Your Opinion Please...

I'd like some other DBA's opinions on this one. When the 7.1x (5.x doesn't do this...) engine transmits a single large BLOB (300KBytes +) over a network to a client, in between writes to the network it spins waiting for the network to become ready for more data again, rather than entering a wait state. Informix claims this improves network throughput. This is all well & good if the network is fast, like Ethernet or faster, the CPU stays pegged, but for very short periods because the network is hauling the data away pretty fast. I can live with that.... However, if the network is a PPP phone line connection, theoretically 350+ times slower, the CPU stays pegged at 100% for the duration of the transfer, which can be hours in some cases. Probably 99% of that CPU is unproductive spinning. Fortunately, when other real work comes along, the CPU is relinquished pretty efficiently. I think this is a substantial operational problem, because my CPU's stay at 100% all the time, even though I know that most of it is totally unproductive for now at least. Performance Monitoring and Capacity Planning are essentially impossible, because I can't distinguish between real and unproductive work. I might as well just turn sar off, for all the good it does now. Informix just doesn't seem to think that this is much of a problem. Does anyone else think this is a problem, or am I just clueless?? Would you find this acceptable in your environment? Call 'em like you see 'em. Greg