Dear Ceph, While discussing @cephdays London about Erasure Coding, one of the main drawbacks of the technology seemed to be the trade off between size and computation requirements even on small footprint operations (concat, append...) One of my colleagues is currently working on homomorphic encryption, which is basically a way to encrypt content that provides a very useful feature: operations made on encrypted content reflects on unencrypted content. Say for example that A and B are the unencrypted objects, A' and B' the encrypted objects, and C = F(A,B) the result of an function applying on A and B that is homomorphic. Then C' = F(A', B'). A simple example of that would be concatenation. If I have 2 strings A and B, then decrypting A+B has the same value as A'+B'. Anyway this made me think that object storage might benefit from such application if and only if the same principles were also working for erasure codes and form "Homomorphic Erasure Coding". Lucky me there seem to be some papers around related to such application of homomorphic codes: http://arxiv.org/abs/1008.0064 (pdf: http://arxiv.org/pdf/1008.0064.pdf) http://blog.notdot.net/2012/08/Damn-Cool-Algorithms-Homomorphic-Hashing http://www.pdl.cmu.edu/PDL-FTP/SelfStar/podc07.pdf Or any result on duckduckgo of research https://duckduckgo.com/?q=homorphic+erasure+code Not being a coder nor mathematician myself I may of course repeat something trivial for you guys (and in that case I apologize) but it seems to me such codes would dramatically improve simple operations on erasure coded objects, and would make sense... It's also my first message to the list, let me know if there is any policy I didn't follow, I'd be happy to comply to best practices... Hope this opens new ideas... Best, Samuel Cozannet -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
