Using MSTP for Load Balancing

[Darius Gražulis]

Hello,

I hope someone can help me.

I have described the situation graphically in the attachment.

The situation is like that: Users from VLAN30 provide a lot of traffic (and we are planning to have more of these users) so I was thinking of connecting K03 and K10 to K01 with additional 1G sfp links (red lines). I would like to put the VLAN30 traffic to these links (the red link between K03 and K10 is for redundancy).

How do I configure the switches to keep the VLAN30 in new links and VLAN10,20 in old links.

right now we are using RSTP. So there is only one spanning tree with the root K01. Is it possible to use MSTP to get several spanning trees but to the same root?

--------Example---------
K10 should use its own link to K01 for VLAN30 and the link to K11 for other VLAN's.
K03, K05 and K04 should use the link between K03 and K01 for VLAN30 and other links for VLAN's 10 and 20.

if any of new links from K03 and K10 to K01 would fail then I want VLAN30 to go through the new link between K03 and K10.

All VLAN's from K05 should travel through the same link to K03. the same
-----------------



Best regards,
Darius.

[Pieter 't Hart]

Hi Darius,
You've got a complex diagram.
Is this based on physical locations?
and lengt of fiber connections?

I would suggest letting the edge switches be real edge switches.
=> connect each edge switch directly to the core. add cabling accordingly.

As you've got only a single core, and all vlan30 traffic go to server2, your ideas about redundancy on the edge seems a little overdone.
=> Remove the redundant conenctions, this will make your life much simpler.
Better connect two links from each edge to the core as a port-trunk.
You may use two different physical paths to add to your "redundancy".

[Antonio Milanese]

Hello Darius,

if I understand correctly you want to balance multiple active links to your core switch and segregate heavy traffic of VLAN 30; traffic engineering,or sort of,using MSTP is possible at least until you have multiple VLAN /multiple paths scenarios:
with MSTP,simplifying,you can build multiple logical topologies (instances)in which you have separated but concurrent virtual RSTP topologies, so altering path cost and path priority in a different MSTP instance you can "engineering" how your traffic is forwarded:

MSTP instance 1 mapping VLAN30:
link1:VLAN30 = forwarding
link2:VLAN20 = blocked

MSTP instance 2 mapping VLAN 20
link1:VLAN30 = blocked
link2:VLAN20 = forwarding

To recap you need to:

1) create a common MSTP config/revision for all switches
2) define 2 different instances and distribute/map VLANs on those instances
3) use stp path cost or priorities to loadbance links
4)use QoS to prioritize traffic of VLAN 20 (or IP source/dest) in the event of one path failure

Anyway as Pieter noted you have a SPOF in your core because loadbancing does not equate automatically to redundancy =)

Regards,
Antonio

[Darius Gražulis]

Hi Pieter,

Thank you for your feedback, but it does not solve my problem regarding load balancing and MSTP. So if you have any relevant experience I would be happy to hear that :).

The length of fiber is anywhere between 120m and 160 m (the buildnig is 500m long). Actually this is approx. half of the whole diagram; only the part which describes my sittuation best.

I would also like to have several Core switches with premium licences and redundant links between them and edge switches. Unfortunatelly this is not finantially acceptable for my management... I hope some day...

We have made the decision to make redundant connections because of harsh industrial environment and big distances. There is some possibility of interruption of the link.



Best regards,

Darius.

[Darius Gražulis]

Hi Antonio,

Thank you for your feedback. :)

You have explained it very well. I have read about MSTP in the Adwanced Traffic Guide and thought that this is possible. So basically Ihave to:

-configure all VLANs on all switches (tag them on all inter-switch ports).
-create two different MSTP instances and map VLANs to them: 1 instance VLAN10 and VLAN20, 2 instance VLAN30.
-Q1: can I set up the path cost's for each MSTP instance individually?
-Q2: Can I set up the root switch for yach MSTP instance individually?

You have mentioned QoS. Is this somehow linked to MSTP or this is configured outside MSTP?

And just to check if I understand correctly:

I can set the higher priority to VLAN10 and VLAN20. this will prevent VLAN30 from consumming all the bandwidth of the inter-switch link in case 4 of 5 links to the core get disconnected? and the MSTP instances won't interfere with that?


One thing that was not clear to me is whether both RSTP' can point to the same root? because in the ATG all the example MSTP instances have different switches as root.

Can you (or anybody else who would like to contribute) please comment my questions above.

Best regards,
Darius.

[Antonio Milanese · Accepted solution]

Hello Darius,

-Q1: can I set up the path cost's for each MSTP instance individually?
yes similarly to "span port_n path-cost" "span port_n priority" which govern IST/CST (and inter-region) topology calc there is a MSTP per-instance equivalent "span instance inst_n port_n path-cost" "span instance inst_n port_n priority"

-Q2: Can I set up the root switch for yach MSTP instance individually?
yes you NEED to setup per-instance priority AND IST/legacy priority even if you have one core:
you should treat each instance switch config as a "virtual rstp bridge" with its own topology but with a common logical (and maybe physical) topology view throught CIST..don't be fooled by global bridge root topology with path cost or priority.. MSTP use a multi hierarchy with multiple physical-to-logical layers: if you have read on documentation this is a MULTIPLIER and serve as addendum to calc regional CIST root,legacy CST,boundary ports,ecc

i suggest you to read this excellent articles:

http://blog.internetworkexpert.com/2010/02/22/understanding-mstp/
http://blog.ine.com/2008/07/27/mstp-tutorial-part-i-inside-a-region/

>You have mentioned QoS. Is this somehow linked to MSTP or this is configured outside MSTP?
outside..i.e. normal QoS: this is to prevent heavy traffic sources from stealing all bandw to "high value" resources when paths coalesce due to a failure

>One thing that was not clear to me is whether both RSTP' can point to the same root? because in the ATG all the example MSTP instances have different switches as root.
see Q2..per port path-cost/priority are different, even if correlated, to root bridge election..yes it's advisable explicitally set how are computed even if a per-instance root will be a accesss/aggregate switch

Best Regards,
Antonio