Now that we know what clustering is about let us get to the lower level details.
And specifically what I know and what I do. >From a commercial point of view if you are a networking service vendor, if you provide a mail server or spam control or firewall or a bug database or something then you can provide 100% uptime using clustering even during catastrophic failures like power outage, someone tripping the cable and so on. It works at separate layers(not OSI) just like everything in the UNIX world. First there is a protocol called CARP or VRRP which has idea of a virtual and physical IP address. The virtual IP address is assigned to multiple nodes and only one of them will be active at a time and respond at the IP layer. So this is an "active passive" clustering. Which means we scale out rather than scale up. Resources are not optimally utilized. Moreover depending upon the use case in question you certainly have to sync the data between the cluster participants. Now to implement real active active clustering in which load sharing is effected you need to some layer 7 redirecting using a 'service redirector' to share the service load between cluster nodes. And this means that the service redirector node now acts as a single point of failure and so that needs to be clustered active passive. Now you have a perfect situation in which load is shared optimally and we also ride on the CARP/VRRP active passive case which is the foundation of clustering. Now if you have been with me so far then I shall be discussing my fresh knowledge in this domain as and when I wrap up my deployment in my customer location in London. Also the service redirector node can bind to separate IP addresses and serve multiple protocols with one physical network card and bare metal. -Girish -- Gayatri Hitech http://gayatri-hitech.com _______________________________________________ ILUGC Mailing List: http://www.ae.iitm.ac.in/mailman/listinfo/ilugc ILUGC Mailing List Guidelines: http://ilugc.in/mailinglist-guidelines
