On 08/06/2011 02:21 AM, Dilshod wrote:
I was doing research on how to keep cache demons in sync across multiple
machines (5-10). In this environment any machine could fall off the
scene and reappear, so address auto-discovery is essential.
Keeping a cluster in sync is a theoretically hard
Hi All,
I was doing research on how to keep cache demons in sync across multiple
machines (5-10). In this environment any machine could fall off the scene
and reappear, so address auto-discovery is essential.
What's a good way to approach the problem?
I have read Brokerless
You could make a distributed discovery service that all daemons know
about (use DNS and a common port). When clients come up, they register
with the service. the discovery service would need to know about the
other discovery service nodes to be able to replicate the data.
Joshua
On 8/5/2011
Hi Joshua, thanks for the reply.
So what's the advantage just having pub/sub on that service daemon that
would simply rebroadcast messages? I was hoping of not having this middle
piece.
Another option I was considering is to have each node to be publisher
and subscriber, and use zpolling to
On 6 August 2011 08:21, Dilshod dils...@gmail.com wrote:
Also I looked into epgm protocol which allows multicasting but Linode
(VPS) host does not support it from I could gather online.
Pretty much all data centres and hosting blocks IPv4 multicast, maybe this
changes with IPv6. I have an