most stable VPLS setup

-headstrong- · Tue Oct 20, 2009 3:56 pm

HI Guys

Quick question,

1) Which is more stable: LDP or BGP based VPLS's? I do not care much about scalability issues just stability....

2) Would TE tunnels make LDP based VPLS more stable?

Thanks

sewlist · Tue Oct 20, 2009 7:46 pm

W've been using LDP based VPLS for a while, very solid in my opinion

using Layer 3 MPLS as well with vrf's , its amazing networking for me, we run about 25+ clients through the layer3 with bgp vpn4-route

we mostly wireless ISP, but we have a solution for ADSL customers coming into the MPLS cloud and link them on the wireless network and its all perfect, very much like the cisco style MPLS providers

my 2 quick cents

S

-headstrong- · Tue Oct 20, 2009 8:49 pm

thanks sewlist

Please have a look at my attached network layout

Basically, sites C,E,F,G all have a VPLS tunnel to Site-A which is the PPPOE A/C
The below attached IP's are the loopback IP's
OSPF is running between the Sites

Do you have any tweaks that I could apply to the below setup?:

admin@Site-A>
/mpls
set dynamic-label-range=16-1048575 propagate-ttl=yes
/mpls interface
add comment="" disabled=no interface=all mpls-mtu=1508
/mpls ldp
set distribute-for-default-route=no enabled=yes hop-limit=255 loop-detect=no lsr-id=1.1.1.1 path-vector-limit=255 transport-address=1.1.1.1 \
use-explicit-null=no
/mpls ldp interface
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s hold-time=15s interface=ether2 transport-address=0.0.0.0
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s hold-time=15s interface=ether3 transport-address=0.0.0.0

/interface vpls
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 comment="" disable-running-check=no disabled=no l2mtu=1500 mac-address=\
02:DE:B3:CC:A7:5E mtu=1500 name=site-01 remote-peer=1.1.1.3 vpls-id=1:0
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 comment="" disable-running-check=no disabled=no l2mtu=1500 mac-address=\
02:C3:5C:7A:13:D3 mtu=1500 name=site-02 remote-peer=1.1.1.5 vpls-id=2:0
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 comment="" disable-running-check=no disabled=no l2mtu=1500 mac-address=\
02:4E:28:C6:B4:39 mtu=1500 name=site-01 remote-peer=1.1.1.6 vpls-id=3:0
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 comment="" disable-running-check=no disabled=no l2mtu=1500 mac-address=\
02:4F:EA:94:76:BB mtu=1500 name=site-02 remote-peer=1.1.1.7 vpls-id=4:0
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 comment="" disable-running-check=no disabled=yes mac-address=02:17:39:8F:B4:3A mtu=\
1500 name=site-03 remote-peer=1.1.1.4 vpls-id=5:0

/interface bridge
add admin-mac=00:00:00:00:00:00 ageing-time=5m arp=enabled auto-mac=yes comment="" disabled=no forward-delay=15s l2mtu=65535 max-message-age=20s mtu=1500 \
name=lobridge priority=0x8000 protocol-mode=none transmit-hold-count=6
add admin-mac=00:00:00:00:00:00 ageing-time=5m arp=enabled auto-mac=yes comment="" disabled=no forward-delay=15s l2mtu=1500 max-message-age=20s mtu=1500 \
name="PPPOE A/C" priority=0x8000 protocol-mode=none transmit-hold-count=6

/interface bridge port
add bridge="PPPOE A/C" comment="" disabled=no edge=auto external-fdb=auto horizon=1 interface=Site-F path-cost=10 point-to-point=auto priority=0x80
add bridge="PPPOE A/C" comment="" disabled=no edge=auto external-fdb=auto horizon=1 interface=Site-G path-cost=10 point-to-point=auto priority=0x80
add bridge="PPPOE A/C" comment="" disabled=no edge=auto external-fdb=auto horizon=1 interface=Site-C path-cost=10 point-to-point=auto priority=0x80
add bridge="PPPOE A/C" comment="" disabled=no edge=auto external-fdb=auto horizon=1 interface=Site-E path-cost=10 point-to-point=auto priority=0x80

