1: 2: More number for the questions about SSL overheads.... 3: 4: The following numbers were generated on a pentium pro 200, running linux. 5: They give an indication of the SSL protocol and encryption overheads. 6: 7: The program that generated them is an unreleased version of ssl/ssltest.c 8: which is the SSLeay ssl protocol testing program. It is a single process that 9: talks both sides of the SSL protocol via a non-blocking memory buffer 10: interface. 11: 12: How do I read this? The protocol and cipher are reasonable obvious. 13: The next number is the number of connections being made. The next is the 14: number of bytes exchanged bewteen the client and server side of the protocol. 15: This is the number of bytes that the client sends to the server, and then 16: the server sends back. Because this is all happening in one process, 17: the data is being encrypted, decrypted, encrypted and then decrypted again. 18: It is a round trip of that many bytes. Because the one process performs 19: both the client and server sides of the protocol and it sends this many bytes 20: each direction, multiply this number by 4 to generate the number 21: of bytes encrypted/decrypted/MACed. The first time value is how many seconds 22: elapsed doing a full SSL handshake, the second is the cost of one 23: full handshake and the rest being session-id reuse. 24: 25: SSLv2 RC4-MD5 1000 x 1 12.83s 0.70s 26: SSLv3 NULL-MD5 1000 x 1 14.35s 1.47s 27: SSLv3 RC4-MD5 1000 x 1 14.46s 1.56s 28: SSLv3 RC4-MD5 1000 x 1 51.93s 1.62s 1024bit RSA 29: SSLv3 RC4-SHA 1000 x 1 14.61s 1.83s 30: SSLv3 DES-CBC-SHA 1000 x 1 14.70s 1.89s 31: SSLv3 DES-CBC3-SHA 1000 x 1 15.16s 2.16s 32: 33: SSLv2 RC4-MD5 1000 x 1024 13.72s 1.27s 34: SSLv3 NULL-MD5 1000 x 1024 14.79s 1.92s 35: SSLv3 RC4-MD5 1000 x 1024 52.58s 2.29s 1024bit RSA 36: SSLv3 RC4-SHA 1000 x 1024 15.39s 2.67s 37: SSLv3 DES-CBC-SHA 1000 x 1024 16.45s 3.55s 38: SSLv3 DES-CBC3-SHA 1000 x 1024 18.21s 5.38s 39: 40: SSLv2 RC4-MD5 1000 x 10240 18.97s 6.52s 41: SSLv3 NULL-MD5 1000 x 10240 17.79s 5.11s 42: SSLv3 RC4-MD5 1000 x 10240 20.25s 7.90s 43: SSLv3 RC4-MD5 1000 x 10240 58.26s 8.08s 1024bit RSA 44: SSLv3 RC4-SHA 1000 x 10240 22.96s 11.44s 45: SSLv3 DES-CBC-SHA 1000 x 10240 30.65s 18.41s 46: SSLv3 DES-CBC3-SHA 1000 x 10240 47.04s 34.53s 47: 48: SSLv2 RC4-MD5 1000 x 102400 70.22s 57.74s 49: SSLv3 NULL-MD5 1000 x 102400 43.73s 31.03s 50: SSLv3 RC4-MD5 1000 x 102400 71.32s 58.83s 51: SSLv3 RC4-MD5 1000 x 102400 109.66s 59.20s 1024bit RSA 52: SSLv3 RC4-SHA 1000 x 102400 95.88s 82.21s 53: SSLv3 DES-CBC-SHA 1000 x 102400 173.22s 160.55s 54: SSLv3 DES-CBC3-SHA 1000 x 102400 336.61s 323.82s 55: 56: What does this all mean? Well for a server, with no session-id reuse, with 57: a transfer size of 10240 bytes, using RC4-MD5 and a 512bit server key, 58: a pentium pro 200 running linux can handle the SSLv3 protocol overheads of 59: about 49 connections a second. Reality will be quite different :-). 60: 61: Remeber the first number is 1000 full ssl handshakes, the second is 62: 1 full and 999 with session-id reuse. The RSA overheads for each exchange 63: would be one public and one private operation, but the protocol/MAC/cipher 64: cost would be quite similar in both the client and server. 65: 66: eric (adding numbers to speculation) 67: 68: --- Appendix --- 69: - The time measured is user time but these number a very rough. 70: - Remember this is the cost of both client and server sides of the protocol. 71: - The TCP/kernal overhead of connection establishment is normally the 72: killer in SSL. Often delays in the TCP protocol will make session-id 73: reuse look slower that new sessions, but this would not be the case on 74: a loaded server. 75: - The TCP round trip latencies, while slowing indervidual connections, 76: would have minimal impact on throughput. 77: - Instead of sending one 102400 byte buffer, one 8k buffer is sent until 78: - the required number of bytes are processed. 79: - The SSLv3 connections were actually SSLv2 compatable SSLv3 headers. 80: - A 512bit server key was being used except where noted. 81: - No server key verification was being performed on the client side of the 82: protocol. This would slow things down very little. 83: - The library being used is SSLeay 0.8.x. 84: - The normal mesauring system was commands of the form 85: time ./ssltest -num 1000 -bytes 102400 -cipher DES-CBC-SHA -reuse 86: This modified version of ssltest should be in the next public release of 87: SSLeay. 88: 89: The general cipher performace number for this platform are 90: 91: SSLeay 0.8.2a 04-Sep-1997 92: built on Fri Sep 5 17:37:05 EST 1997 93: options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) idea(int) blowfish(ptr2) 94: C flags:gcc -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -m486 -Wall -Wuninitialized 95: The 'numbers' are in 1000s of bytes per second processed. 96: type 8 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 97: md2 131.02k 368.41k 500.57k 549.21k 566.09k 98: mdc2 535.60k 589.10k 595.88k 595.97k 594.54k 99: md5 1801.53k 9674.77k 17484.03k 21849.43k 23592.96k 100: sha 1261.63k 5533.25k 9285.63k 11187.88k 11913.90k 101: sha1 1103.13k 4782.53k 7933.78k 9472.34k 10070.70k 102: rc4 10722.53k 14443.93k 15215.79k 15299.24k 15219.59k 103: des cbc 3286.57k 3827.73k 3913.39k 3931.82k 3926.70k 104: des ede3 1443.50k 1549.08k 1561.17k 1566.38k 1564.67k 105: idea cbc 2203.64k 2508.16k 2538.33k 2543.62k 2547.71k 106: rc2 cbc 1430.94k 1511.59k 1524.82k 1527.13k 1523.33k 107: blowfish cbc 4716.07k 5965.82k 6190.17k 6243.67k 6234.11k 108: sign verify 109: rsa 512 bits 0.0100s 0.0011s 110: rsa 1024 bits 0.0451s 0.0012s 111: rsa 2048 bits 0.2605s 0.0086s 112: rsa 4096 bits 1.6883s 0.0302s 113: