[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On 2015-03-02 14:50, Matt Stevenson wrote:
I'm interested in anyones experience of transitioning from dedicated Servers to Linux Virtual Servers what Load Balance either at the switch or Application level?Most of what I've done has load balanced using multiple reverse proxy caches, these days it makes more sense to push that layer to a CDN, who'll keep 99% of the rubbish out of your way for less than it will cost you to provision one virtual squid box with decent bandwidth, and provide bandwidth cheaper than you can buy it elsewhere. They'll also be faster.
You can also provision fairly meaty virtual machines with the usual virtual machine providers, and these providers also provide high performance database engines for common purposes.
So you have to get pretty big to even need to think of load balancing, unless it is purely for availability or manageability. Thinks thousands of concurrent users on most web applications, which puts you quite high up the chain in terms of user base.
As such I've mostly managed to avoid virtual load balancing, but at a push I'd probably use the virtual server providers tools, because they'll fail over between bits of their cloud better than anything I'm likely to set-up.
Switches don't really exist at this level, so nearly all the load balancing is layer 7, based on proxies and cookies or source address. These proxies are often application agnostic, at the pain of logging in again if the specific host your session is with dies mid session. The only application level changes I've seen are to put session information into shared storage, to avoid that repeated login. See also what the CDN providers offer.
-- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq