Citrix Provisioning Thoughts

I'd like to briefly share my thoughts about Citrix Provisioning Server which I have come to love and respect over the past 6 years as a product.  First of all, I love Provisioning Server; it is a dynamic technology that has just gotten better and better over the years.  It is easy for me to install, configure, and troubleshoot this powerful, reliable, efficient and amazing product after having installed it for over fifty customers in the D.C. area.  And from installation and configuration to documentation and technical support, Citrix has made significant improvements including the Virtual Desktop Handbook.  So, I could pretty much "slip and fall" and set up a solution that includes Provisioning Server.  However, there are some other things to consider when implementing awesome technology of Provisioning Server for my customers who require rapid provisioning capability.

There is a certain amount of complexity to install, configure, troubleshoot, maintain and upgrade Provisioning Server.  Anyone familiar with PVS can attest to that, especially when they have to coordinate with a team to configure the DHCP server, have to work with a different team to coordinate the hypervisor virtual switch, another team to configure the physical networking, another team to verify there are no other PXE or TFTP services on the subnet, another team to coordinate the physical servers, another team or individual to coordinate permissions on vCenter, another individual to coordinate storage to present to the PVS server and yet another team to establish a naming convention for the session host target devices.  And when none of those people are the primary technical and project contacts, there can be challenges making all that happen.  There is some additional infrastructure required and there are many different things that can interfere with something as simple as a reboot.  And those things might not be Provisioning Server’s fault, but if a target device can’t get an IP address for whatever reason, that target device isn’t going anywhere.

Something that needs to be considered is that there may be a lot of concepts that are new to a customer when introducing a virtualization solution into their environment.  For example, it may be that small team’s first attempt at virtualizing workloads in general.  Most customers already have a virtualization presence of some sort, but you'd be surprised at some of the smaller shops that are doing some very big things and have not yet broken into the workload virtualization space.  So, working with the hypervisor and management of that hypervisor and the network and shared storage demands may be new to them.  It also may be the customer's first time configuring PIV authentication for remote access purposes.  It may be the first time they have to deal with remote profiles (that is, profiles that are not persistent on the end point device).  It is most probably the first time for the customer to deal with Server and Desktop OS Delivery Groups.  It may be their first exposure to personal vDisk objects.  And, when I go onsite, it is definitely the first time for the customer to deal with desktops in the datacenter.  There are a lot of shifts in mentality and expectation right there and none of them have anything to do with Provisioning Server.  You can probably see where I’m going with this.  There are a lot of new things going on with one of these projects in the first place, even before you make the decision for provisioning your desktops & servers with either PVS or MCS.

One of the reasons for this may be that the customers see one of us highly skilled virtualization consultants come to their site and they want to have us do all the things that they have been interested in doing but don't have FTE's that can do it.  Whether it's a skill situation or a priority focus on other tasks is inconsequential:  if you are a CIO who now has a consultant who's an expert in all things that are virtualization-related and they’re already onsite, why not do *everything* while that expert is there? Those experts aren't cheap, are they?  And they can do it, right?  So, I can see how someone might make a decision to do Active/Passive Global Server Load Balancing with Citrix XenDesktop 7.6 with Personal vDisks, Provisioning Server, StoreFront and the latest 10.5 NetScaler Gateway.  And let's do PIV authentication, and replication for SQL and fileshares and upgrade VMware while we're at it.  And, those new Converged Network Adapters should be added while we're doing that!  So, a lot of things get thrown into a project since it is a logical boundary, rather than give thought to "Do I have the right people in place to support this solution once that expert who sets it all up is gone?"  And right there, at Accelera, we have the answer.  We recommend a Managed Services offering no matter what we provide. This allows us to put all the latest and most powerful technology into play at a customer site and support it smoothly.

Don't get me wrong: I love all of the virtualization technology.  I love having a job where I get to play with technology like toys.  It is cool to use Provisioning Server and vSphere and Hyper-V on blade servers with Windows Server 2012 R2 using Profile Management with DFS-R and a NetScaler or a Big-IP F5 LTM/GTM/APM for remote access.  We can deliver virtualized applications with App-V or ThinApp and provide impressive diagrams and spreadsheets that document and show the configuration from many different views for all the different products.  And I love being able to provide a thousand virtual desktops with persistency and solid profile management in days instead of weeks.  I think it's cool; I started off my desktop virtualization focus 6 years ago with Citrix Provisioning Server 4.5 SP2 and XenDesktop 3.0, and I consider PVS to be my friend.  But, it is *my* friend, and what may be my friend may not necessarily be my customer's friend, even though PVS has inarguably gotten better and better year after year.

Besides the pre-boot environment, TFTP/PXE server and BOOTP-compliant DHCP server dependencies, there are also reverse-imaging/direct-boot method updates for NIC drivers and the PVS agent itself. Sure, an expert can very easily account for and explain and troubleshoot these items if necessary, but I wouldn't wish this on my customers.  And there are other complexities, of course, such as providing suitable shared storage & replication scripts/technology for VHD file system objects, vDisk versioning, sites, networking, updating, training, and managing multiple target device collections in addition to the XD Machine Catalogs which are a must.  Once again, we need to provide that Managed Services Offering for our customers, even if the solution is without PVS: hopefully I've shown a few instances where the customer is introducing new technology and concepts even without PVS.

Look, I personally love PVS and think it's the best.  It is the best and there is nothing in my humble opinion that can compare to it when you need to provision thousands of virtual desktops or RDS servers.  But my love and admiration and respect doesn't mean it is the best fit for my customers.  In those cases where we are sub-2000 virtual desktops, and that's a lot of very valuable customers that I'm talking about, I'd have to recommend MCS for Citrix and Linked Clones for VMware Horizon View.  I might even recommend MCS for 4000 virtual desktops using two machine catalogs.  With all the optimizations that have been made in reference to networking and storage, and with the reduction of complexity and with the additional capability of provisioning Server OS Delivery Groups through MCS, I am now more inclined to recommend Machine Creation Services.