Very poor performance on Cat 6000 gigabit? [7:57695]
Here's something annoying that I came across yesterday... any clues as to what's going wrong would be very much appreciated. Scenario: HP NetServers with built-in 100M NICs, based on an Intel chipset. With the HP drivers, the performance is fine - as you'd expect from a 100M connection. With Intel drivers, nothing changes. Still fine. Add a 1G NIC, again HP badged but with an Intel chipset (Intel Pro/1000TX), and bind them together into a fault-tolerant set using the Intel drivers that were priovided by HP (they don't provide HP badged drivers for this card, though they are happy to sell it with an HP sticker on it for twice the cost of the "Intel" card). My intention of course is that the 1G adapter is the primary (and set so in the "teamed adapter" settings) and the 100M would only be used as a fallback if the 1G fails. That's where things go wrong. With both cards connected to the same switch (long-term intention of course is that the 100M card will connect to a standby switch) it insists on using the 100M card, even when the 1G is set as the "preferred primary" and the 100M is the "preferred secondary". Both cards definitely work... if I unplug the connection to the 100M, the 1G takes over. With only the 100M connected, it works. Now, here's the very odd bit. You'd expect better performance from the 1G card. But no. Testing with file copies to or from another server that has been working fine with a 1G card for a year or so (attached via fiber to a GBIC on the supervisor card on the switch), I get several times times better performance with the 100M NIC than I do with the 1G (both UTP). I've tried different cables. All are BICC GigaPlus. The 100M connection goes through a patch panel, but I've run a 20M flylead direct from the server to the switch for the 1G connection. The switch is a Cisco Catalyst 6000 with the 100M connections going to 48-port 100M cards, and the 1G connections going to a 16-port 1G card. Software, firmware, etc versions pasted below. Seeing much worse performance from Gigabit adapters compared to 100M is something of a disappointment, to say the least. Any ideas? The hardware and versions: WS-C6006 Software, Version NmpSW: 7.2(2) Copyright (c) 1995-2002 by Cisco Systems NMP S/W compiled on Jun 3 2002, 18:30:10 System Bootstrap Version: 5.3(1) System Web Interface Version: Engine Version: 5.3.4 ADP Device: Cat6000 ADP Ver0 Hardware Version: 1.0 Model: WS-C6006 Serial #: XXX PS1 Module: WS-CAC-1300WSerial #: XXX PS2 Module: WS-CAC-1300WSerial #: XXX Mod Port Model Serial #Versions --- --- --- -- 1 2WS-X6K-SUP1A-2GEXXX Hw : 3.1 Fw : 5.3(1) Fw1: 5.1(1)CSX Sw : 7.2(2) Sw1: 7.2(2) WS-F6K-PFC XXX Hw : 1.0 3 8WS-X6408-GBIC XXX Hw : 2.1 Fw : 4.2(0.24)VAI78 Sw : 7.2(2) 4 48 WS-X6248-RJ-45 XXX Hw : 1.1 Fw : 4.2(0.24)VAI78 Sw : 7.2(2) 5 48 WS-X6248-RJ-45 XXX Hw : 1.4 Fw : 5.4(2) Sw : 7.2(2) 6 16 WS-X6316-GE-TX XXX Hw : 1.3 Fw : 5.4(2) Sw : 7.2(2) 15 1WS-F6K-MSFC XXX Hw : 1.3 Fw : 12.0(7)XE1, Sw : 12.0(7)XE1, [GroupStudy.com removed an attachment of type application/ms-tnef which had a name of winmail.dat] Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=57695&t=57695 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
RE: Very poor performance on Cat 6000 gigabit? [7:57695]
If you get this to work keep me/us informed as I am sure you will. Because I could never get this to work, I actually had to buy another 1gig nic and still the drivers did not work correctly actually eneded up just using fast etherchannel which is working great. -Original Message- From: Martin Reilly [mailto:[EMAIL PROTECTED]] Sent: Tuesday, November 19, 2002 6:45 AM To: [EMAIL PROTECTED] Subject: Very poor performance on Cat 6000 gigabit? [7:57695] Here's something annoying that I came across yesterday... any clues as to what's going wrong would be very much appreciated. Scenario: HP NetServers with built-in 100M NICs, based on an Intel chipset. With the HP drivers, the performance is fine - as you'd expect from a 100M connection. With Intel drivers, nothing changes. Still fine. Add a 1G NIC, again HP badged but with an Intel chipset (Intel Pro/1000TX), and bind them together into a fault-tolerant set using the Intel drivers that were priovided by HP (they don't provide HP badged drivers for this card, though they are happy to sell it with an HP sticker on it for twice the cost of the "Intel" card). My intention of course is that the 1G adapter is the primary (and set so in the "teamed adapter" settings) and the 100M would only be used as a fallback if the 1G fails. That's where things go wrong. With both cards connected to the same switch (long-term intention of course is that the 100M card will connect to a standby switch) it insists on using the 100M card, even when the 1G is set as the "preferred primary" and the 100M is the "preferred secondary". Both cards definitely work... if I unplug the connection to the 100M, the 1G takes over. With only the 100M connected, it works. Now, here's the very odd bit. You'd expect better performance from the 1G card. But no. Testing with file copies to or from another server that has been working fine with a 1G card for a year or so (attached via fiber to a GBIC on the supervisor card on the switch), I get several times times better performance with the 100M NIC than I do with the 1G (both UTP). I've tried different cables. All are BICC GigaPlus. The 100M connection goes through a patch panel, but I've run a 20M flylead direct from the server to the switch for the 1G connection. The switch is a Cisco Catalyst 6000 with the 100M connections going to 48-port 100M cards, and the 1G connections going to a 16-port 1G card. Software, firmware, etc versions pasted below. Seeing much worse performance from Gigabit adapters compared to 100M is something of a disappointment, to say the least. Any ideas? The hardware and versions: WS-C6006 Software, Version NmpSW: 7.2(2) Copyright (c) 1995-2002 by Cisco Systems NMP S/W compiled on Jun 3 2002, 18:30:10 System Bootstrap Version: 5.3(1) System Web Interface Version: Engine Version: 5.3.4 ADP Device: Cat6000 ADP Ver0 Hardware Version: 1.0 Model: WS-C6006 Serial #: XXX PS1 Module: WS-CAC-1300WSerial #: XXX PS2 Module: WS-CAC-1300WSerial #: XXX Mod Port Model Serial #Versions --- --- --- -- 1 2WS-X6K-SUP1A-2GEXXX Hw : 3.1 Fw : 5.3(1) Fw1: 5.1(1)CSX Sw : 7.2(2) Sw1: 7.2(2) WS-F6K-PFC XXX Hw : 1.0 3 8WS-X6408-GBIC XXX Hw : 2.1 Fw : 4.2(0.24)VAI78 Sw : 7.2(2) 4 48 WS-X6248-RJ-45 XXX Hw : 1.1 Fw : 4.2(0.24)VAI78 Sw : 7.2(2) 5 48 WS-X6248-RJ-45 XXX Hw : 1.4 Fw : 5.4(2) Sw : 7.2(2) 6 16 WS-X6316-GE-TX XXX Hw : 1.3 Fw : 5.4(2) Sw : 7.2(2) 15 1WS-F6K-MSFC XXX Hw : 1.3 Fw : 12.0(7)XE1, Sw : 12.0(7)XE1, [GroupStudy.com removed an attachment of type application/ms-tnef which had a name of winmail.dat] Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=57700&t=57695 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
RE: Very poor performance on Cat 6000 gigabit? [7:57695]
Have you tried the URL "NIC Issues..." on Cisco pages? There are a list of problems related with Intel, Compaq and so on. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=57833&t=57695 -- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
Re: Very poor performance on Cat 6000 gigabit? [7:57695]