/interface bridge settings
set use-ip-firewall=yes use-ip-firewall-for-pppoe=yes use-ip-firewall-for-vlan=no

admin@Site-C>
/mpls
set dynamic-label-range=16-1048575 propagate-ttl=yes
/mpls interface
add comment="" disabled=no interface=all mpls-mtu=1508
/mpls ldp
set distribute-for-default-route=no enabled=yes hop-limit=255 loop-detect=no \
lsr-id=1.1.1.3 path-vector-limit=255 transport-address=1.1.1.3 \
use-explicit-null=no

/mpls ldp interface
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s \
hold-time=15s interface="WIRELESS BACKBONE TO SITE-B" transport-address=\
0.0.0.0
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s \
hold-time=15s interface="ETHERNET TO SITE-E" transport-address=0.0.0.0
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s \
hold-time=15s interface="ETHERNET TO SITE-D transport-address=0.0.0.0
/mpls ldp neighbor
add comment="" disabled=no send-targeted=yes transport=1.1.1.2

/interface vpls
add advertised-l2mtu=1500 arp=enabled cisco-style=no cisco-style-id=0 \
comment="" disable-running-check=no disabled=no l2mtu=1500 mac-address=\
02:A5:23:76:FB:C7 mtu=1500 name=VPLS-TO-SITE-A remote-peer=\
1.1.1.1 vpls-id=1:0

-headstrong- · Thu Oct 22, 2009 1:39 pm

I was reading this page about MTU: http://wiki.mikrotik.com/wiki/MTU#VPLS_ingress

For VPLS, should the MPLS interface MTU be set to 1508 or 1512?

Mplsguy · Thu Oct 22, 2009 10:01 pm

I assume you want to avoid fragmentation. Then mpls-mtu of outgoing interface (the one over which VPLS encapsulated frames will get sent) should be set to: 1514(ethernet frame)+4(vpls control word)+8(2 labels)=1526. Note that all routers in MPLS cloud should be configured with the same mpls-mtu. Also remember that you should not configure mpls-mtu that would exceed l2mtu.

-headstrong- · Thu Oct 22, 2009 10:25 pm

Thanks mpls guy.

I have just noticed that I am using RB433 and some sites are joined via ether2 which has a max MTU 1522 compared to 1526 of ether1....I have 3x RB433 linked at some site, hence why ether2 is used

Would it be advantages for me to rather use RB600's? Would I see an improvement?

EDIT:
Should I set a transport address for the LDP interfaces? I currently have them as default
/mpls ldp interface
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s hold-time=15s interface=ether2 transport-address=0.0.0.0

Mplsguy · Fri Oct 23, 2009 9:53 am

Thanks mpls guy.

I have just noticed that I am using RB433 and some sites are joined via ether2 which has a max MTU 1522 compared to 1526 of ether1....I have 3x RB433 linked at some site, hence why ether2 is used

Would it be advantages for me to rather use RB600's? Would I see an improvement?

It depends. The difference will be in amount of frames forwarded across network - assuming that you will transport full size frames across VPLS tunnel, you will have 2 times more frames if you use devices that do not support big enough l2mtu (because every big frame will get fragmented). I guess you should decide based on wether your current bw requirements for network are close to its capacity and what are your future plans. If, let's say, only 20% of your available network bandwidth is used, it is fine to go with fragmentation, if you already are at the network capacity limits (especially number of frames it can forward), you should see improvement. If you have wireless links as part of your MPLS cloud, I would strongly suggest avoiding frame fragmentation. Also take into account that l2mtu of 1526 bytes will not allow to transport full size ethernet frame with VLAN tag over tunnel w/o fragmentation - if you plan on doing that, it may be good idea to use devices that support jumbo frames.

EDIT:
Should I set a transport address for the LDP interfaces? I currently have them as default
/mpls ldp interface
add accept-dynamic-neighbors=yes comment="" disabled=no hello-interval=5s hold-time=15s interface=ether2 transport-address=0.0.0.0

I don't think so - you should only set it if there is specific need for that - in most cases "global" transport address is fine. You have to set interface transport address only when neighboring routers have partial routing table and therefore can not route to "global" transport address (can only route using connected routes). Beware that 2 routers must communicate the same transport address over all links that connect them (in other words - if 2 routers are connected by 2 separate links, the same transport address must get advertised over both links) - this is also good reason to use "global" transport address - there are less chances to misconfigure something.

