This post is live-blogged from the VMTurbo presentation at #VFD4. The structure and organization will undoubtedly be poor, since this is all captured on the fly from our chat.
Overview
VMTurbo likes buzzword bingo. The official title of their introduction is “Demand-Driven Control for the Software-Defined Universe.” Lol. Either way, it’s GREAT to have Eric Wright back with us, this time as a presenter instead of a delegate.
VMTurbo aims to analyze and optimize the environment to achieve “desired state” which they’ve defined as a balance of maximum efficiency and application performance. They paint the data center as an economy, and all the moving parts as either supply (providers) or demand (consumers). The software will attempt to achieve desired state by allocating/reallocating resources based on that supply and demand. It’s really cool to think about this; if we consider migrating a VM to get two VMs on the same host to achieve high network throughput, Operations Manager will do a cost-benefit analysis and determine if the gain from that move would outweigh the cost from using up an additional X amount of CPU and RAM on that same box. That’s just a simple example.
Deep-dive
When I first used Operations Manager and heard from VMTurbo, it was focused on the hypervisor. Now, the product encompasses networking, hybrid cloud, OpenStack, and more. In talking about how VMTurbo views that ‘desired state’ we started digging into VMTurbo vs. DRS for a minute, and Frank Denneman piped up on Twitter which cracked me up. Eventually, Christopher asked them to move on to the demo, so now we’re getting to see the good stuff. We spent a lot of time going through the interface and looking at setting up policies and viewing reports and also some neat forecasting. I noted that I have always disliked, and still dislike the interface. It’s too dense, and to me it seems like it’s so complex that it’s hard to use. I have actually used it, so I’m not just speculating. I’d like to see a more elegant interface out in front of the admittedly powerful features. We did hear that they’re aware of this and working on improvements.
Found something really cool, which was a follow-up on a question Rick Schlander asked on Twitter about DRS rules. VMTurbo will treat your licensing constraints (for example, sockets licensed for SQL) as just another commodity in the economy! This is even better in my opinion than traditional DRS affinity/anti-affinity rules in vSphere.
Update: We talked more offline at dinner, and I came to a realization I wish I’d had earlier in the day – the algorithms, models, etc that make this whole ecosystem analysis work are already built. So what it takes to add an additional ecosystem member is just to get info from it via API, and determine its role as a supplier or a consumer in its relationship with other entities. Is essence, what this means is that its incredibly fast for development at VMTurbo to add tie-ins to new features that vendors release (VVOLs, for instance) or to new vendors/products entirely.
Wrap Up
That’s it for today! All the sponsors gave fantastic presentations and set the bar high for the rest of the week. Engagement was great, and nothing really came off the rails. Making it through a whole day with no hiccups at TFD is quite an accomplishment! See everyone again tomorrow for some more awesome sessions.