hi, as far as i was aware you CAN`T team to different speed network cards we use the intel/Compaq/HP (the same cards/drivers) and i have not been able to get the teaming to work with 100/1000 . if you put 2 1g`s togther ...no problem2 100`s ...again no problem but different speed`s NOPE.. HTH steve - Original Message - From: "Elijah Savage III" To: Sent: Tuesday, November 19, 2002 2:31 PM Subject: RE: Very poor performance on Cat 6000 gigabit? [7:57695] > If you get this to work keep me/us informed as I am sure you will. > Because I could never get this to work, I actually had to buy another > 1gig nic and still the drivers did not work correctly actually eneded up > just using fast etherchannel which is working great. > > -Original Message- > From: Martin Reilly [mailto:[EMAIL PROTECTED]] > Sent: Tuesday, November 19, 2002 6:45 AM > To: [EMAIL PROTECTED] > Subject: Very poor performance on Cat 6000 gigabit? [7:57695] > > > Here's something annoying that I came across yesterday... any clues as > to what's going wrong would be very much appreciated. > > Scenario: > > HP NetServers with built-in 100M NICs, based on an Intel chipset. > > With the HP drivers, the performance is fine - as you'd expect from a > 100M connection. With Intel drivers, nothing changes. Still fine. > > Add a 1G NIC, again HP badged but with an Intel chipset (Intel > Pro/1000TX), and bind them together into a fault-tolerant set using the > Intel drivers that were priovided by HP (they don't provide HP badged > drivers for this card, though they are happy to sell it with an HP > sticker on it for twice the cost of the "Intel" card). My intention of > course is that the 1G adapter is the primary (and set so in the "teamed > adapter" settings) and the 100M would only be used as a fallback if the > 1G fails. > > That's where things go wrong. > > With both cards connected to the same switch (long-term intention of > course is that the 100M card will connect to a standby switch) it > insists on using the 100M card, even when the 1G is set as the > "preferred primary" and the 100M is the "preferred secondary". Both > cards definitely work... if I unplug the connection to the 100M, the 1G > takes over. With only the 100M connected, it works. > > Now, here's the very odd bit. You'd expect better performance from the > 1G card. But no. Testing with file copies to or from another server that > has been working fine with a 1G card for a year or so (attached via > fiber to a GBIC on the supervisor card on the switch), I get several > times times better performance with the 100M NIC than I do with the 1G > (both UTP). > > I've tried different cables. All are BICC GigaPlus. The 100M connection > goes through a patch panel, but I've run a 20M flylead direct from the > server to the switch for the 1G connection. > > The switch is a Cisco Catalyst 6000 with the 100M connections going to > 48-port 100M cards, and the 1G connections going to a 16-port 1G card. > Software, firmware, etc versions pasted below. > > Seeing much worse performance from Gigabit adapters compared to 100M is > something of a disappointment, to say the least. > > Any ideas? > > The hardware and versions: > > WS-C6006 Software, Version NmpSW: 7.2(2) > Copyright (c) 1995-2002 by Cisco Systems > NMP S/W compiled on Jun 3 2002, 18:30:10 > > System Bootstrap Version: 5.3(1) > System Web Interface Version: Engine Version: 5.3.4 ADP Device: Cat6000 > ADP Ver0 > > Hardware Version: 1.0 Model: WS-C6006 Serial #: XXX > > PS1 Module: WS-CAC-1300WSerial #: XXX > PS2 Module: WS-CAC-1300WSerial #: XXX > > Mod Port Model Serial #Versions > --- --- --- > -- > 1 2WS-X6K-SUP1A-2GEXXX Hw : 3.1 > Fw : 5.3(1) > Fw1: 5.1(1)CSX > Sw : 7.2(2) > Sw1: 7.2(2) > WS-F6K-PFC XXX Hw : 1.0 > 3 8WS-X6408-GBIC XXX Hw : 2.1 > Fw : 4.2(0.24)VAI78 > Sw : 7.2(2) > 4 48 WS-X6248-RJ-45 XXX Hw : 1.1 > Fw : 4.2(0.24)VAI78 > Sw : 7.2(2) > 5 48 WS-X6248-RJ-45 XXX Hw : 1.4 > Fw : 5.4(2) > Sw : 7.2(2) &
RE: Very poor performance on Cat 6000 gigabit? [7:57695]
