Help with QoE setup

We have put in a new network with a bunch of redundancy. I'm having a hard time figuring out the best way to implement QoE. I want to see all traffic from both our HQ and remote sites hitting our servers. My end goal is when someone reports something is slow, I want to be able to look from say the remote site, across the WAN, through the HQ network down to the server and application to see what might have been causing the performance issue. In general we have an RDS farm and document management application that I would like to be able to see performance from different location and where the trouble spots might be.

We have several remote sites, lets say 20. Typical setup is a local switch, router and a few servers. I'm not really worried about performance between PCs and servers locally at each site right now. Most complaints are from accessing resources across the WAN. Management wants to start putting data back out locally at remote sites. I need some data to show that coming across the WAN is the issue, or is it SAN performance or VM performance back at HQ.

So, local site --> switch --> router -- WAN -- HQ router --> dual core 4500s --> dual Nexus switch (all servers hang off these).

All our server sit in one vlan. Our 4500s run VSS. Our Nexus 5Ks are linked together via Port Channel and we have an FEXs hanging off each one of those. Everything has redundant paths as you can see in the image below. Most our environment is Virtual. It seems I would want to look at traffic on the 4500s. If I was to say monitor/span the server vlan or a port, I would need to do that on both 4500s as traffic could be hitting either. Also then what sort of destination would/could I use? We have mulitple ESX hosts, so not sure how using one VM could see traffic on another ESX host.

Right now we just have the free licenses for QoE, so 11 SPAS and 1 NPAS. Other products we have are NPM, NTA, SAM and SRM. If need be, I could invoke the 30 day trial of QoE if we need more to do this. I want to make sure and not waste those 30 days trying to get the configuration right though. Also if this turns out to give us good data, I have something to take to management to maybe purchase more QoE license in the future if need be .

Thanks

core network overview.png

Find more posts tagged with

qoe

quality of experience

network performance monitor

Accepted answers

All comments

silverbacksays

I'm not sure how up you are on how to physically implement the two kinds of sensors for QoE. That might be a good place for you to start, as it'll define part of the 'where'. Have a look at this thread:

NPM 11.0 Release Candidate Now Available!

It goes into detail about how QoE Packet Analysis Sensors (PAS) work, what the two types are and where they need to deployed.

Hope it helps you a little mate!

BakerD

Thanks silverback but I have seen that document and looked at the deployment guides etc. Nothing talks about what I'm trying to do.

So here is where I'm at today after trying a few things. On both 4500s I created a monitor sessions (span) with a source of the vlan that my servers sit in. This is sent to a destination port on each 4500. Those destination ports are linked to an ESX host. On there I created a new dbswitch for those two links and connected those to a new VM I built just for QoE. So the VM has 3 links, one to my server vlan so I can access/manage it etc. The other two are to each 4500 destination ports.

I added my new VM node to SW and then set it up as the Network Sensor, NPAS. I left it set to auto and it picked up 17 devices nodes which I can see in the Manage QoE Sensors section of settings. After letting it run for a few days it was seeing applications and servers, although listing a bunch of them as N/A. So this morning I went into settings and added our main WAN router as a node. It went from showing 100+ nodes with XX traffic to just 29 on the QoE summary page.

So still wandering if I have it setup right, so any help appreciated.

Thanks

adatole

That's exactly the way I would have suggested you do it. As long as you are monitoring the servers involved in the conversations, you will see the data start to show up in the QoE dashboard.

BakerD

Something is not right, looking at my QoE summary page I just now it isn't seeing much. I put wireshark on my VM and looked at both links and not much is going on. I would think data would just be flying. All I see are some broadcasts here and there. On the QoE summary, The All business productivity rating is showing only 8.4MB of volume.

If I have 3 NICs on my NPAS, how does it know which ones to look at that are connected to the span ports?