Thanks Colin, We’ve already taken packet traces of both, and sent to IBM. They see the send window being reduced automatically by TCPIP because the remote end cannot keep up. the problem only surfaces when doing EBCDIC and BLOCK transfers.
Dave Jousma Vice President | Director, Technology Engineering From: IBM Mainframe Discussion List <[email protected]> on behalf of Colin Paice <[email protected]> Date: Thursday, March 28, 2024 at 8:45 AM To: [email protected] <[email protected]> Subject: Re: Slow FTP's Dave, I've been working on a TCP session monitor. It reports MB/Second + changes to the TCP parameters, such as change in window size, buffer size etc. all of which affect throughput. I can let you have a copy of the program if you are interested. Dave, I've been working on a TCP session monitor. It reports MB/Second + changes to the TCP parameters, such as change in window size, buffer size etc. all of which affect throughput. I can let you have a copy of the program if you are interested. [email protected] it reports //SYSPRINT Using --PORT 22 --SLEEP 10 --COUNT 10 --TCPIP TCPIP HH:MM:SS,remote ,BytesI,BytesO,,SegsI , SegsO,,BI/Sec,BO/Sec,,ReXmtC, 08:33:23,10.1.0.2:41984 , 0, 0,, 0, 0,, 0, 0,, 0, 08:33:23,10.1.0.2:55554 , 676, 45832,, 54, 63,, 67, 4583,, 0, 08:33:33,10.1.0.2:41984 , 0, 0,, 0, 0,, 0, 0,, 0, 08:33:33,10.1.0.2:55554 , 52, 27452,, 15, 28,, 5, 2745,, 0, 08:33:43,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, Freeing entry 10.1.0.2:41984 08:33:53,10.1.0.2:37728 , 156, 1164,, 14, 11,, 15, 116,, 0, 08:33:53,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:04,10.1.0.2:37728 , 416, 71932,, 50, 77,, 37, 6539,, 0, 08:34:04,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:14,10.1.0.2:37728 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:14,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:24,10.1.0.2:37728 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:24,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:34,10.1.0.2:37728 , 0, 0,, 0, 0,, 0, 0,, 0, 08:34:34,10.1.0.2:55554 , 0, 0,, 0, 0,, 0, 0,, 0, The first time the program sees a connection it dumps out 10.1.0.2:41984 JOBNAME SSHD4 Interface TAP0 10.1.0.2:41984 ResourceId: 45 InSegs: 37 OutSegs: 31 SSThresh: 65535 10.1.0.2:41984 OutBuffered: 0 InBuffered: 0 10.1.0.2:41984 ReXmtCount: 0 CongestionWnd: 41760 RoundTripTime: 4 RoundTripVar: 11 10.1.0.2:41984 SndBufSize: 65536 SndWnd: 64128 MaxSndWnd: 64256 SendMSS: 1452 10.1.0.2:41984 RcvBufSize: 65536 RcvWnd:131020 Lcl0WindowCount: 0 Rmt0WindowCount: 0 10.1.0.2:41984 If TCPIP parameters change, such as window size you get in //CHANGE 08:33:23 10.1.0.2:55554 NWMConnRoundTripTime n:2 - o:4 = -2 08:33:23 10.1.0.2:55554 NWMConnRoundTripVar n:2 - o:7 = -5 08:33:33 10.1.0.2:55554 NWMConnRoundTripTime n:3 - o:2 = 1 08:33:33 10.1.0.2:55554 NWMConnRoundTripVar n:3 - o:2 = 1 08:33:53 10.1.0.2:37728 NWMConnCongestionWnd n:36000 - o:20160 = 15840 08:33:53 10.1.0.2:37728 NWMConnRoundTripTime n:6 - o:10 = -4 08:33:53 10.1.0.2:37728 NWMConnRoundTripVar n:17 - o:29 = -12 08:33:53 10.1.0.2:37728 NWMConnRcvWnd n:131020 - o:130176 = 844 On Thu, 28 Mar 2024 at 12:31, Jousma, David < [email protected]> wrote: > All, > > Grasping at straws here, IBM support center is baffled too. > > To clone z/OS maintenance to various disconnected sysplex’s I do a DFDSS > dump, and FTP it everywhere it needs to be. Its roughly a 50Gb file > transfer. There is some environmental issue causing slow file transfers to > some systems (40mb’s a sec) and fast file transfers (150Mb/sec) to other > systems on the same CEC. With IBM support help, we’ve narrowed down the > problem to the specification of MODE B and EBCDIC on the transfer since it > is a DSS dump. Remove those, and the transfer is fast on the slow > systems, and still fast on the fast systems. Obviously that isn’t a > solution though. > > So, we are a GDPS shop. The oddity is that all the “fast” transfers are > to the K systems(control systems), and all the “slow” transfers are to the > traditional application systems. TEST, DEV, PROD makes no difference, nor > does LPAR busy or not busy. > > It seems there is something configured differently on the “slow” systems > that is affecting mode b, ebcdic file transfers, but for the life of me, I > cannot put my finger on what, nor can the support center, except that the > issue is at the remote end, in that the OS cannot offload the data fast > enough, so TCPIP/FTP is slowing the transfer pace. > > A virtual adult beverage of choice to the one that can point in a > direction to look…. > > > Dave Jousma > Vice President | Director, Technology Engineering > > > > > > This e-mail transmission contains information that is confidential and may > be privileged. It is intended only for the addressee(s) named above. If > you receive this e-mail in error, please do not read, copy or disseminate > it in any manner. If you are not the intended recipient, any disclosure, > copying, distribution or use of the contents of this information is > prohibited. Please reply to the message immediately by informing the sender > that the message was misdirected. After replying, please erase it from your > computer system. Your assistance in correcting this error is appreciated. > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN This e-mail transmission contains information that is confidential and may be privileged. It is intended only for the addressee(s) named above. If you receive this e-mail in error, please do not read, copy or disseminate it in any manner. If you are not the intended recipient, any disclosure, copying, distribution or use of the contents of this information is prohibited. Please reply to the message immediately by informing the sender that the message was misdirected. After replying, please erase it from your computer system. Your assistance in correcting this error is appreciated. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
