home *** CD-ROM | disk | FTP | other *** search
- On Tue, 26 Mar 1996, Bob Nixon wrote:
-
- > Appology accepted, I admittedly don't understand the inner workings of how
- > Win-95 handles the com ports or if the DTE settings above 115200 are real
- > but concerning the mutitasking while moving files both in and out plus
- > Bi-directional transfers here are my nunbers for single and by-directional
- > transfers of highly compressable files while running a DOS graphics based
- > game running in demo mode in the background. My system is AMD486DX4-120
- > w/32megs of ram and Diamonds Stealth-32 graphics card. The modem is USR
- > Courier with V.34+ code. Tranfers for single or one way file transfer of a
- > Coreldraw.cdr(border file) 10600cps both inbound and outbound with a
- > 28800/28800 connection. Bidirections transfer slowed to about 8000cps on
- > each direction. Both these numbers are nearly identical to the speeds that
- > I've gotten on previous tests. BTW transfer method was WS-FTP32 with a
- > Win-95 ppp connection to my ISP, Primenet in Phoenix, AZ at 9:30PM(busy
- > prime hours) to my home directory.
- >
- > PS. Comm overruns were not observed during any testing.
-
- Ok, let me try to explain what's going on here. First, the modems between
- the computers are controlling the transfer in conjunction with the CPU's
- at each end. In addition, the 115200 DTE rate is bottle-necked down to
- 28800 between the modems. This is true even though the throughput rate is
- above 10600 cps. Why? you might ask. Well, the modems compress the data
- first *before* sending it. This also helps explain why the throughput
- in each direction drops when doing bi-directional transfers on
- uncompressed files; the modems each are doing double-duty compressing and
- uncompressing each data stream.
-
- Your numbers pretty much match mine on modem to modem transfers both in
- uni-directional and bi-directional transfers. I presume you used HS/Link
- for the bi-directional (and, from the rates on uni-directional, probably
- on those also)? I need to try this with Smodem (in my opinion, a more
- efficient and fault-free bi-directional protocol) and see what I get on a
- bi-directional transfer with that.
-
- Now, if you have two computers, hook up a null modem cable between them
- and open each port at 115200. Take a file and transfer it. That was my
- bench test for comparison between the linux and DOS platforms. No modems,
- no LapLink, just two comm programs (Commo on one side and minicom on the
- other) using zmodem (DSZ at one end, the built-in zmodem at the other)
- I would be interested to see results of a 230k port setting at each end
- with DOS or Win95 at each end also.
-
- The tests were made in one direction only; from the DOS box to the
- linux/DOS box, with the linux/DOS box booted to DOS and then booted to
- linux. When the linux/DOS box was in DOS, the comm program was Commo
- w/DSZ as the zmodem protocol drive (also did it with DSZ as a standalone).
-
- Again, I am not using Win95 but I do know a little about that platform.
- It isn't really a true operating system, it runs on DOS 7.0 so it's
- subject to similar problems that any DOS box might have. It is, I am
- informed, much improved over Windows 3.1 (and WFWG 3.11) but I still see
- complaints about throughput and overruns.
-
- What amazed me about linux was the absolute lack of errors during a true
- hi-speed transfer (115200 from end to end). I was used to limiting the
- DOS-to-DOS connections to 57600 in order to limit the errors during transfer.
-
-