Part 59: My vCloud Journey – Design Thoughts – Provider vDC
People who know me well, will know its rare that someone says something that shuts me up. Being a congenital blabbermouth that I am, if you say something interesting to me – its likely to inspire me to say something. So its rare that something is said that makes me take a step back with a realization that the way I’ve been thinking has either a fatal flaw or an assumption that at its heart I’d not questioned before. That’s the other thing – by my nature I’m a very questioning person. Sometimes I wish I wasn’t, and that my brain would give me the rest of me a chance – just accept stuff around me. Sadly, that rarely happens…
Such a situation happened on the recent design course I attended with Eric Sloof and gang of students who were as equally quizical as I am. What gave me pause for thought was a discussion about the Provider Virtual Datacenter (pVDC). In case you don’t know the Provider vDC is how vCloud Director presents the physical resources to the largely logical construct of the Organization. It’s common to talk of a Platinum, Gold, Silver and Bronze classes of Provider vDC. In my minds eye I often see the Provider vDC a bit like a building. Each building offers a different class/quality of experience to my tenants who occupy it. Of course, a tenant can choose to have their Organizations distributed amongst my facilities as they see fit. So they could choose to have their Test/Dev Org in my Bronze building and their Production workloads in my Platinum building. I’ve often said to people the job of the “Cloud Admin” is like the Supervisor or Facilities Management. I’m kind of thinking of the supervisor in the popular sit-com “Friends” who fixes (or doesn’t fix) the heating their apartment. Hopefully I’m a bit more helpful to my tenants than our man in the boiler suit.
The provider vDC points to a vSphere cluster OR importantly clusterS in vSphere 5.1/vCloud Director 5.1. Yes, that’s right a single Provider vDC can now contain multiple “buildings” or clusters – just in case 32 hosts in a single cluster with 1TB of RAM isn’t enough for you. 😉 So critically the Provider vDC is abstracting the complexity of the cluster away from the tenant – they just consume They don’t need to know anything about the underlying virtual or physical infrastructure. In my building analogy my tenants expect to pickup the phone; connect to the internet; turn on the air-con – and for it all to just work without having to think about what it takes to make those services happen. In fact for me its this whole Provider vDC, Organization and Organization vDC that makes vCloud Director stand head above the competition – who are happy to just “point” to the vSphere Cluster. To me their solutions don’t offer the actual level of abstraction needed to be truly defined as cloud. Couple with the automation creation of networks and firewalls in the vCNS Edge Gateway – I personally think VMware are on the right track.
Okay. So far so good. In my environment I’ve created a Gold and Silver Clusters. The Gold Cluster has 42Ghz of CPU/79GB RAM and 5.8TB of Storage, where as my “Silver Cluster” has 38Ghz/47GB of RAM and 5.64TB of storage. The storage presented to the “Gold Cluster” also includes access to fibre-channel storage. So it it makes perfect sense to classify them as Gold/Silver. Finally, the Gold Servers are based on AMD Chipset and the Silver Servers based on Intel chipsets. So even if I wanted to I could have them in the same DRS cluster.
So here’s what was said that stopped me in my tracks. If you took the storage difference away – and allocate 1Ghz of CPU and 1GB of RAM to a VM does matter which cluster or Provider vDC the VM runs in? The answer is it makes no difference at all. 1Ghz of CPU and 1GB RAM from an AMD Server or Intel Server is the same. But I’d never thought of it that way. I’d got so hung up on the classifications of “Gold” and “Silver”. That I just assumed that if I had VM that required a lot of physical resources “Gold” would be the best place to run. But is it? Really my Gold/Silver clusters are only different – in that one has more CAPACITY than the other, from a performance perspective so long as the VM gets the CPU/Memory the Provider vDC’s I’ve created are functionally (to the VM) the same. The only other difference is the CPU chipset – but that’s about chipset features not raw cycles per second. So what if the AMD processors don’t support VMware FT, but the Intel ones do. Big deal. FT is hardly a “Cloud Feature” given the number of primary/secondaries it supports per host.
What I think is interesting is how corrosive the terms “Gold” and “Silver” are. They imply one is better than the other. When in many case they are just DIFFERENT amounts of RAM/CPU. True my “Gold” Provider vDC has more physical resources, but so long as my tenants VMs get their resources they need does it really matter which cluster they come from? Despite me using the terms “Gold” and “Silver” I actually quite hate these terms. Mainly because for the last 2-3 years its all that we have talked about – and just getting a bit bored with the way we talk about cloud right now. This isn’t about the concept, its more about the language. You see I tire very quickly of stale language that is repeated ad nauseam, especially if I find myself repeating it as well… But I also think how these terms (gold, silver yadda yadda) are dangerous in the mind of tenants. I fear if perhaps these terms have been developed as way of artificially “sell” one class over another… As a tenant wouldn’t you prefer to be in the “Gold” Provider vDC than the “Silver” one – despite the fact we know there is no performance difference?