One of the exciting new features of the recent 4.0 release of Virtualization Manager is “Time Travel”.  The goal of Time Travel is to allow administrators to understand how the virtual environment configuration, relationships and performance has changed over time.  The reality of virtual environments is that they are dynamic – and becoming more so – with things like VMware DRS (and recently announced Storage IO DRS), enabled by vMotion and Storage vMotion.  For example, if I was experiencing slowness on a VM yesterday, there’s no guarantee that the VM in question is on the same host or datastore right now as it was when I was experiencing a problem – hence the need to be able go back in time for that VM and understand the historical performance in context of where the VM was at that time.  Let’s take a closer look.


The easiest way to see the power of Time Travel is to open the map view within Virtualization Manager.  The map view shows the different objects in the virtual infrastructure and how they are related, their alert status, and also allows sorting of performance data.  I know we were having an IO issue on one of our datastores last night – it was probably caused by VM back-ups but I’d like to understand what VMs were on that datastore and which ones were generating the most IO last night.


Firstly, let’s put the datastore in question “in context” so that we focus on that datastore and understand the other pieces of the infrastructure its related to.




When you put a single object in context like this – you’ll see a little Time Travel box appear in the top left corner of the map view.  Before we go back in time to last night, let’s see what VMs on this datastore are generating the most IO right now – simply click on the “Sort By” link and choose to sort the VMs by IOPs.  Click the “descending” check box to sort from high to low.  Looks like “Palm Beach” is our top talker right now.


If we now go back to 1AM around when our backups kick in (notice the Time Travel box has changed to a night view!) – we can see that the Miami VM was our top talker back at 1AM on this particular datastore.




“Time Travel” is also exposed in the performance analyzer (performance charting) within Virtualization Manager.  In this case – I have selected the same datastore Tick and plotted its IOPs for a time window between 12AM and 3AM in the morning – to catch again our back-up activity.




I would like to understand the VMs that were on this datastore within this time window, and which ones were generating most IO.  If I click on the related line button here – I am able to see Time Travel in action.  I am able to sort the (dependent) VMs that were on this datastore in this time window, by their peak or minimum IOPs (or any other collected metric).




This way I can now see a plot of the IOPs over time for these VMs relative to the datastore itself for the time window of interest.




So with Time Travel – you can troubleshoot what actually happened in the past without assuming the environment looked the same as it does now!