Data collected on a QNX 4 system (600 MHz P3) is copied to a WinXP based
system (1.0 GHz Laptop )for post-processing. Currently the copy happens
across a 100 Mbit LAN connection using nfs (MS Services for UNIX). The copy
uses a PERL script with several "cp -Rcvn " in it. The data being copied
consists of a mix of binary files sized from 200MB down to several KB.
Several tests have been performed to determine the effective data transfer
rate of a number of different transfer methods using a fixed set of 36 files
totaling 605MB (5 files account for 550MB, the remainder are small):
QNX 4 to Laptop via 100 Mb LAN/nfs - Effective Transfer Rate =
1.0 MB/s
QNX 4 / USB 1.0 / USB FireLite SmartDisk - Effective Transfer Rate =
0.5 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate =
1.6 MB/s
QNX 4 / USB 2.0 / USB Flash Memory Stick - Effective Transfer Rate = 1.4
MB/s
By way of comparison, copying the 605MB data from Windows to Windows
produced transfer rates of:
WinXP Laptop via 100 Mb LAN to W2k Network Server - Effective
Transfer Rate = 4.0 MB/s
WinXP Laptop via 100 Mb LAN to 366 MHz W2k Desktop - Effective Transfer
Rate = 6.4 MB/s
The -O option for cp has been tried and produced no improvement in copy
times. The USB SmartDisk is capable of max sustained transfer rates of
15MB/s. A 100 Mb LAN should be capable of at least 10 MB/s.
What can be done from the QNX 4 end to improve the data transfer
performance?
Data collected on a QNX 4 system (600 MHz P3) is copied to a WinXP based
system (1.0 GHz Laptop )for post-processing. Currently the copy happens
across a 100 Mbit LAN connection using nfs (MS Services for UNIX). The
copy uses a PERL script with several "cp -Rcvn " in it. The data being
copied consists of a mix of binary files sized from 200MB down to several
KB.
Several tests have been performed to determine the effective data transfer
rate of a number of different transfer methods using a fixed set of 36
files totaling 605MB (5 files account for 550MB, the remainder are small):
QNX 4 to Laptop via 100 Mb LAN/nfs - Effective Transfer Rate =
1.0 MB/s
QNX 4 / USB 1.0 / USB FireLite SmartDisk - Effective Transfer Rate =
0.5 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate =
1.6 MB/s
QNX 4 / USB 2.0 / USB Flash Memory Stick - Effective Transfer Rate = 1.4
MB/s
By way of comparison, copying the 605MB data from Windows to Windows
produced transfer rates of:
WinXP Laptop via 100 Mb LAN to W2k Network Server - Effective
Transfer Rate = 4.0 MB/s
WinXP Laptop via 100 Mb LAN to 366 MHz W2k Desktop - Effective Transfer
Rate = 6.4 MB/s
The -O option for cp has been tried and produced no improvement in copy
times. The USB SmartDisk is capable of max sustained transfer rates of
15MB/s. A 100 Mb LAN should be capable of at least 10 MB/s.
What can be done from the QNX 4 end to improve the data transfer
performance?
What disk driver (Fsys.xxx) are you using in QNX4?
PK
“Gord Sipko” <> gsipkoATnxtenergyDOTcom@xyz.com> > wrote in message
news:e61mfm$adl$> 1@inn.qnx.com> …
Data collected on a QNX 4 system (600 MHz P3) is copied to a WinXP based
system (1.0 GHz Laptop )for post-processing. Currently the copy happens
across a 100 Mbit LAN connection using nfs (MS Services for UNIX). The
copy uses a PERL script with several "cp -Rcvn " in it. The data being
copied consists of a mix of binary files sized from 200MB down to several
KB.
Several tests have been performed to determine the effective data transfer
rate of a number of different transfer methods using a fixed set of 36
files totaling 605MB (5 files account for 550MB, the remainder are small):
QNX 4 to Laptop via 100 Mb LAN/nfs - Effective Transfer Rate =
1.0 MB/s
QNX 4 / USB 1.0 / USB FireLite SmartDisk - Effective Transfer Rate =
0.5 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate =
1.6 MB/s
QNX 4 / USB 2.0 / USB Flash Memory Stick - Effective Transfer Rate = 1.4
MB/s
By way of comparison, copying the 605MB data from Windows to Windows
produced transfer rates of:
WinXP Laptop via 100 Mb LAN to W2k Network Server - Effective
Transfer Rate = 4.0 MB/s
WinXP Laptop via 100 Mb LAN to 366 MHz W2k Desktop - Effective Transfer
Rate = 6.4 MB/s
The -O option for cp has been tried and produced no improvement in copy
times. The USB SmartDisk is capable of max sustained transfer rates of
15MB/s. A 100 Mb LAN should be capable of at least 10 MB/s.
What can be done from the QNX 4 end to improve the data transfer
performance?
What disk driver (Fsys.xxx) are you using in QNX4?
PK
“Gord Sipko” <> gsipkoATnxtenergyDOTcom@xyz.com> > wrote in message
news:e61mfm$adl$> 1@inn.qnx.com> …
Data collected on a QNX 4 system (600 MHz P3) is copied to a WinXP based
system (1.0 GHz Laptop )for post-processing. Currently the copy happens
across a 100 Mbit LAN connection using nfs (MS Services for UNIX). The
copy uses a PERL script with several "cp -Rcvn " in it. The data being
copied consists of a mix of binary files sized from 200MB down to several
KB.
