ProCurve / ProVision-Based
cancel
Showing results for 
Search instead for 
Did you mean: 

Aggregation/Failover between two bridged networks

John_Chapman
Occasional Visitor

Aggregation/Failover between two bridged networks

We have a site with a pair of 5412zl L2/L3 ProCurves - one located in each building.

 

The two buildings are linked with two different wireless bridge paths.

 

One is a microwave link that currently offers roughly 125Mbps symmetric and will be upgraded to 250Mbps symmetric  with a license update.

The second is an 802.11 WiFi link that nominally operates at about 125Mbps symmetric, although typical bandwidth is closer to 80Mbps. We will be moving these higher on pole mounts which may improve line of site and eliminate some of the foliage that is reducing signal strength.

 

We would like to combine these links to provide load sharing aggregation as well as failover in case one of the link paths fails (or possibly starts underperforming).

 

In each case, there is a 1000SX Fiber SFP running from the 5412xl; from that fiber link, we hit either a 1000SX/1000T transceiver or another SFP module directly into a radio device. The WiFi and Microwave radios are acting as directional point-to-point bridges.

 

We have a mix of traffic we need to feed over these links including:

1) Critical priority: Latency-sensitive VoIP SIP trunk between two ShoreTel phone systems

2) Critical priority: TN3270 and Telnet traffic to critical manufacturing systems

3) High priority: Lotus Notes/Domino connections to server for email and Notes DBs

4) High priority: AD authentication and printing

5) Medium priority: SMB File transfers

6) Medium priority (but very high bandwidth, low latency): Closed-circuit security camera monitoring (primarily tunneled over HTTP video feeds) between about 100 cameras, 1-2 DVRs and a handful of monitoring workstations that display at least a handful of rotating feeds from the DVRs and may need to pull in detail from a given set of cameras.

7) Medium priority; moderate latency: Internet and WAN access from Bldg 1 via Bldg3's Internet and WAN connections

 

We need to define QoS/traffic policies to guarantee #1-4, provide medium priority and bandwidth constraints on #5-6 (especially during congestion or a link outage such as one of the radios failing)  and a fair/equitable use for #7.

 

What is the best strategy (or strategies and trade-offs) for handling flow of the traffic over the links? Is there any way to prioritize one type of traffic (e.g. file transfers, AD, etc.) over the slower WIFi link and the other (e.g. camera traffic) over the low-latency/fast Microwave link?

 

Is there a way to have two bandwidth caps for the camera video - one when both links/paths are operational and another when a path is down?

 

Thanks,

 

John

3 REPLIES
Vince-Whirlwind
Honored Contributor

Re: Aggregation/Failover between two bridged networks

The first thing I would be doing is determining the bandwidth requirements of:

1/ VoIP traffic 

2/ Manufacturing traffic

3/ CCTV traffic

 

You will then know how much bandwidth you have left to play with.

 

You can't really have (variable) application traffic and file transfers prioritised above your CCTV feeds, as those feeds are useless if they start dropping packets due to congestion and you might as well not have them at all. It would be a bit naff to have the CCTV feeds dropping out every time somebody sends a print job, for example.

And you can't really "guarantee" your Notes and Printing traffic as they could use any amount of bandwidth.

 

As your bandwidth is limited, you will be wanting to use both your links. 

As your links are different, you can't aggregate them. (Although you could try - would save you a lot of pain if it works, but you'd lose access to a lot of your bandwidth on the faster link).

As you also want to use your links as redundant for each other, you will need to be using Policy-Based-Routing.

(Read "Route Policy" in the 5400 Routing manual).

 

So two steps:

1. prioritise your traffic on ingress to the network according to a scheme, eg:

VoIP=5

Manuf=4

CCTV=3

App traffic=2

File transfers=1

Internet=0

2. setup your two links as distinct Layer3 hops and setup PBR to feed the traffic at your links depending on source/destination in order to share the predictable traffic as sensibly as possible, and to spread the unpredictable traffic across both links.

 

You can track interfaces to change VRRP priorities, but I don't think you can do it for PBR on these switches.

 

 

 

John_Chapman
Occasional Visitor

Re: Aggregation/Failover between two bridged networks

The site manager has asked whether we can use simple link aggregation to bind the two switches over two different (radio) bridged networks.

 

I'm assuming the LACP/802.11ad does not work across  bridged networks (especially not ones with different bandwidth/latency characteristics)?

 

I know he wants to keep this as brain-dead simple and bulletproof as possible.

 

Thanks,

 

John

Vince-Whirlwind
Honored Contributor

Re: Aggregation/Failover between two bridged networks

As you are using Procurve switches, if you configure your two interfaces as "Trunks" but not as "LACP Trunks", then that should give you the effect of distributing your traffic across the two links.

 

Just make sure you disable one of the links (by disabling one of the switchports on the 5412) then configure both interfaces at both ends as Trunks, then re-enable the switchport so the second link comes online.

Then just observe the traffic stats on the interfaces and see if all four are seeing Tx packets.