Today is the first day of VMworld Europe 2019 in Barcelona! The team and I have been here since last saturday to prepare for the event and relax a little bit in Barcelona! So since we are all relaxed, lets kick of this first day!
Since Monday isn’t actually the “real” start of the conference, most of the booths in the main hall (VMvillage) such as the Hands-On-Labs (HOL’s), Cloud City, Community area’s etc weren’t all open or build up yet. But that’s not a problem. There’s still plenty to do and to visit! So after we’ve checked out the main hall we started to go to our session(s).
Technical Advisory Board (TAB) for VMware Cloud Providers
Just like last year we had the opportunity to join the VMware TAB session for Cloud Service Providers. This session brings the attendees a large piece of information regarding key VMware topics, innovations and upcoming strategies which they want to have some feedback on. I feel that this feedback isn’t pointless and that it’s actually being used to enhance the visions and ensure that we (the Cloud Providers) get the opportunity to launch the services we need to to grow our businesses.
The contents of this session are unfortunately under NDA so I can’t blog about it. But I can tell you that there are some awesome things coming in the future regarding NSX-T, vCloud Director, vCloud Availability and other Multi-Cloud related software packages and services.
A fun fact to close out the notes for this session: All of the VMware Cloud Provider virtual machines (more than 10 million), make us the 4th largest Megacloud in the world. This is also something Rudolf and I told our audience when we did our presentation on the vMA Techcon a few weeks ago.
Supporting Machine Learning workloads and GPUs on vSphere (MLA3014TE)
I’ve been to a couple of ML and DL sessions before and I try to educate myself on this area, but I’ve noticed that this topic is so complex to understand. Well, for myself that is! This session explained how ML and DL is done and how it works on a global level. I highly recommend to watch this session at home if you are interested in ML and DL in general.
The other part of the session went on about how GPUs can be used on vSphere environments. As most know there are a couple of options on how you can use the GPUs. Below I will explain them very shortly:
- DirectAccess I/O
- Directly passthrough the physical GPU to your virtual machine. This however means that you can only let one virtual machine use the GPU and you lose some vSphere features such as vMotion.
- Nvidia vGPU
- With Nvidia vGPUs you can slice up the physical GPU into virtual GPUs. These can be assigned to your virtual machines in vGPU profiles.
- You can assign vGPU profiles to multiple virtual machines.
- You don’t lose vSphere features such as vMotion and Suspend.
- Remote GPU Access with Bitfusion FlexDirect
- With Bitfusion you can remotely serve your GPUs to virtual machines. You can de-couple the GPU’s from the production compute environment.
- It slices any GPU’s into virtual GPU in any size.
- You do need RoCE and a 100Gbit network connection.
- You also have vSphere vMotion since the vGPUs are connected through the network.
The last one is specifically for AI workloads, but I think it also works great if you don’t have any hardware that supports GPUs, such as blades or small form factor servers that don’t have any physical space to house the GPU cards.
This is just a very short explanation how you can use them in your vSphere environment. I want to get some blogs up and running in the future in which I will explain them in more detail including some installation guidelines.
Partner General Keynote session
This session is always something we just have to visit. During this session VMware expresses their “love” for their partners. They dive deeper into the vision and strategy that they have for partners. And since we work for one of the largest VMware Cloud Provider partners in the BeNeLux…. Well you guessed it.
VMware CEO Pat Gelsinger started by telling everybody that he climbed mount Kilimanjaro last year and how that journey was very hard for him. The reason Pat told us this story is because he actually didn’t climb the mountain himself. He climbed with expert guides, which carried most of this luggage and because of that him climb the mountain. VMware wants to position itself just like these expert guides. Guiding all of the global VMware partners into their Multi-Cloud journey. Which as far as I’m concerned, they are doing for the company where I work. The VMware services allow us to easily manage and deploy workloads across clouds.
After this Pat told us the VMware vision again. It’s still the same as it was a couple of years ago. Any device, any application on any cloud. They are refining the vision once new insights (or acquisitions, like Carbon Black) are gained, but in essence the vision is still the same as it was since it’s been announced five or six years ago.
Next they made the below statement, which I completely understand and agree.
Technologists who master multi-cloud will own the next decade.
I’ve been seeing a market shift to a Multi-Cloud strategy for a while now. The customers no longer want to place their workloads solely on one of the Megaclouds like Azure, Google or AWS, but also want to have some workloads in national datacenters or in their own on-premise environments. One of the main reasons for that is cost (especially hidden costs), the migration pain and the other one I see is data locality requirements. Migrating complete IT environments to a cloud, whatever type of cloud that is, is extremely difficult and takes a lot of work, money and time.
VMware based services for partners such as HCX, NSX-V and NSX-T, vCloud Director and vCloud Availability helps us to assist the customer in this journey. The migrations to other cloud environments are easier because of this. In some situations customers can even migrate their complete IT environment on their own. Great job VMware!
After this we had some drinks at the Partner Reception and went on with our evening with a nice dinner!
Stay tuned for more tomorrow!
0 Comments