Several tests have been performed to determine the effective data transfer
rate of a number of different transfer methods using a fixed set of 36
files totaling 605MB (5 files account for 550MB, the remainder are small):
QNX 4 to Laptop via 100 Mb LAN/nfs - Effective Transfer Rate =
1.0 MB/s
QNX 4 / USB 1.0 / USB FireLite SmartDisk - Effective Transfer Rate =
0.5 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate =
1.6 MB/s
QNX 4 / USB 2.0 / USB Flash Memory Stick - Effective Transfer Rate = 1.4
MB/s
By way of comparison, copying the 605MB data from Windows to Windows
produced transfer rates of:
WinXP Laptop via 100 Mb LAN to W2k Network Server - Effective
Transfer Rate = 4.0 MB/s
WinXP Laptop via 100 Mb LAN to 366 MHz W2k Desktop - Effective Transfer
Rate = 6.4 MB/s
The -O option for cp has been tried and produced no improvement in copy
times. The USB SmartDisk is capable of max sustained transfer rates of
15MB/s. A 100 Mb LAN should be capable of at least 10 MB/s.
What can be done from the QNX 4 end to improve the data transfer
performance?
Before changing to Fsys.atapi I did a baseline copy test writing to the host computer’s own hd:
QNX 4 to HardDisk - Effective Transfer Rate = 1.2 MB/s
Changing Fsys.eide to Fsys.atapi the results for two of the tests are:
QNX 4 to HardDisk - Effective Transfer Rate = 5.8 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate = 3.6 MB/s
A significant improvement of 2x - 5x with just the Fsys driver change! This is a step in the right direction. Now to find other improvements…
The hardware/cabling being used supports UDMA 2. With this setting in Fsys.atapi the max rate is now 6.6M/s - a 13% improvement! I do have an 80 conductor cable here, so I might try it as well.
What can be done to boost network performance?
Gord
5.8 with Fsys.atapi isn’t that much ( try playing with udma option)
The hardware/cabling being used supports UDMA 2. With this setting in Fsys.atapi the max rate is now 6.6M/s - a 13% improvement! I do have an 80 conductor cable here, so I might try it as well.
What can be done to boost network performance?
I’m pretty sure the issue is not the network itself. Make sure you are using the lastest version of TCP/IP (5.0), it contains many improvement in NFS.
Just to make sure it’s not NFS, try to ftp to and from the windows machine and compare the performance. Note that the harddisk could have been the bottleneck. Have you try network copy with the new setting.
Gord
5.8 with Fsys.atapi isn’t that much ( try playing with udma option)
Before changing to Fsys.atapi I did a baseline copy test writing to the host
computer’s own hd:
QNX 4 to HardDisk - Effective Transfer Rate = 1.2 MB/s
Changing Fsys.eide to Fsys.atapi the results for two of the tests are:
QNX 4 to HardDisk - Effective Transfer Rate = 5.8 MB/s
QNX 4 / USB 2.0 / USB FireLite SmartDisk - Effective Transfer Rate =
3.6 MB/s
A significant improvement of 2x - 5x with just the Fsys driver change! This
is a step in the right direction. Now to find other improvements…
Why does QSSL setup the Fsys.eide as the default when the Fsys.atapi is
faster?
That (Fsys.atapi) driver is far newer than any QSSL’s install CD you could
Using the 80 conductor cable and udma=4 doesn’t change a thing. The disk drive supports up to udma 5 but I’m not sure about the disk controller. With UDMA enabled in the BIOS, boot messages say udma2 with all hardware combinations. The PIII motherboard has the Intel 440BX chipset with a 7111 eide controller. The 440BX chipset appears to support ATA-66 but who knows…
We don’t have TCP/IP 5.0 so that may be a limiting factor as well. Given the speed of the USB 2.0 copy, I might persue that route instead of the network - the end objective is to get the data on the USB disk.
I have tried to retest the network side of things and currently cannot even come close to my previous test results. The onboard PCI network card and the add-on USB 2.0 card have been sharing the same interrupt and I’m in the early stages of trying to get this sorted out. The USB card has been removed and the network speed is still way down.
I’m pretty sure the issue is not the network itself. Make sure you are using the lastest version of TCP/IP (5.0), it contains many improvement in NFS.
Just to make sure it’s not NFS, try to ftp to and from the windows machine and compare the performance. Note that the harddisk could have been the bottleneck. Have you try network copy with the new setting.
“Gord Sipko” <gsipkoATnxtenergyDOTcom@xyz.com> wrote in message news:e6n442$lt$1@inn.qnx.com…
Using the 80 conductor cable and udma=4 doesn’t change a thing. The disk drive supports up to udma 5 but I’m not sure about the disk controller. With UDMA enabled in the BIOS, boot messages say udma2 with all hardware combinations. The PIII motherboard has the Intel 440BX chipset with a 7111 eide controller. The 440BX chipset appears to support ATA-66 but who knows…
i don’t think UDMA over 2 were support by the 440BX chipset. I do think 6.6 Meg very low though. My 6 years old laptop gives me 10Mbytes/sec raw read speed.
We don’t have TCP/IP 5.0 so that may be a limiting factor as well. Given the speed of the USB 2.0 copy, I might persue that route instead of the network - the end objective is to get the data on the USB disk.
I have tried to retest the network side of things and currently cannot even come close to my previous test results. The onboard PCI network card and the add-on USB 2.0 card have been sharing the same interrupt and I’m in the early stages of trying to get this sorted out. The USB card has been removed and the network speed is still way down.
I’m pretty sure the issue is not the network itself. Make sure you are using the lastest version of TCP/IP (5.0), it contains many improvement in NFS.
Just to make sure it’s not NFS, try to ftp to and from the windows machine and compare the performance. Note that the harddisk could have been the bottleneck. Have you try network copy with the new setting.
Using the 80 conductor cable and udma=4 doesn’t change a thing. The
disk drive supports up to udma 5
I get stable ~18MBps throughput on “cp -V /dev/hd0t77 /dev/null”.
Tomorrow I’ll post the Advantech’s mobo model and the HDD details.