Best practice | Load Balancing/Sharing | MPLS |

[badr_alobaidi]

Dears

Greetings

I'm seeking your ideas and support, I need to know how can I get the benefit of load balancing to reduce the operational cost of the MPLS circuits. Attached is the topology of the MCS connection from IT to branch office. 1 Mbps for each link and the I'm planning to reduce the bandwidth for each to 512 kbps and do load sharing to be as one link 1 Mbps.

Kindly advise a smart solutions even hardware solution,,

Regards

[Richard Burts]

There is not a solution that will take a 512 connection from R1 to ISP1 and a 512 connection from R2 to ISP2 and produce a single link of 1 Mbps. And it is even more of a challenge to do that when you specify that it will be MPLS. Are you suggesting one MPLS with ISP1 and another MPLS with ISP2 or are you imagining a single MPLS through both ISPs?

I believe that the best you can do is to use some routing protocol between you and the ISPs that will direct some flows through one and some flows through the other. Be aware that trying to do per packet load share is likely to produce out of order packets and some applications are adversely affected by out of order packets.

HTH

Rick

[badr_alobaidi]

Dear Richard,,

Greetings,,

Thanks for providing a such ideas. It's 2 different MPLS (ISPs) and we need to reduce the copex by downgrade the bandwidth for each and aggregate it. You mentioned that using routing protocol among  R1 and R2. Does that mean to connect theses routers together and do routing?

For example, R1 is the primary link so the traffic come from the LAN to R1 and then divide them through ISP1 MPLS and another through the link that made between R! and R2. 

[Richard Burts]

I understand the need to reduce expense by reducing bandwidth. It makes aggregating the traffic more difficult because there are two ISP. It would be easier to achieve if there was a single ISP. If you want to aggregate the traffic in a way that uses both links then I believe that you need some routing protocol between R1 and R2. And you need some routing policy that will take traffic coming from the LAN and process it in a way that will send some through R1 and will send other through R2.

HTH

Rick

[Joseph W. Doherty]

Disclaimer

The Author of this posting offers the information contained within this posting without consideration and with the reader's understanding that there's no implied or expressed suitability or fitness for any purpose. Information provided is for informational purposes only and should not be construed as rendering professional advice of any kind. Usage of this posting's information is solely at reader's own risk.

Liability Disclaimer

In no event shall Author be liable for any damages wha2tsoever (including, without limitation, damages for loss of use, data or profit) arising out of the use or inability to use the posting's information even if Author has been advised of the possibility of such damage.

Posting

If your branch routers support it, PfR can dynamically load balance between your two links.

If not, many routers support ECMP, and in a branch topology as in your diagram, you can make one router the primary gateway and often configure an IGP to treat the other router as another equal cost path.  Or, you might try GLBP.

[e.ciollaro]

Hi, a way to have a two links that seems to be a single link could be to configure a GRE tunnel between branch and IT and configure per-packet load balancing. so that GRE packtes are balanced on links. Obviously you have an overhead due to the GRE header and out of order packet as Richard has highlighted. I think that Cisco implement GRE according to RFC 2890 so may be reordering is done bye GRE but I'm not sure (I did a capture with wireshark and I saw the extended header but Key-bit and sequence-number bit are always set to 0, i.e not used)

And some scenario it could also be useful to advertise different networks  to ISPs, this ways MPLS netwroks are used for different type of traffic. In some case one network is for data and the other for video and VoIP. Backup could be manual or based on dynamic routing protocols.

Bye,

enrico