eva.tuczai

Turbo Tip: Reservations Update

Blog Post created by eva.tuczai on Feb 14, 2015

Update - for a great example on how to use VMTurbo's Economic Scheduling Engine to drive the best placement for new workload, check out this GREAT post from one of our customers Andre lingenavd who is using our Market via the REST API to drive deployment actions!  Deployment recommendations by VMTurbo (Part 1) Deployment recommendations by VMTurbo (Part 2)

 

If you run a virtualized infrastructure, you have to be able to answer the question of how can I reserve resources for projects and workload that are coming in the future?  How can you determine if you have the capacity TODAY for FUTURE workload, and at the same time assure that you can run today's workload with tomorrow's project, while assuring performance and efficiency?  And if you do not have the capacity, what will tell you the minimal amount you need?

 

The Challenge is complexity of fluctuating workload demand in the current environment and managing future RISKS of over-provisioning.  This combination can put not just future projects but even existing workload at risk when you are ready to deploy. You also want to have visibility into the reservations that are in the system so everyone can see exactly why additional reservations cannot be met, or increasing demand from current workload requires you to add more compute or storage to maintain your reservations. 

 

VMTurbo has the Solution to Intelligently Plan and Manage the allocations of your resources for future demand, that allows you to reserve capacity for new projects, and gives you the answer of when additional compute or storage is needed.

 

The reservations feature in the DEPLOY tab was first introduced in v4.7 (VMTurbo Ops Mgr v4.7 What's New: Intelligent Planning with Reservations ) and in version 5.1 has been enhanced with the following capabilities:

 

1. Reservation Date: You can decide when you want to ask the Market if there is enough supply to meet the demand by creating a FUTURE reservation date (default is today).  The use case is to delay asking the system about available capacity when you know there is none today, and you have a project to add more resources.  Note, even if you ask today and the reservation cannot be met, when a reservation is in a pending state, the system will retry the reservation and will change the state when the supply is available to meet demand.

2. Fulfilled, Pending and Future States: Better visualization of which reservations are able to be fulfilled.  Note a fulfilled reservation may also be a partial one, where not every VM can be placed.  Click on the reservation to see which VMs are placed.

3. Run a Plan on the existing pending reservations. Use the Plan and Generate Report icons in the upper right.  Note, these will not include reservations in a future state.

4. Want to see what happens when the VMs deploy?  Run a Continuous Projection Plan, introduced in v5.1, from the Plan tab, and keep the default option to "Include Reservations".  If the projection period goes over the Deploy Date, then the VMs will "deploy" in the plan, and you can see the resulting utilization.

 

Considerations:

1. Over-commit factor: The default overprovisioning factor which drives decisions on available capacity is set to 10x Memory and 10x CPU.  These commodities are configured in the Policy tab and are global values:

 

2. VM Deployment: Note when you specify the DEPLOY date on a reservation, the VMs will be automatically deployed using the defined VM template by VMTurbo on that date, and then the reservation is deleted.  If you do not want the VMs deployed by VMTurbo, then CANCEL the reservation before midnight of the VMTurbo Server time (default GMT).  You can also use the VMT REST API to modify reservations.

Outcomes