hypervisor.