Tech Field Day 10 (TFD10) - VMTurbo Use Case Hybrid Cloud

Video created by megan.gay Expert on Mar 8, 2016

    Q&A Transcript:

    2:24: “Do you guys provide the sFlow collectors as well as part of the product or is that something that I need to collect and then send it to you?...So are there specific collectors that you support?

    Answer: “Well we add it as a target. Everything we're doing is through rest API’s so you’ll give us your credentials and we’ll instrument it that way.

     

    4:50: So right now you're talking about being able to detect VMs that are communicating and moving them closer. What about doing the opposite?

    Answer: “It all depends on the the price. That’s similar to two VMs that consume a lot of memory and that would cause to drive prices of that commodity up on whatever location they are running on and then they would start shopping and if they have anywhere else that is cheaper to run the economics scheduling engine will generate an action to move them there.”

     

    5:30: Do DRS rules take precedence over VMTurbo? Let’s say we have a clusters of servers and we’ve created a DRS rule but for some reason they have overflowed the boundary and VMTurbo wants to put them...”

    Answer: “First of all, we work very well with DRS. I will say a lot of time people will lower the aggressiveness of DRS. It’s a very different approach to deciding when to move workloads, but there’s no interference.”

     

    8:07: “In order to understand how this works, you seem to have to have somebody very intelligent to drive it and I just wonder does that not take away from some of the value or is that not the case because what you’re describing in some of the corner cases some of the things you can and cannot do you have to know how the system is reacting and how it will do certain things and I sorta get the impression that in order to drive this and make full use of it you have to be incredibly skilled and spend a lot of time using the product...”

    Answer: “I think it’s the opposite, the system is so smart very intelligent it requires ideally no one to run it. If you turn on automation then all you need to do is sit back and watch all your actions disappear. For a system administrator it’s about first of all getting him trust that these decisions are better than them going and investigating what to do on their own.”

     

    10:05 “So with your existing customers then what do they tend to do? Do they tend to trust the system and if so how long how long does it take them to get to that point and have they been burned?

    Answer: “60-70% are prospects not customers, they are still going through the evaluation process, automate before they buy. So they are trusting and automating motions. Typically cause they are used to that capability.”

     

    11:20 “When you say you move a workload from my data center across to the cloud what exactly do you mean by workload because data takes time to copy around...

    Answer: “We only burst out workloads to the cloud that are horizontally scalable. Either you define them as such or if we detect that they are connected to a load balancer. These are the only types of workloads that we will recommend to burst to the cloud and bring them back once resources are freed.”

     

    12:35 “So you just said if it detects a load balancer how are you detecting that? Based on source destination something like that? You have to manually do that?”

    Answer: “No you add it as a target. You just give us an IP.”

     

    14:06 “So can you tie into ACI as well?”

    Answer: “Not yet that is coming up in the next release.”

     

    15:43 “So do you actually recommend expanding the physical resources as well on prem?

    Answer: “So that depends on the type of demand pattern that we detect so if you consistently need more resources and you don’t have enough in your private data center we will first recommend that you clone to assure application performance right away, but we will also generate an action to provision you hardware and once you have that available, workloads that can move back to run on prim we will move back.”

     

    16:37 “How am I notified, do I have to come to the console to know that there is an alert waiting or do you get an email, can I configure, how can I be notified, can I be paged?...And then the second part of that question is if I am automating everything how can I go back and view what has happened?”

    Answer: “We do have a reporting capability. Reports on what’s happened, what’s pending, what’s happened over the last time period whether that’s a couple days, a month.”

     

    19:50 “So do you can actually kick off the workload to autoscale the environment for me and how are you doing that?

    Answer: “Yes if you automate.”

     

    22:08 “But what about the predictable flippiness, black friday, super bowl sunday, seasonal, do get all that ahead of time, like you can teach it?...like I know when this super bowl sunday is going to be and when the next five are and they're not always the same day.”

    Answer: “One of the things we have is a planning capability and one of the things you can do with this planner is to set it on a different demand baseline. So again the economics scheduling engine it works on the idea of supply and demand. The supply is the infrastructure that you currently have but the demand is what current demand you have on your workload which could be very low right now but because we have historical data we also calculate the worst and largest demand you have ever had in your system. We can reconfigure and plan based on that worst case.”

     

    25:02: “What I want to do is say I am going to spin up some resources ready for this workload to arrive so I’m prepared. I just want to make sure VMTurbo doesn’t turn it off.”

    Answer: “So you can add reservations to the system. You can tell the system to reserve whatever you think it needs. So you can say this type of Virtual Machine I think will be needing a thousand more of those and you can create reservations for it. What the reservation means for the economics scheduling engine, we basically create fake virtual machines, add them to the economic scheduling engine and everything behaves as if those virtual machines are already there.”

     

    31:51 “How do you understand what the demand for that workload is on AWS or Azure?

    Answer: “We have the information that we get from the public cloud providers and then we have our application control where you can add applications as a target just as you can do on your private cloud.”

     

    33:20 “But if I had my own custom map I would have to somehow come up with a way to spit out results or metrics, connections? I’m out of memory CPU whatever ever I wanted to present.”

    Answer: “We have an SDK, you can write one, it's very simple and we have a lot of examples of how to use it and people are sharing their templates on the Green Circle Community.”

     

    39:20: “I’m just thinking BCO workflows, Jenkins, things like that, being able to do things right, that maybe are not built into the product but have those hooks to where I can generate additional workflows that can’t be accomplished by this.”

    Answer: “VMTurbo is really the brain, the decision engine, it’s going to tell it what to do. Now we're not in the business in going out and building all those automation scripts. As those platforms and tools grow our platform can connect into all of them. We just want to stay focused on driving the best possible decisions in the most interesting environments possible.”

     

    42:20 “We talked a lot about bursting into the cloud for horizontal applications and I am wondering how the model works in the back end considering data needs to be moved to the cloud to be be able to do something with the application itself on that location, so how does VMTurbo actually consider these stateless applications as a whole?”

    Answer: “We don’t recommend to move stateless applications between private and public clouds or between the different public cloud providers just because it is very inefficient and the downtime can be very very long and it can be very costly.”

     

    43:40 “Do you consider the different economics for the different stateless applications? Because stateless applications for instance you don't need to move back, if you don't need the recourses any more you just kill it? Do you take those different economics into account?”

    Answer. “Yes, definitely.”

     

    45:30 “You talk about the container part, is that like what you like talked about earlier on? Where virtual machines are communicating with each other? Is that how you defined a pod again”

    Answer: “No this is a container pod. It’s a unit in kubernetes. It’s the smallest kubernetes unit. This is specifically something that is discovered via kubernetes.”

     

    47:00 “I’d be interested to know how many of your customers are using containers and kubernetes?

    Answer: “Just a few, maybe more than a single percentage.”