Encryption

Encryption

Postby Dandy » Sat Nov 29, 2014 1:46 am

Looking for a guide measure the processing power when encrypting. I have a 1Gb file of data. The size is fixed, not variable. I need to encrypt that file and before doing it, I need to know how long it would take. The encryption is AES (asymmetric) on 256bit. Let's consider the Epiphany IV. Can someone show me how you compute the time? Also, I'd be interested to know how that relates to an i5 (i5-4430, 4 cores, running at 3.0Ghz) or i7-4770(4 cores @3,4Ghz).
Dandy
 
Posts: 4
Joined: Sun Mar 02, 2014 10:09 pm

Re: Encryption

Postby Melkhior » Sat Nov 29, 2014 7:46 am

Dandy wrote:Looking for a guide measure the processing power when encrypting. I have a 1Gb file of data. The size is fixed, not variable. I need to encrypt that file and before doing it, I need to know how long it would take. The encryption is AES (asymmetric) on 256bit. Let's consider the Epiphany IV. Can someone show me how you compute the time? Also, I'd be interested to know how that relates to an i5 (i5-4430, 4 cores, running at 3.0Ghz) or i7-4770(4 cores @3,4Ghz).


You're going to need to be more accurate.

1) AES is a symetric cypher (it uses the same key for encryption and decryption);
2) You don't use AES directly, that would be ECB mode, which is woefully insecure;
3) Some modes are parallel, some are not. The non-parallels mode are unlikely to be helped by an Epiphany chip;
4) Most x86-64 chip have instructions dedicated to AES, so implementations using those instructions are very fast.

OpenSSL let you test the speed of algorithm, so it will give you an idea for CPUs. For instance you can try:
Code: Select all
openssl speed -evp aes-256-ecb

to measure AES with 256 bits key in ECB mode. For this algorithm, an older Core 2 Duo E8200 shows about 150-160 MB/s (single core, no dedicated instructions), whereas an i5-4570s goes up 3600 MB/s (single core, AES-NI instructions). The Cortex A9 in the Parallella shows about 18 MB/s. Getting data back from the Epiphany to the Cortex A9 memory in the Parallella seems limited to at the very best 100 MB/s (I achieve 45 MB/s for AES-256-CTR, which doesn't need to send data to the Epiphany).

HTH
Melkhior
 
Posts: 39
Joined: Sat Nov 08, 2014 12:19 pm


Return to Scientific Computing

Who is online

Users browsing this forum: No registered users and 1 guest