One of the best parts of my job is that I get to work with very smart people, providing me with the opportunity to learn something new or look at things from new angles, every single day! Recently, I had the pleasure of hosting the anchor leg in the 2017 cloud computing predictions video roundtable, in which I asked cloud and networking analyst Dan Conde and systems management analyst Edwin Yuen about their thoughts on cloud trends in 2017 as they apply to converged and hyperconverged systems.
Edwin answers my questions about whether we can and will have true workload portability between on-premises and public cloud services, and Dan answers my questions about the impact of hyperconverged deployments on networking. Enjoy—and feel free to send along comments or questions!
Terri: Hi, I'm Terri McClure, Senior Analyst for Cloud Infrastructure and Storage. I wanna ask the other analysts who cover cloud-related technologies what their thoughts are on what will unfold in 2017 when it comes to managing and scaling private and hybrid clouds. Joining me today are Edwin Yuen, who covers systems management and Dan Conde, who covers networking. Edwin, we're seeing many IT organizations looking for hybrid cloud solutions. They wanna do test and dev in the cloud or use the cloud as a backup target or for burst capacity or...or for apps, they don't feel the need to manage in-house. And while converged and hyper-converged solutions are ideal for building an on-premises portion of this, how easy is it really to manage and move workloads between on-premises infrastructures and the cloud? And do you think we really...We'll have in 2017 true workload portability that runs on a common management stack, for on-premises and cloud-based apps?
Edwin: The question you ask is a very important one. Right now, in its current state, there's relatively very little workload migration or bursting happening, primarily because the workloads are VM-based and unless you have a common hypervisor platform, there are conversions to take place. The conversions themselves were to prevent bursting, which requires movement not only up, but also back down from the cloud. However, in 2017, there will be new options. We have the new VMware on AWS cloud option, which allows for ESX-based VMs to migrate and burst up to the AWS and then back down again. And we also have new options like Azure Stack, which will provide an application API compatibility so that applications can be developed on-premises and then be moved back and forth. As well as container technology, which offers portability of format for existing applications. Depending on how all those options mature, 2017 may be the year we see true portability in workloads.
Terri: Dan, while I know you cover several cloud-related areas, networking is probably the one you cover that is most overlooked when it comes to converged and hyper-converged solutions. With converged solutions, we see networking components integrated into the solution. Hyper-converged solutions often don't include networking though. What the hyper-converged vendors don't discuss is the fact that because of the scale-out nodal architecture, there's really no north-south network associated with these systems. All the network traffic they generate is east and west on the communications networks. So what should IT teams be thinking about when implementing hyper-converged solutions from a networking angle?
Dan: I think IT teams really need to understand the workload characteristics and look at the networking based on that. Most modern leaf spine networking architectures will fit into a hyper-converged solution where there's a lot of east-west traffic such as Big Data. What's more important is whether or not IT teams are to look and think of SDN systems that complement their systems software which may include virtualization like leaf sphere [SP] or app container systems which have many alternative networking solutions. Adoption of SDN in data centers has been slower than many of us expected. And I think people are looking for a quick reduction in TCO, which was not on the cards. The bigger savings actually happens in the edge with SD-WAN. But back to hyper-converged, I think customers can start with what they have and slowly adopt new networking architectures. The old systems will work and they don't have to do a wholesale swap out. But I'd say that if they are going to make a switch to a leaf-spine, it would be a great time today to do that.