> From: Aaron Grant > Sent: Friday, March 14, 2014 6:03 AM > > We were considering replicating our CAS tickets using ehCache and were > wondering how many other folks did this and did you see a considerable > jump in your memory or cpu usage on your CAS servers?
We originally deployed a two node cluster using ehCache; it mostly works but I wasn't particularly happy with it. If you are using unicast, as opposed to multicast, there is an as yet unfixed bug that severely impacts scalability, it reduces performance in a two node cluster and renders a three node cluster simply intractable. I recently tried out Unicon's relatively new hazelcast ticketRegistry, it's a lot simpler to configure compared to ehCache and works a lot better. https://github.com/Unicon/cas-addons/wiki/Configuring-HazelcastTicketRegistry I'm planning to swap out our existing two node ehCache cluster in a couple of weeks with a three node hazelcast cluster. I don't like to pass credentials in plaintext over the wire, as ehCache didn't support any type of encryption, I ended up tunneling its traffic with ssh. Hazelcast supports two different modes of encryption, SSL and symmetric, but unfortunately neither of them seem reliable under load. For my new cluster, I ended up setting up ipsec tunnels between the nodes and leaving the encryption to the OS rather than the replication library, that turned out to work very well. If you are just starting, personally I would recommend trying out hazelcast before ehCache. -- You are currently subscribed to [email protected] as: [email protected] To unsubscribe, change settings or access archives, see http://www.ja-sig.org/wiki/display/JSG/cas-user
