SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    iSCSI: mandating markers (was Nailing down CRC-32C)


    • To: "iSCSI" <ips@ece.cmu.edu>
    • Subject: iSCSI: mandating markers (was Nailing down CRC-32C)
    • From: "m nima" <m_nima_1@lycos.com>
    • Date: Mon, 14 May 2001 11:35:34 -0700
    • Content-Language: en
    • Content-Transfer-Encoding: 7bit
    • Content-Type: text/plain; charset=us-ascii
    • Organization: Lycos Mail (http://mail.lycos.com:80)
    • Reply-To: m_nima_1@lycos.com
    • Sender: owner-ips@ece.cmu.edu

    
    Marker - I'm confused now. Assuming the server side uses 10G and has 
    marker in HW, What is the fate of its client running 1G/S and has a 
    SW marker?
    
    Repeat of garbled comment (sorry) - I think the key for iSCSI success 
    is to make sure it's low weight protocol that scales well even to 
    10Gb/S and beyond. For that, everything needs to be computable at the 
    receiver, on a per TCP segment basis and not span multiple segments! 
    
    Adding markers, making security mandatory to implement and use of 
    Data CRC (as proposed now where it is not friendly with out-of-order) 
    leaves iSCSI far behind.
    
    SW support - well today you have to provide your own TCP stack in the 
    MS environment. Every vendor that wishes can implement these changes 
    in his stack. So I don't see the point about use of legacy. 
    
    I believe, the standardization effort HAS to enable a new way, such 
    that one could do a better job (like you said - so that iSCSI gets 
    cheaper and better than FC), and framing is a better way for the 
    future
    
    Nima
    
    
    
    
    
    Get 250 color business cards for FREE!
    http://businesscards.lycos.com/vp/fastpath/
    


Home

Last updated: Tue Sep 04 01:04:43 2001
6315 messages in chronological order