These examples come from my 11/73 just now with a DELQA Ethernet and CMD CQD-220/TM SCSI adapter with a SCSISD model 5.2 (I think) with firmware version 4.2 disk on ra0 and a DEC RZ26 disk on ra1. For the purposes of these tests, it is not important to get exact timings and I/O rates, but to understand approximately what is the ballpark for I/O throughput.
The CQD-220/TM manual states it is capable of transfers up to 4.8 MiB/s in synchronous mode and 3.0 MiB/s in asychronous mode. And I don't know what sort of disk fragmentation my file system has. Contiguous sequential I/O usually performs better than random I/O for older storage devices.
You can use the dd command to see how well the disk and file system perform. First we'll test the read speed for the SCSI2SD ra0 hard drive.
Code:
# time dd if=/dev/rra0e of=/dev/null bs=16k count=1024
1024+0 records in
1024+0 records out
25.7 real 0.2 user 6.6 sys
16 MiB in 25.7 seconds gives a read rate of 652,810 bytes/second from the raw disk device.
Now for the RZ26 ra1 hard drive.
Code:
# time dd of=/dev/null bs=16k count=1024 if=/dev/rra1a
1024+0 records in
1024+0 records out
22.6 real 0.2 user 6.0 sys
A somewhat better rate of 742,355 bytes/sec.
Next we'll write and read through the file system on the SCSI2SD ra0 disk. My 11/73 only has 3 MiB of RAM, so writing 16 MiB will push data through the file system cache to the disk. This version of dd doesn't have an option to use direct I/O bypassing the disk cache, so we'll run a sync afterwards to flush the cache to disk.
Code:
# time dd if=/dev/zero of=testfile1 bs=16k count=1024
time sync
1024+0 records in
1024+0 records out
268.7 real 0.2 user 255.7 sys
# 0.3 real 0.0 user 0.1 sys
Giving a write rate of 62,369 bytes/sec. Not very encouraging given what today's hardware is capable of, but it is what this combination is capable of.
Next we read back the data (some of the end of the written file may still be in the cache, but reading from the start of the file will evict it).
Code:
# time dd if=testfile1 of=/dev/null bs=16k
1024+0 records in
1024+0 records out
149.5 real 0.3 user 146.5 sys
Giving a read rate of 112,222 bytes/sec. There are a variety of tools to characterize file system I/O performance, a search will turn some up, such as ioperf or fio (unknown how easy to port to old non-ANSI C compiler).
Next we'll test FTP reads from my MacBook Pro, across a 5 GHz WiFi through two Cisco 1 Gb/s small office 8 port switches to the RJ-45 to AUI adapter on the 11/73.
Code:
ftp> get testfile1
200 PORT command successful.
150 Opening BINARY mode data connection for testfile1 (16777216 bytes).
226 Transfer complete.
16777216 bytes received in 567 seconds (28.9 kbytes/s)
You can see the transfer rate is similar to the one you had. In this case since we aren't sending data from my MacBook to the 11/73, we don't have to be concerned with the link speed disparity caused by the DELQA 10 Mb/s link rate.
If you really want to test the network stack and Ethernet adapter rates, you could use something like iperf3 (or an older version, I don't know offhand if any will compile on old non-ANSI C compilers). Iperf3 is a very flexible tool which will show any issues with the network connection pretty clearly, including lost packets.
