I'm not entirely certain I buy the argument that software which has lower CPU and memory utilization on average is better (on the server).
The server is sitting there with all the ram and CPU and 2% or 20% doesn't make a difference if there is nothing else to take advantage of that other 98 to 80%. The only thing that matters is the ability to handle peaks (and being reliable).
You want server software that is optimized around the worst case scenario (unless you want to run your software on servers with little resources in which case it's the other way around).
So if nginx can handle higher loads than pound, it's a better load balancer. But if it keeps my servers at 2% most of the time but does worse under high loads then it's no good.
Now I wouldn't use Pound either for load balancing and I believe nginx should be better, but the data in this article doesn't really support that argument.
Of course the evidence is far from conclusive without more data. But Pound has had terrible performance so far, and based on what data I have nginx is an order of magnitude more efficient.
i prefer a proxy balancer that is smart enough to know when the back-end servers (mongrels in this case) are busy or not, and smart enough that it doesn't send requests to back-end servers that are down (i'm looking at you, apache mod_proxy_balancer)
I've been really happy using apache2 in front of a mongrel cluster. Apache serves up static files and load balances requests to the mongrels. Can anyone give me a good reason why I should be using something other than apache on a production server?
The server is sitting there with all the ram and CPU and 2% or 20% doesn't make a difference if there is nothing else to take advantage of that other 98 to 80%. The only thing that matters is the ability to handle peaks (and being reliable).
You want server software that is optimized around the worst case scenario (unless you want to run your software on servers with little resources in which case it's the other way around).
So if nginx can handle higher loads than pound, it's a better load balancer. But if it keeps my servers at 2% most of the time but does worse under high loads then it's no good.
Now I wouldn't use Pound either for load balancing and I believe nginx should be better, but the data in this article doesn't really support that argument.