It's old news, but the NIC"s on MP3000's were advertized as "limited", not for production load. IIRC, I think "maybe 7-9 Mbit" even though they were physically 100Mbit. Weren't they emulated by the OS2 code? The OS2 emulation would probably bottleneck bandwith before TCPIP CPU was an issue. The team that sized our R25 replacement MP3000 said we should NOT use those NICS.
Our metric was that TCP/IP was using as much CPU as our medium CICS regions fed thru TN3270. Look at the paths thru the dispatcher saved in offloading. Offload trafic is LU, enters VTAM, exits VTAM. Onboard TN3270 adds trips thru the TCPIP stack. Our peak loads were driven by 3270 traffic, when load increased, TCPIP, VTAM, SuperSession, CICS (x5 prod regions) and IMS ALL increased at the same time. We were marginal but OK until we had to SSL encrypt TN3270 traffic (No crypto co-pro). With the CISCO CIP, there are Cisco content switches "beyond" the CIP that are doing encryption. THey were already there for PeopleSoft and other web traffic, so no cost to us to use. The CIP can encrytp, but Cisco FUD says it's limited as well. I suspect it would limit out faster for web SSL, but 3270 transactions are usually less than 2k bytes. > Interesting. Not long ago, we also had an MP3000. No OSA, but it did > have three NIC's that worked just like OSA's. > > I don't recall the TN3270 load being noticeable, let alone significant. > We did a boatload of other IP traffic, but, again, no noticeable load. > > How metric did you use that suggested there was a problem? > --- [This E-mail scanned for viruses by Declude Virus] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