-headstrong- · Fri Oct 23, 2009 10:38 pm

Thanks mpls guy.

I have just noticed that I am using RB433 and some sites are joined via ether2 which has a max MTU 1522 compared to 1526 of ether1....I have 3x RB433 linked at some site, hence why ether2 is used

Would it be advantages for me to rather use RB600's? Would I see an improvement?
It depends. The difference will be in amount of frames forwarded across network - assuming that you will transport full size frames across VPLS tunnel, you will have 2 times more frames if you use devices that do not support big enough l2mtu (because every big frame will get fragmented). I guess you should decide based on wether your current bw requirements for network are close to its capacity and what are your future plans. If, let's say, only 20% of your available network bandwidth is used, it is fine to go with fragmentation, if you already are at the network capacity limits (especially number of frames it can forward), you should see improvement. If you have wireless links as part of your MPLS cloud, I would strongly suggest avoiding frame fragmentation. Also take into account that l2mtu of 1526 bytes will not allow to transport full size ethernet frame with VLAN tag over tunnel w/o fragmentation - if you plan on doing that, it may be good idea to use devices that support jumbo frames.

I have wireless and ethernet links as part of my MPLS cloud but I do not have VLAN tags, so l2mtu of 1522 should be fine right? I do run pppoe over the VPLS tho

Doe MPLS TE tunnels add more overhead than the 2*4= 8 mpls bytes?

Mplsguy · Mon Oct 26, 2009 11:03 am

I have wireless and ethernet links as part of my MPLS cloud but I do not have VLAN tags, so l2mtu of 1522 should be fine right? I do run pppoe over the VPLS tho

Doe MPLS TE tunnels add more overhead than the 2*4= 8 mpls bytes?

Like I said, you need l2mtu 1526 for VPLS forwarding 1514 byte (1500 bytes data + 14 bytes ethernet header) ethernet frames.

TE tunnel itself adds just one label (4 bytes), so it does not matter (from frame size point of view) if you forward VPLS across TE tunnel or across LDP established LSP.

-headstrong- · Mon Oct 26, 2009 11:31 am

I have wireless and ethernet links as part of my MPLS cloud but I do not have VLAN tags, so l2mtu of 1522 should be fine right? I do run pppoe over the VPLS tho

Doe MPLS TE tunnels add more overhead than the 2*4= 8 mpls bytes?
Like I said, you need l2mtu 1526 for VPLS forwarding 1514 byte (1500 bytes data + 14 bytes ethernet header) ethernet frames.
.

Thanks for the clarity, however, I am running VPLS over LDP with an l2mtu of 1522 and the VPLS tunnels still work. Does it by default drop the PPPOE MTU to 1476?

EDIT: I just did a ping test, over the pppoe over the VPLS without fragmentation, with a size of 1480 get I reply

Mplsguy · Mon Oct 26, 2009 5:35 pm

Thanks for the clarity, however, I am running VPLS over LDP with an l2mtu of 1522 and the VPLS tunnels still work. Does it by default drop the PPPOE MTU to 1476?

EDIT: I just did a ping test, over the pppoe over the VPLS without fragmentation, with a size of 1480 get I reply

I don't know to what exactly you refer as "PPPOE MTU" of 1476 bytes, therefore can not comment on your PPPoE case, but you can easily calculate resulting size of packet knowing what headers are added.

RouterOS VPLS implementation automatically fragments (and reassembles at egress point) packets that exceed mpls-mtu after VPLS encapsulation. Therefore it will work, but fragment packets - you can confirm this e.g. by watching number of outgoing packets.

-headstrong- · Mon Nov 02, 2009 12:54 am

I don't know to what exactly you refer as "PPPOE MTU" of 1476 bytes, therefore can not comment on your PPPoE case, but you can easily calculate resulting size of packet knowing what headers are added.

ie what value should I set the max MTU for pppoe clients?

/interface pppoe-client set pppoe-out1 max-mtu=1492

thanks

most stable VPLS setup

most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Re: most stable VPLS setup

Who is online