ESG AND DLR

[Sonikashah]

I am actually new to NSX and have some doubts in :

1.)  Lets suppose i have 5 compute cluster , 1 management and edge cluster for a DC.

How many DLR should need to be their as we are planning to go with 4 ESG.

I am not sure that why we have taken 4 edges . If i am not wrong then a dlr have 1000 LIF with 8 uplinks .

These uplinks can be used to connect to 4 diffent edges. However , will they be of same subnet or different subnet.

2.) How if the overall architehture.

According to my understanding :

Cluster()---------->>> logical switch----->>> DLR----->>> EDGE ------->>> VDS------->>> Physical switch .

3.) If i have 4 edges what is the best practice to connect these edges to physical environment and DLR. I know their roles and functions but still stuck on some designing chunks.

Regards ,

Sonika

[larsonm]

1.  You could have as few as one DLR, depending on what you're trying to accomplish.  One may speculate that 4 edges would be deployed for the purpose of to leverage ECMP. 

2.  You wouldn't connect the DLR directly to the Edge.  You'd connect the DLR uplink interface to a logical switch that would be your transport logical switch.

Internal interfaces on the edges would connect to the transport zone.

3.  See page 72 and 131-137 of the design guide for more on this matter HERE

Answers could be different if you're speaking of multi-tenancy.

[Sonikashah]

Thanks Larsonm .. However , bit confused as I am a core network person .

Lets say if i have 4 chassi with 15 host each. I have managament & edge and compute cluster 1 and compute cluster 2 on all 4 chassis.

I am putting NSX manager , controller , and all NSX related component in managemnt cluster.

Now lets talk about configuration part :

I have one dvs which is connected to all 4 chassis .

Now if i am going with one DLR on chassis 1 host 1. Then how other host will work . Is it like this that when we do host preparation then it gives DLR instances to all clusters which we prepare .

Only control vm is active standy by which is helping routing to be done .

I just to know the sizing of this ..

chassis wise just for my app db and web traffic.

4 chassis with 15 hosts each.

each chasis have 20 gb bandwidth.

4 edges will pear ospf with how many dlr ???

4 edges will pair with how many control vms.

and how. If u can show i a diagram tht would be great

[larsonm]

Rather than re-invent what has already been done, take a look the Design Guide HERE.

If you're thinking something like a single tenant with a high amount of north-south bandwidth then this is your answer - Using ECMP, 1 DLR can pair with 8 ESGs as shown in page 149

If you're thinking multiple tenants, then this is your answer - Up to 9 DLR control VMs can pair with 1 ESG, as long as they are pairing on separate edge interfaces as shown on page 71.  Using trunk interfaces, this number goes up to 200, as up to 200 sub-interfaces are supported across all available vNIC interfaces deployed as trunks.