Barracuda 340s – What We Like, Dislike and Our Wish List

We have been using a pair of Barracuda Load Balancers — model 340 — for few years. Normally, we don’t tinker with them too much as they are sort of the “invisible” devices that no one remembers after setting them up. However, recently we had to make a few configuration changes and had to revisit the beasts. In this review, we will share what we like, dislike about the Barracuda 340s and perhaps what they can improve in the future.

What We Like about Barracuda Load Balancer 340s

Here is a laundry list of what we like about them:

• Cheaper than F5 Network devices – we used F5 before and they are supervly overpriced Linux boxes
• Simple to Setup – compared to F5 devices
• Pleasant Support Encounters – the people behind the pizza boxes seem to respond fast and getting to a real engineer is not too difficult
• Direct Path Route – this is what we love the most and it is often not played up by Barracuda itself!

Those of you who are unfamiliar with load balancers, here is a quick introductory lesson. All income traffic flows through the load balancers and gets distributed among your real servers behind them load balancers. The responses from the real servers are then sent back to the clients through the load balancers as well. So if you have a TON of traffic, your load balancer’s network interface (port) and the network stack will be super busy managing both incoming and outgoing traffic. That’s what you get in a vanilla load balancer and even on USD $15K+ F5 boxes that we used in the last few years.

Barracuda 340s offer a different option — direct path route. This means, that incoming traffic hits the 340s but the outgoing traffic can go out on a different interface on your real servers. So for example, say you have three real servers behind a pair of 340s. If each of the real servers belong to a routable subnet than the 340s can handle the incoming traffic and you can have a loopback device that ties the virtual IP to the other IP on the other interface.

Why bother? You can get greater throughput as 340s are busy with only incoming traffic, the outgoing traffic is served directly from the real servers using a different network port.  Just think logically — if a 340’s network port can handle N incoming and outgoing packets  per second, would be better only dealing with incoming packets and letting the real servers handling the outgoing? You got it!

Now why is Barracuda downplaying this wonderful feature when selling these pizza boxes? Here are some of our guesses:

• If you found out that you can do direct path route to increase throughput, perhaps you won’t buy their higher-end models which simply have better throughput?
• Exposing real servers with real IP address is considered bad security practice by some?
• They can only deal with one subnet and therefore it is not too practical for customers with loads of different networks?
• They just do not have enough customers using these features? We do!

What We Don’t Like about Barracuda Load Balancer 340s

Sure, we like the 340s but there are also things that we do not like about them, which are below:

• The Web interface really belongs in the 90s.  You have to go to page B to return to page A to see a status update! Hello Ajax!
• The Web interface has a low-end feel to it – if you are used to F5 Web interface, you will certainly feel that you bought something cheap!
• Sometimes we have to reboot the box to get the settings to take effect — which sucks.

What We Wish for Future Barracuda Load Balancer 340s

Overall, we will give 340s a B+ rating and would like Barracuda to think about the following in updating them:

• Have a Web interface company review your interface and make them more user-friendly
• Please write better documentation and make them context sensitive for your Web interface
• Use more higher-end visuals — use real graphs and charts for a change
• How about hooking up some fun stats?