VxWhat? Making Sense of this VxWorld

VxRail, VxRack, VxFridge, VxWho? We have definitely found a naming scheme and stuck with it. However I am finding that it has started confusing a lot of our customers. I was at VMworld 2017 and were asked a lot of the same basic questions. I hope to clear this up in an avenue that can be reached by many, rather than standing by my booth and chatting you up one by one. So lets get started by looking at some of the technical requirements and best practices around our products, then decide how we should approach our portfolio. Up first is VxRack.

VxRack:
VxRack is not a single product. It is a product line consisting of two products; VxRack Flex and VxRack SDDC. Both of these include a multitude of configuration options around core found and drive type (SSD vs HDD) / size (1 TB vs 10 TB). They  are Rack-Scale Solutions. Rack-Scale means that these products are the rack itself and all the components within them. Don’t think of VxRack as a server, but more so an entire rack. Included in these are Cisco top of rack switches and a management cluster to start the rack deployment. You do not manage the switches in this. They are part of the engineered solution. The only thing you’ll provide here is the uplinks to your network, the power for the rack(s), and the floor space.

VxRack SDDC utilizes VMware Cloud Foundation software built on VMware vSAN for it’s software defined storage. WIt requires NSX and VMware within its product. There’s not much flexibility in this, however, you now have a complete Cloud Foundation software that will support your infrastructure… neat.

VxRack Flex is built upon Dell EMC’s software defined storage application, ScaleIO. Think of this as taking your local server storage and turning it into a block storage array. The cool part of this is you can put whatever hyperviser you want on this; KVM, Hyper-V, or our bestest friend, VMware. You could also do a bare metal install of Linux on top of the ScaleIO software and gain all the data services and speed of ScaleIO while bypassing a hyperviser.
 

VxRail:
Now let’s move onto VxRail. This is my personal favorite solution as there’s a ton of engineering in it, as well as a ton of flexibility and options. Also Aaron Buley is bonkers for it; and that guy’s great. Check out his YouTube channel, SEV Ops, buncha dorks.

VxRail is not a single product. Just like VxRack, it’s a line of products. However, VxRail is standardized on software, vSAN. But where it differentiates from VxRack is literally at the switch. We have a pretty great BYOS (Bring Your Own Switch) policy on VxRail. Now we do have a few requirements for your switch, but its far more lax than the rigidity of VxRack switches. VxRail is a jointly engineered solution between our friends at VMware and Dell EMC. VxRail also does not have a management cluster for the VxRail deployment. You can run vCenter on the VxRail appliance or outside of it; its up to you. vSAN requires a minimum of a 3 node cluster to start. However we at Dell EMC recommend a 4 node cluster. This way if you lose a node, you don’t go into read only mode, which is the case if you only have 3 nodes. For more info on vSAN, go see Chad’s blog for an intro.
Within the VxRail product line, there are multiple node types. Each node is purpose built for a solution. These node types are G-Series, E-Series, S-series, V-series, P-series. These products have different node density and physical node size. However they all can come in either SSD or HDD storage types. As well as multiple different CPU configs. So let’s break these out a bit.

G-Series
G-Series, as in General Use.
Original VxRail.
4 VxRail nodes (N) into a 2 rack unit (U) form factor.
Great for small RU requirements

E-Series
Entry node series
Dell R630
1 node per 1 rack unit / 4 nodes per 4 rack units
Perfect for cost effective requirements

S-Series
Storage dense node
Dell R730XD
1 node per 2 Rack Units / 4 nodes per 8 rack units
When you need storage but don’t need extra chips and ram

V-Series
VDI optimized node
Dell R730
Option to have up 2 GPU cards
1 node per 2 Rack Units / 4 nodes per 8 rack units
VDI Workload beast

P-Series
Performant node
Dell R730
1 node per 2 Rack Units / 4 nodes per 8 rack units
As speedy as they get!

Now let’s talk about the misconceptions:
“VxRack is expensive”
Not really. You can’t compare this to just servers. You’re getting the whole package here. Servers, network, racks, cables, validated code levels. All of this is included in your price. Now there is a management cluster in the first rack to help manage the software and provide a home for vCenter, so you do have some overhead there. But as this grows, that cost deminishes as it becomes a smaller percentage of your overall VxRack deployment. The sticker shock of a management cluster is less on a 1,000 node VxRack vs a 4 node VxRack.

“VxRail is the little brother to VxRack SDDC”
FALSE. They both have vSAN, yes. But VxRail does not require NSX, nor is the switch part of the product for VxRail (remember BYOS?). VMware Cloud Foundation is also not a requirement in VxRail.

“vSAN is super taxing”
Nope. Here’s a white paper stating some best practices. I usually size for around 10% on the safe side. But I’m a pretty conservative person <insert southern joke here>.

“Data locality is the only way”
No its not. Or maybe it does… What’s your switching environment? 1 GB? I would make sure ALL 4 1GbE ports are ready for your specific VxRail models. However with the introduction of 10 GbE switching, this argument has died. Here’s a paper from a super smart dude about it.

“We are going to outgrow our VxRail deployment. Rack is the right choice”
Okay let’s talk some strategy here. How quickly are you looking at scaling? Remember we’re breaking some purchasing habits here. You can purchased incrementally if you’d like! So if you aren’t going rack scale before the end of the life-cycle management of the Rail… why bother? Also, remember you can totally still use VxRail for other products if you decide to first use a VxRail then a VxRack! Thanks to the power of vMotion, you can move all your workloads around easy peasy.

“VxRail is more ROBO.”
No. and yes. While it is a great ROBO solution, lets not forget that it can scale pretty far as well. The max scale is 64 ESXi hosts, which is the VMware maximum as well.

“VxFridge has 5 9’s of availability”
I’m not entirely sure about it. VxFridge is literally a refrigerator.

I hope this has help cleared the air. If it hasn’t, please leave a comment or shoot me a tweet. We can setup a webex and chat some more.

Til next time,
Kyle