You are right you can't team 2 different speed nics. But like I said I could not even get teaming to work with the hp drivers with 2 of the same nics, that is why I recommended getting another 1 gig nic and using gigachannel or either use fast etherchannel with 2 100 meg nics and you do not have to worry about flaky software. -Original Message- From: steve [mailto:[EMAIL PROTECTED]] Sent: Thursday, November 21, 2002 12:41 PM To: [EMAIL PROTECTED] Subject: Re: Very poor performance on Cat 6000 gigabit? [7:57695] hi, as far as i was aware you CAN`T team to different speed network cards we use the intel/Compaq/HP (the same cards/drivers) and i have not been able to get the teaming to work with 100/1000 . if you put 2 1g`s togther ...no problem2 100`s ...again no problem but different speed`s NOPE.. HTH steve - Original Message - From: "Elijah Savage III" To: Sent: Tuesday, November 19, 2002 2:31 PM Subject: RE: Very poor performance on Cat 6000 gigabit? [7:57695] > If you get this to work keep me/us informed as I am sure you will. > Because I could never get this to work, I actually had to buy another > 1gig nic and still the drivers did not work correctly actually eneded > up just using fast etherchannel which is working great. > > -Original Message- > From: Martin Reilly [mailto:[EMAIL PROTECTED]] > Sent: Tuesday, November 19, 2002 6:45 AM > To: [EMAIL PROTECTED] > Subject: Very poor performance on Cat 6000 gigabit? [7:57695] > > > Here's something annoying that I came across yesterday... any clues as > to what's going wrong would be very much appreciated. > > Scenario: > > HP NetServers with built-in 100M NICs, based on an Intel chipset. > > With the HP drivers, the performance is fine - as you'd expect from a > 100M connection. With Intel drivers, nothing changes. Still fine. > > Add a 1G NIC, again HP badged but with an Intel chipset (Intel > Pro/1000TX), and bind them together into a fault-tolerant set using > the Intel drivers that were priovided by HP (they don't provide HP > badged drivers for this card, though they are happy to sell it with an > HP sticker on it for twice the cost of the "Intel" card). My intention > of course is that the 1G adapter is the primary (and set so in the > "teamed adapter" settings) and the 100M would only be used as a > fallback if the 1G fails. > > That's where things go wrong. > > With both cards connected to the same switch (long-term intention of > course is that the 100M card will connect to a standby switch) it > insists on using the 100M card, even when the 1G is set as the > "preferred primary" and the 100M is the "preferred secondary". Both > cards definitely work... if I unplug the connection to the 100M, the > 1G takes over. With only the 100M connected, it works. > > Now, here's the very odd bit. You'd expect better performance from the > 1G card. But no. Testing with file copies to or from another server > that has been working fine with a 1G card for a year or so (attached > via fiber to a GBIC on the supervisor card on the switch), I get > several times times better performance with the 100M NIC than I do > with the 1G (both UTP). > > I've tried different cables. All are BICC GigaPlus. The 100M > connection goes through a patch panel, but I've run a 20M flylead > direct from the server to the switch for the 1G connection. > > The switch is a Cisco Catalyst 6000 with the 100M connections going to > 48-port 100M cards, and the 1G connections going to a 16-port 1G card. > Software, firmware, etc versions pasted below. > > Seeing much worse performance from Gigabit adapters compared to 100M > is something of a disappointment, to say the least. > > Any ideas? > > The hardware and versions: > > WS-C6006 Software, Version NmpSW: 7.2(2) > Copyright (c) 1995-2002 by Cisco Systems > NMP S/W compiled on Jun 3 2002, 18:30:10 > > System Bootstrap Version: 5.3(1) > System Web Interface Version: Engine Version: 5.3.4 ADP Device: > Cat6000 ADP Ver0 > > Hardware Version: 1.0 Model: WS-C6006 Serial #: XXX > > PS1 Module: WS-CAC-1300WSerial #: XXX > PS2 Module: WS-CAC-1300WSerial #: XXX > > Mod Port Model Serial #Versions > --- --- --- > -- > 1 2WS-X6K-SUP1A-2GEXXX Hw : 3.1 > Fw : 5.3(1) > Fw1: 5.1(1)CSX > Sw : 7.2(2) >