[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference irocz::terminal_servers

Title:	Terminal Servers
Notice:	See Note 2 for Directory of important notes. Please use keywords.
Moderator:	LAVC::CAHILLON

Created:	Tue May 14 1991
Last Modified:	Fri Jun 06 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	3547
Total number of notes:	12300

3456.0. "DS900 Max Port Throughput" by CSC32::L_DEGROFF () Thu Mar 06 1997 10:15

    Have a customer who is experiencing limited throughput on his DS900. 
    With one port in use, he expects to see a sustained flow (no flow
    control) of at least 38,400 bps (4,800 characters).  Instead, the port
    is toggling flow control (INPUT XOFFED: Yes) almost continually.
    
    Found only limited information in the DNAS v2.0 SPD.  Could not find
    any DS900 SPD, but the DS700 SPD shows a maximum sustainable throughput
    of 215,000 (8-bit) characters per RS422 port.  Given a Ethernet con-
    sisting of only a DS900 and a host with no LAT service annoucements
    (QIO connection to the DS900 port to retrieve a continuous stream of
    data from a connected device,)  please advise the following:
    
    1.  Port buffer size:
    
    2.  Server Buffer size (outgoing packets):
    
    3.  What is the maximum sustainable, demonstrated port throughput
        without flow control for one port?  Ten ports?
    
    4.  Is there any way, other than adjusting the CIRCUIT TIMER, to increase
        the throughput?
    
    Thanks
    Larry

T.R	Title	User	Personal Name	Date	Lines
3456.1		IROCZ::D_NELSON	Dave Nelson LKG1-3/A11 226-5358	`Thu Mar 06 1997 10:40`	9
	RE: .0 What is the customer doing with the port: LAT, LAT Service, Telnet, Telnet Listener, Raw TCP Listener, SLIP, CSLIP, or PPP? Regards, Dave
3456.2	LAT	CSC32::L_DEGROFF		`Thu Mar 06 1997 17:03`	5
	Dave, The port is providing a LAT connection for a modem. Larry
3456.3		TWICK::PETTENGILL	mulp	`Wed Apr 30 1997 01:19`	23
	Assuming a circuit timer of 80ms this is about 12 messages per second with two slots per message and a maximum slot size of 250, or about 6000 characters per secondor about 60 kilobits. However, the terminal server will send XOFF before its buffer is close to filling so that will limit the rate further. The above makes some optimistic assumptions about the buffer and credit management in the terminal server that might not be true. It certainly was NOT true a decade ago when I was involved and we were supporting 50 terminals in 128KB and 4 meg was more like the size of a disk. You can decrease the circuit timer which will allow the terminal server to send messages more often. What's wrong with this? LAT does have an idle mode that it enters after to empty messages where messages are not exchanged for the keepalive time which is typically 20 seconds. I can see that our focus on terminals when working on the design of the Local Area Terminal protocol blinded us to some tuning parameters. We also focused on the disparity between the rate that users can type and the rate that systems would generate output. ;-)