Project

General

Profile

Persistent data » History » Revision 10

Revision 9 (Greg Burri, 09/02/2009 09:23 PM) → Revision 10/18 (Greg Burri, 09/02/2009 09:25 PM)

h1. Persistent data Structures 

 h2. Files cache 

 This file contains all the shared files and theirs SHA-1 hashes. This file is saved in user space after an update of the file list, it is completely loaded at the start of Aybabtu. If some paths of some files doesn't match the shared directory they are discarded. 

 This file is saved as a binary file serialized by protocol buffer. 

 You can see the proto file here : source:/application/Protos/files_cache.proto 

 h3. Size 

 Here is a size approximation for 50'000 files stored in 1 TB. The overhead of the file format is not taken in account. 

 <pre> 
 Nb of files : 50'0000 
 Total size : 1 TB 
 Chunk Size : 32 MB 
 Average file per directory : 50 
 Average filename and dirname length : 50 char 
 Size of a hash : 20B 
 </pre> 

 <pre> 
 AverageFileSize = 1*1024^4 / 50'000 = 21MB 
 Sha1Size = floor(AverageFileSize / 32MB) * 20 * 50'000 = 1 * 20 * 50'000 = 977kB 
 FilenameSize = 50'000 * 50 * 2 = 4.8MB (2 bytes per char) 
 DirnameSize = (50'000 / 50) * 50 * 2 = 97kB 

 TotalSize = 977kB + 4.8MB + 97kB = 5.9MB 
 </pre> 

 h2. Queued files and directories 

 A file containing the queued files and directories is saved in user space. An item from the list may contains a remote peer id and/or the hashes which identify a file.  

 As the file cache file (see above), this file is saved as a binary file serialized by protocol buffer. 

 For more details see the file : source:/application/Protos/queue.proto 


 h2. Configuration file 

 TODO..