BGP vs Appliance?

Personal opinion only here, but you asked:
BGP is used extensively for this purpose and works very well for dual-homing and load balancing. You have the expertise in-house, and the router IOS comes with BGP at no extra cost. Assuming that you also own your own IP subnet and have your own BGP AS number, why on earth would you want to throw in another piece of hardware that is just supposed to 'magically' load balance without using BGP? The whole purpose of having multi-homed connections and redundancy is to reduce the single points of failure. Why in heck would you put some device out front and CREATE another single point of failure?

MCI engineers like the FatPipes, so there must be something to it. If you don't have an IP address block and BGP AS number already, the cost of getting that setup vs the cost of the FatPipes is about a wash, and if you don't have the inhouse expertise in BGP, the extra consulting $$ can add up.

I'd say you already have the best situation, most cost effective, and highest ROI of any potential alternative using a 3rd party hardware device.

Thanks I agree. Do you know anything about Cisco OER(Optimizer Edge Routing)? This device is suppose to be assist with the limitation of BGP where by default BGP alone will not chose the best path based on latency, delay, jitter, etc etc. Basiclly the device creates a iBGP neighbor with your edge dual-homed router. It runs its own tests on the above mentioned criteria(and a few other besides latency, delay, etc.) detemines which is the best provider and then injects the route to your edge router along with a prefered local preference attribute. Let me know if you've heard of this new Cisco product and what your thoughts are. I think sounds like a decent idea that is basically an extension to bgp. Let me know your thoughs. Thanks again.

This is one that I have not heard of yet. I have, however, been playing around with SAA, Service Assurance Agent, to adjust routing based on availability of an upstream host (ie. if my ISP loses it's peering connection, my interface stays up, but I'm bound to only what we can reach within the same ISP) so that if I lose that connectivity, I can swing my default to an alternate ISP without using BGP. Lots of potential with this. Search CCO for "SAA routing" and you'll find several good articles.

Bottom line, there are several Cisco methods to achieve the goal without adding some unknown 3rd party device into the fray.

BTW, what IS your exact goal/mandate?

redundancy and best path.

What is your gut feeling? Unfortunately I don't know anyone who has actually put any of these devices in place to get first-hand reaction on their performance.

What WAN technology(s) are you using? T1? DS3? Multiple ISPs? How many lines do you have? Do you host publicly accessible servers (inbound IP requirements)? Where are your single points of failure now? How do you mitigate them? How much will redundant/best path WAN links give you in terms of uptime of those servers/services that must be up? What will you have to give up in your budget and do without if you go with something like the FatPipes?

There are lots of ways to spin it.