Implementing Memcached on EC2

Erik Osterman e at
Thu Jan 4 17:25:11 UTC 2007

Vincent van Leeuwen wrote:
> On 2007-01-03 15:42:48 -0800, Erik Osterman wrote:
>> We could use tinc, which provides fullmesh routing, but we'd
>> still have the single point of failure for authorization (I believe).
> No you don't, tinc doesn't have single points of failure if properly set up. 
> Just make sure each node knows of more than one other node for the initial 
> connection. Once a first connection is made the other hosts are 
> auto-discovered.
> That said, tinc certainly isn't perfect, and I'd be hesitant to tunnel
> low-latency packets over it. Once the persistent connections between all hosts
> are in place everything runs fine, but when connections fail (for whatever
> reason, mostly network problems) and have to be reopened you have extra
> latency and packet loss at that time. We're very happy with using tinc between
> various networks, but we don't do any high-performance (be it latency or
> bandwidth) data over it.

Thanks for that useful bit of data. We'll keep that in mind, if it comes 
down to implementing it....

Erik Osterman

More information about the memcached mailing list