Hi,

Just curious — In production How often would we need to restart NextGen server? In other words, what’s the mean time to failure, and what’s the service level agreement? I’m not advocating any change. I understand our approach is perhaps the only option. I just want to know the expectation on NextGen server.

My perspective — My previous teams (%% mmail%%) created client-server systems in-house where a home-made daemon would run inside a Unix host for a few days before it eventually stops working, for various reasons

* request queue grow and grow and overflow
* no more output in the server log
* extremely slow response
* excessive network latency
* perhaps memory leak
* perhaps deadlocks
* perhaps some threads get stuck due to exessive synchronization and locking
* perhaps thrashing
* core dump

Such degradation symptoms are common in Windows desktops — most of us do a reboot at least once a week or so. Even commercial-grade servers can suffer the same fate. In a high-volume system, we used to restart our iplanet servers every night. Robust and resilient server design is a software industry challenge.

tan bin