Results 1 to 10 of 12

Thread: Linux DM-Crypt Being Parallelized

Hybrid View

  1. #1
    Join Date
    Jan 2007
    Posts
    14,590

    Default Linux DM-Crypt Being Parallelized

    Phoronix: Linux DM-Crypt Being Parallelized

    Red Hat's Mikulas Patocka has posted a set of Linux kernel patches that parallelize the dm-crypt sub-system...

    http://www.phoronix.com/vr.php?view=MTM0NzQ

  2. #2
    Join Date
    Jan 2012
    Posts
    57

    Default

    On modern hardware, I suspect this is pretty much moot anyway, thanks to AES-NI, right?

  3. #3

    Default

    Quote Originally Posted by pdffs View Post
    On modern hardware, I suspect this is pretty much moot anyway, thanks to AES-NI, right?
    probably, but that's only available since the last 2 processor "generations", or something like that. for normal workloads most people don't need anything faster than a core2quad and these don't have it.

  4. #4
    Join Date
    Dec 2011
    Posts
    2,023

    Default

    AES-NI can be used by dm-crypt, TrueCrypt, 7-Zip, OpenSSL, and by libraries such as Network Security Services (NSS), and the Crypto API in the Linux kernel.
    Last edited by uid313; 04-10-2013 at 05:36 AM.

  5. #5
    Join Date
    Mar 2013
    Posts
    2

    Default

    Not everything is x86 or x86_64. Recent quad-core ARM CPUs obviously lack AES-NI and could conceivably benefit from this.

    Also, AES-NI only improves performance for AES. AES is probably the most common algorithm for bulk encryption, but it is not the only algorithm out there.

  6. #6
    Join Date
    Jan 2012
    Posts
    18

    Default

    AES-NI isn't available in low-end intel processors. Not even the last generation of i3's has it (i3-3240 for example: http://ark.intel.com/products/65690/...Cache-3_40-GHz). Most i5 has it, but not all (2nd gen i5: i5-750 http://ark.intel.com/products/42917/...Cache-2_40-GHz). Intels "market segmentation is a bit weird and not always easy to follow. I always check ark.intel.com before pressing "buy" on processors to make sure I get the features I want.

    Either way, parallelization seems reasonable if it can be made to not hurt efficiency when io-loads are low.

  7. #7
    Join Date
    Jan 2009
    Location
    Vienna, Austria; Germany; hello world :)
    Posts
    629

    Default

    Quote Originally Posted by swrx View Post
    Not everything is x86 or x86_64. Recent quad-core ARM CPUs obviously lack AES-NI and could conceivably benefit from this.

    Also, AES-NI only improves performance for AES. AES is probably the most common algorithm for bulk encryption, but it is not the only algorithm out there.
    exactly

    my core i7 e.g. lynnfield doesn't have aes-ni or avx - so I'm grateful that all the x86_64 optimizations exist (CONFIG_CRYPTO_BLOWFISH_X86_64, CONFIG_CRYPTO_CAMELLIA_X86_64, CONFIG_CRYPTO_SERPENT_SSE2_X86_64, CONFIG_CRYPTO_TWOFISH_X86_64, CONFIG_CRYPTO_TWOFISH_X86_64_3WAY)


    will try those patches out - let's see if it cuts down latency on these

    dm-crypt/luks -> lvm -> filesystem layouts


    edit:

    ok seems like it's still heavily worked on, debated

    so will try it a later time
    Last edited by kernelOfTruth; 04-29-2013 at 01:39 PM.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •