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

Re: IPsec and TCP

To: Phil Karn <karn@qualcomm.com>
Subject: Re: IPsec and TCP
From: Craig Metz <cmetz@inner.net>
Date: Tue, 07 Jan 1997 08:32:27 -0500
cc: ipsec@tis.com, naganand@ftp.com, ho@earth.hpc.org
In-reply-to: Your message of "Mon, 06 Jan 1997 19:20:50 PST." <199701070320.TAA03077@servo.qualcomm.com>
Sender: owner-ipsec@ex.tis.com

In message <199701070320.TAA03077@servo.qualcomm.com>, you write:
>>This is interesting.  I knew that TCP/IP will be miserably slow when
>>you throw DES into it (about 1 Mbit/sec @ 200 Mhz on a Pentium) but I 
>
>I don't know where you get such pessimistic figures. My own DES (in a
>mix of C and 386 assembler) gets 15.9 megabits/sec on a 133 (not 200)

	In a *highly* unscientific experiment, I was able to move a 100MB
file via FTP (TCP/IPv4) between a P120 and a P90 with both ESP DES-CBC and AH
HMAC-SHA on every packet and got an average throughput of about 3.18Mb/s. This
is with a software implementation that is unoptimized and has lots of debug
code in place. It would not at all surprise me if a well-optimized
implementation could at least double that. You want faster? Use hardware
assists for the crypto.

									-Craig

Follow-Ups:

Re: IPsec and TCP

From: andrade@netcom.com (Andrade Software Andrade Networking)

References:

Re: IPsec and TCP

From: Phil Karn <karn@qualcomm.com>

Prev by Date: Re: IPsec and TCP
Next by Date: Re: [graydon@gabber.multinet.net: Thinkin' about identifiers...]
Prev by thread: Re: IPsec and TCP
Next by thread: Re: IPsec and TCP
Index(es):
- Main
- Thread