The Story of Project Fargo

The Story of Project Fargo


Project Fargo, aka VMFork, aka Instant Clone was among the biggest revelations of the vSphere 6 launch build-up, but come GA it was nowhere to be seen…

Wind the clock back to August 2014 and you’ll find Daniel Beveridge and John Dodge from VMware presenting a tech preview of VMFork as part of a VDI for Next-Gen Desktops presentation (EUC2551) to an eager audience at VMworld in San Francisco.  Formerly known as ‘Project Fargo’, VMFork was an implementation of University of Toronto and Carnegie Mellon University developed tech SnowFlock which was built to offer rapid virtual machine cloning for Cloud computing.  Touted by VMware as ‘A Fresh Approach to Application Publishing’, VMFork offered rapid cloning of running VMs fully loaded with Applications by spawning Child VMs from a quiesced VM using Copy-on-Write (COW), giving you a new, customised (in terms of hostname/IP/SID etc using a variation of View’s QuickPrep tech, only without the need for a reboot) clone that starts right where the parent left off in under 4 seconds.  Predominantly pitched at EUC use cases, VMFork would allow you to provision hundreds of desktops in a  matter of minutes (500 in under 5 minutes, in fact) complete with applications, all with a minimal footprint and boasting 120 sessions per-host (a miserly dual socket 8 core host at that) and an impressive 30% reduction in CPU against comparable RDSH sessions.  Not too shabby I’m sure you’ll agree, and the scene was set for Fargo (or VMFork) to replace View Composer with it’s cumbersome in comparison Linked Clones, and much more besides with the promise of vCenter APIs for the much anticipated ‘2015 platform release’ (these were still the times when referring to the 2015 platform release as vSphere 6 would get you feeling the wrong end of a VMware ‘Fork’).

Screen Shot 2015-03-27 at 09.00.59 (2)
Slide from EUC2550, VMWorld 2014

Mere days prior to the VMworld unveiling of VMFork however, VMware HQ announced the acquisition of fellow CA-based company CloudVolumes, a provider of real-time application delivery tech that would offer VMware, as VMware VP & CTO of Cloud Native Apps Kit Colbert put it, ‘a better way’ of delivering applications across all three of it’s key technology focal areas: EUC, SDDC and Hybrid Cloud.  CloudVolumes (later to be known as App Volumes) is essentially a Read Only VMDK with one or more applications installed on it (known as an ‘AppStack’) which can then be delivered to a user’s desktop (physical, RDSH or VDI-based mind) with a single click and hey presto, the AppStack will be made available to the user instantly as if the applications were natively installed- all nicely isolated with per-user data and settings, registry etc. stored in deltas leaving the underlying desktop blissfully unaware of it’s existence.
Hang on though, I thought Fargo and VMFork were about to revolutionise application delivery with it’s  ‘Fresh approach’ and ‘superior functionality’??  Step up ‘Project Meteor‘, bringing Fargo and CloudVolumes together to bring about a ‘paradigm shift’ in VDI provisioning and a ‘Just in Time’ assembly of a fully configured, app-laden Virtual Desktop on the fly.  A bold statement indeed.

Project Meteor, VMWorld 2014
Project Meteor, VMWorld 2014

Fast-forward to December 2014 and VMware releases App Volumes 2.5, which introduced Horizon integration to the CloudVolumes product it purchased not 6 months previous.  Compatible with Citrix XenDesktop, XenApp and RDSH App Volumes is a standalone application it’s worth noting, and one that comes with it’s own price tag…

But what of Fargo?  It was February 2nd 2015 and the day of VMware’s self-proclaimed Mega Launch and sure enough Fargo (now Instant Clone) gets special mention in VMware’s hugely hyped ‘One Cloud, Any Application, Any Device’ showpiece presentation where CEO Pat Gelsinger and CTO Ben Fathi walked us through the biggest ever vSphere release- vSphere 6, and the tech also took pride of place in Press Releases and Official Announcement docs.

Ben Fathi - One Cloud, Any Application, Any Device
CTO Ben Fathi shows off Instant Clone during Feb 2nd One Cloud launch

Here, Instant Clone’s reach is extended.  Alongside it’s obvious EUC use case to include Big Data and Containers- makes sense, rapid scale out and tiny footprints make Instant Clone an obvious tech to apply here, so all the more reason to get excited!  Also with that promise of APIs there are countless other opportunities to take advantage of what is without doubt an exciting piece of kit- dev/test for example, why spin up a multiple dev or test instances from scratch when you can Fork a VM, test what you need to test and then kill it in minutes??

So expectations were set, vSphere 6 in all it’s glorious glory was announced with it’s 650 new features and promises of a better world (and a better web client, the world rejoiced) of which Instant Clone was just one drop in a bucket of potential.  Then it came, vSphere 6 GA, on March 16th 2015.
Amid the excitement of SMP-FT, Long Distance vMotion and Godzilla-esque VM capabilities however, Instant Clone was notable by it’s absence in, well, just about everything.  A read through the ‘What’s New in the vSphere 6 Platform’ whitepaper yields no mention of it.  Hands-On-Labs?  Nope.  An allusion to it in the newly improved web client surely?  Well, no.  So where was Instant Clone?

Okay so no need to panic- Instant Clone was for EUC and Big Data right?  And VMware Horizon 6.1 was released almost parallel to the vSphere 6 GA so maybe Instant Clone is surfaced there?  A read through of the Horizon 6.1 Release Notes however reveals plenty of new vSphere 6 features Horizon 6.1 can leverage, but still no Instant Clone.
What about the Big Data?  Um, no, the last release of vSphere Big Data Extensions (BDE 2.1) was back in October 2014 there’s no Instant Cloning fun to be had there either.

So for all the build up and hype around Project Fargo, VMFork and Instant Clone, nearly two weeks from vSphere 6 GA there appeared to be no sign, not even a mention of the tech from VMware.  A dig around the net only finds a comment from Duncan Epping on an old post where he states that Instant Clone was currently for View and BDE only and was leveraged via Private APIs, and confirms that there was no Whitepaper to be found.  This seemed a far distance from the VMworld presentation back in August, not to mention the big One Cloud launch where the implication was the functionality will be built-in to vSphere itself.
Yet, App Volumes was alive and kicking and available to purchase…. has Instant Clone been scaled back to let it’s acquisition-brought brother take centre stage on the Application Delivery platform?  Did VMware show off it’s tech too soon and are having trouble bringing it to market?  Possibly.  What of the paradigm shift touted in the Project Meteor vision??
We were all left awaiting enlightenment, salivating over the use cases for this tech and aching to show it off as on the face of it, Instant Clone could have a major impact in some key business areas.

*UPDATE July 2016*

…And then it arrived!  Here it is in Fling form, on July 26th 2016 the VMFork for pyVmomi Fling was released bringing the VMFork APIs to your existing vSphere environment!  Enjoy!  Oh the possibilities….


Related posts