Need to work on my titles, but that’s what I have for now. I want to spend some time reviewing how NSX Distributed, Edge, and DLR Firewall services behave in respect to vCloud Director and vCloud Usage Meter.
Some may know vCloud Usage Meter provides automated feature detection on a per-VM basis – essentially allowing granular-level billing for tenants. For example, tenant ACME may have 5 VM’s that need Distributed Firewall out of 20 – why charge them for all if they aren’t using it? Since version 3.6 of Usage Meter, this feature has been applied on a per VM level, for the most part.
There are three different versions of NSX available in the Cloud Provider Program –
We can see that your basic fundamentals are available in SP Base: your distributed switching and routing, Edge Firewall services, NAT’ing, etc. Advanced adds in Distributed Firewalling, Service Insertion and other advanced functionality while Enterprise is adding in X-vCenter NSX, HW VTEP integration.
Second, let’s take a look at the chart for NSX Editions by Feature. This comes from our vCloud Usage Meter Feature Detection whitepaper. Don’t have it? Get on VMware Partner Central and grab a copy.
Take note at the statement of “All VMs on all networks serviced by the edge” – this means even if you have a VM on that edge that does not use that feature, it will charge for it since it still has access to it. For example, if you set up a L2VPN on an Edge, be aware that any other connected networks will be charged for NSX Enterprise even though they may not be able to traverse to that tunnel!
I wanted to design a DLR environment within vCD along with using Edge and Distributed Firewall rules. I came up the following design in my 9.x environment:
- Creating an NSX Distributed Logical Router (DLR) is pretty easy. Select the Advanced Gateway box and check Enable Distributed Routing. You can also enable Distributed Routing on any existing Advanced Edge –
- The great thing about how vCloud Director creates Distributed Routing is it’s actually two Edges – a standard perimeter edge along with the DLR southbound. The transit interface is created as a /30 while applying static routes for any connected DLR interface. DLR’s default gateway is the Edge transit interface. Below is what I can see inside of NSX.
Using the NSX Distributed Logical Router is in the NSX Base bundle and covered by Advanced SP Bundle, which is great for providers. However, Edge Firewall services are only available in the NSX Base bundle, where some providers may want to provide “high-powered” firewalling services.
The key difference between Edge and Distributed Firewall services is we do not hairpin for forwarding decisions in the DFW model: the decision is made inside of the hypervisor before the packet ever egresses anywhere. Hence, provides better throughput and capability compared to Edge or traditional firewall services – check out Wissam’s discussion on DFW capabilities here.
vCD Access to Edge Firewall vs Distributed Firewall
I think Tomas did a great job of summarizing vCD Firewall capabilities and pointing out some of the architecture differences between the two firewall technologies.
From a vCD UI access point of view, you have to access the firewall services in two different places.
- Edge Firewall services are accessed by right-clicking on the Edge Gateway and selecting Edge Services. Makes sense, right?
- Now Distributed Firewall management is accessed by right-clicking on the organization VDC and clicking on Manage Firewall. Not my favorite and doesn’t state DFW, but this is how it’s accessed today in the Flex UI.
NSX Distributed Firewall Changes
Okay, let’s get into a few scenarios and test out how Usage Meter will put up the changes.
- I freshly created my three vApps: T1-App-1, T1-DB-01, T1-DLR-tclinux-01/02a (Web Servers). We can run a VM History Report and pick them up and see they all have been associated with the Advanced Bundle initially. I will try to point out the vROps and NSX column in the following discussion, but we will be looking for a checkmark and an “A” that refers to Advanced usage.
- So let’s go ahead and crack open the DFW and start carving up some rules. This is a simple 3-tier architecture so I’m going to lay out allowing web, web to app traffic, app to DB, allowing SSH to App and DB, and blocking everything else in Web, App, and DB Tiers.
- As you can see, the Applied To for my deny rule is only applied to my three Distributed Routing tiers (or routed Org VDC Networks). Not only is this a good NSX practice, this cuts down on who gets charged for NSX Advanced!
- Now looking at the DFW rules from vCenter, I can see vCD created the tenant folder and applied the policies in order. Very streamlined.
- Now let’s check to see if my new rules work. Yep, I can still SSH to my VM’s but do not receive any ICMP replies. Perfect!
How did Usage Meter detect the changes?
- So let’s now take a look at the Usage Meter VM History Report. What’s fascinating is Usage Meter tracks every move called state change. We can see in my tclinux-vApp (or Web) that it tracked when it started using Base NSX usage, then switched to Advanced, then my vROPs Enterprise instance picked it up and added it to its inventory. All in a very short timespan. Very useful!
- With the DB and App layers, we see something very similar – vROps Enterprise, however, picked up these VM’s first and then we see Advanced NSX usage.
- Now, I can see a few new line items populated on my Customer Monthly Usage and Monthly Usage report. I now see Advanced Bundle with Networking, Advanced Bundle with Management, Advanced Bundle with Networking and Management.
- This is to be expected since my VM’s went through a few iterations – my Web VM’s initially went with NSX Advanced without vROps which falls in the Advanced with Management while DB/App started using NSX Advanced without vROps registration (Advanced with Networking bundle). Last of all, all three are now assigned to the Advanced w/Networking and Management since we are using NSX Advanced features along with vROps Enterprise.
- While I see the new line items, I don’t see any units to be reported. Why? Well, I have some very small VM’s (256MB in vRAM) and they just ran a few hours during this testing. Keep in mind Usage Meter averages out the usage over the entire calendar month (720 hours for a 30 day period or 744 hours in a 31 day period). Again, to be expected but was pleased that new categories were added immediately.
In summary, Usage Meter does a great job of detecting NSX feature usage and bills it accordingly based on a specific service – Distributed Firewall being one of them. Thanks!