Data Center

Hi everyone ,

I  have been mulling over Mass withdrawal feature in EVPN VXLAN   using  type 1 EAD/ES route.  

 From page# 151 Chapter 4 of  book "Deploying Juniper Data Centers with EVPN VXLAN"

 "For fast convergence and to quickly signal to other VTEPs (PEs), a VTEP that loses its connection to an Ethernet Segment simply withdraws
the associated Type-1 EAD per ES route first before individually withdrawing each of the addresses."

And I do see that behavior in my lab ( shown below where VTEP ( 3.3.33)  is withdrawing EAD/ES route first, then  individual MAC /IP route  ( 00:05:22:22:22:22) next).

What if we have 1 millions BGP routes, then  VTEP has to withdraw 1 million  MAC routes ? Not  very efficient, why withdrawing EAD/ES route is not enough  because  neighboring VTEP  will remove all MAC entries tied to ESI  upon receiving withdrawal BGP update for EAD/ES. Why do we  still have to withdraw 1 millions BGP routes ?

MAC/IP route is withdrawn next:

If that link remains down forever, then surely those Type-2 MAC/IP routes (1M+) should be removed from the network. The network should not keep T2 routes for hosts that might never return. 

Hi everyone ,

I  have been mulling over Mass withdrawal feature in EVPN VXLAN   using  type 1 EAD/ES route.  

 From page# 151 Chapter 4 of  book "Deploying Juniper Data Centers with EVPN VXLAN"

 "For fast convergence and to quickly signal to other VTEPs (PEs), a VTEP that loses its connection to an Ethernet Segment simply withdraws
the associated Type-1 EAD per ES route first before individually withdrawing each of the addresses."

And I do see that behavior in my lab ( shown below where VTEP ( 3.3.33)  is withdrawing EAD/ES route first, then  individual MAC /IP route  ( 00:05:22:22:22:22) next).

What if we have 1 millions BGP routes, then  VTEP has to withdraw 1 million  MAC routes ? Not  very efficient, why withdrawing EAD/ES route is not enough  because  neighboring VTEP  will remove all MAC entries tied to ESI  upon receiving withdrawal BGP update for EAD/ES. Why do we  still have to withdraw 1 millions BGP routes ?

MAC/IP route is withdrawn next:

Thanks , but why do need to withdraw each BGP route individually?  We could have implemented simple logic:  If a vtep  receives  EAD/ESI  BGP withdraw update, then receiving VTEP must flush all BGP routes tied to the ESI listed in EAD/ESI BGP withdraw update.

If that link remains down forever, then surely those Type-2 MAC/IP routes (1M+) should be removed from the network. The network should not keep T2 routes for hosts that might never return. 

Hi everyone ,

I  have been mulling over Mass withdrawal feature in EVPN VXLAN   using  type 1 EAD/ES route.  

 From page# 151 Chapter 4 of  book "Deploying Juniper Data Centers with EVPN VXLAN"

 "For fast convergence and to quickly signal to other VTEPs (PEs), a VTEP that loses its connection to an Ethernet Segment simply withdraws
the associated Type-1 EAD per ES route first before individually withdrawing each of the addresses."

And I do see that behavior in my lab ( shown below where VTEP ( 3.3.33)  is withdrawing EAD/ES route first, then  individual MAC /IP route  ( 00:05:22:22:22:22) next).

What if we have 1 millions BGP routes, then  VTEP has to withdraw 1 million  MAC routes ? Not  very efficient, why withdrawing EAD/ES route is not enough  because  neighboring VTEP  will remove all MAC entries tied to ESI  upon receiving withdrawal BGP update for EAD/ES. Why do we  still have to withdraw 1 millions BGP routes ?

MAC/IP route is withdrawn next: