JOHNC%CAD2.DECnet@GE-CRD.ARPA.UUCP (03/04/87)
>From: nagy%bsndbg.hepnet@lbl.arpa >When the HSC70 was announced, Digital published some >performance information indicating when one should make the >move from an HSC50 to an HSC70. The result of staring at >the graph was that an HSC50 has lots and lots of throughput. >If you only have a few 78x systems on your cluster, your >HSC50s are perfectly capable of handling the load (i.e., one >HSC50 can handle all your tape/disk I/O needs with no >performance bottleneck). I've seen those graphs and charts, and did a bit of studying of our cluster disk I/O rates. We have 19 RA81's dual pathed between a pair of HSC50s. Our I/O rates are _way_ below the levels DEC says make an HSC busy, but I can very decidely FEEL the difference in performance on the cluster with the disks split so that about half the IOs go thru one HSC and half thru the other. Yes, we push buttons at boot and any time an HSC reboots. Fun. :-) ------------------------------------------------------------------------- "This isn't right. This isn't even wrong!" John Child - Wolfgang Pauli General Electric Aircraft Engines