1 ============================================
2 Performance costs for some common operations
3 ============================================
5 1. `Publishing an A-byte immutable file`_
6 2. `Publishing an A-byte mutable file`_
7 3. `Downloading B bytes of an A-byte immutable file`_
8 4. `Downloading B bytes of an A-byte mutable file`_
9 5. `Modifying B bytes of an A-byte mutable file`_
10 6. `Inserting/Removing B bytes in an A-byte mutable file`_
11 7. `Adding an entry to an A-entry directory`_
12 8. `Listing an A entry directory`_
13 9. `Performing a file-check on an A-byte file`_
14 10. `Performing a file-verify on an A-byte file`_
15 11. `Repairing an A-byte file (mutable or immutable)`_
17 ``K`` indicates the number of shares required to reconstruct the file
20 ``N`` indicates the total number of shares produced (default: 10)
22 ``S`` indicates the segment size (default: 128 KiB)
24 ``A`` indicates the number of bytes in a file
26 ``B`` indicates the number of bytes of a file which are being read or
29 ``G`` indicates the number of storage servers on your grid
31 Publishing an ``A``-byte immutable file
32 =======================================
34 when the file is already uploaded
35 ---------------------------------
37 If the file is already uploaded with the exact same contents, same
38 erasure coding parameters (K, N), and same added convergence secret,
39 then it reads the whole file from disk one time while hashing it to
40 compute the storage index, then contacts about N servers to ask each
41 one to store a share. All of the servers reply that they already have
42 a copy of that share, and the upload is done.
52 when the file is not already uploaded
53 -------------------------------------
55 If the file is not already uploaded with the exact same contents, same
56 erasure coding parameters (K, N), and same added convergence secret,
57 then it reads the whole file from disk one time while hashing it to
58 compute the storage index, then contacts about N servers to ask each
59 one to store a share. Then it uploads each share to a storage server.
67 memory footprint: N/K*S
69 Publishing an ``A``-byte mutable file
70 =====================================
72 cpu: ~A + a large constant for RSA keypair generation
76 memory footprint: N/K*A
78 notes: Tahoe-LAFS generates a new RSA keypair for each mutable file that it
79 publishes to a grid. This takes up to 1 or 2 seconds on a typical desktop PC.
81 Part of the process of encrypting, encoding, and uploading a mutable file to a
82 Tahoe-LAFS grid requires that the entire file be in memory at once. For larger
83 files, this may cause Tahoe-LAFS to have an unacceptably large memory footprint
84 (at least when uploading a mutable file).
86 Downloading ``B`` bytes of an ``A``-byte immutable file
87 =======================================================
93 notes: When Tahoe-LAFS 1.8.0 or later is asked to read an arbitrary
94 range of an immutable file, only the S-byte segments that overlap the
95 requested range will be downloaded.
97 (Earlier versions would download from the beginning of the file up
98 until the end of the requested range, and then continue to download
99 the rest of the file even after the request was satisfied.)
101 Downloading ``B`` bytes of an ``A``-byte mutable file
102 =====================================================
110 notes: As currently implemented, mutable files must be downloaded in
111 their entirety before any part of them can be read. We are
112 exploring fixes for this; see ticket #393 for more information.
114 Modifying ``B`` bytes of an ``A``-byte mutable file
115 ===================================================
121 memory footprint: N/K*A
123 notes: If you upload a changed version of a mutable file that you
124 earlier put onto your grid with, say, 'tahoe put --mutable',
125 Tahoe-LAFS will replace the old file with the new file on the
126 grid, rather than attempting to modify only those portions of the
127 file that have changed. Modifying a file in this manner is
128 essentially uploading the file over again, except that it re-uses
129 the existing RSA keypair instead of generating a new one.
131 Inserting/Removing ``B`` bytes in an ``A``-byte mutable file
132 ============================================================
138 memory footprint: N/K*A
140 notes: Modifying any part of a mutable file in Tahoe-LAFS requires that
141 the entire file be downloaded, modified, held in memory while it is
142 encrypted and encoded, and then re-uploaded. A future version of the
143 mutable file layout ("LDMF") may provide efficient inserts and
144 deletes. Note that this sort of modification is mostly used internally
145 for directories, and isn't something that the WUI, CLI, or other
146 interfaces will do -- instead, they will simply overwrite the file to
147 be modified, as described in "Modifying B bytes of an A-byte mutable
150 Adding an entry to an ``A``-entry directory
151 ===========================================
157 memory footprint: N/K*~A
159 notes: In Tahoe-LAFS, directories are implemented as specialized mutable
160 files. So adding an entry to a directory is essentially adding B
161 (actually, 300-330) bytes somewhere in an existing mutable file.
163 Listing an ``A`` entry directory
164 ================================
170 memory footprint: N/K*~A
172 notes: Listing a directory requires that the mutable file storing the
173 directory be downloaded from the grid. So listing an A entry
174 directory requires downloading a (roughly) 330 * A byte mutable
175 file, since each directory entry is about 300-330 bytes in size.
177 Performing a file-check on an ``A``-byte file
178 =============================================
184 memory footprint: negligible
186 notes: To check a file, Tahoe-LAFS queries all the servers that it knows
187 about. Note that neither of these values directly depend on the size
188 of the file. This is relatively inexpensive, compared to the verify
189 and repair operations.
191 Performing a file-verify on an ``A``-byte file
192 ==============================================
198 memory footprint: N/K*S
200 notes: To verify a file, Tahoe-LAFS downloads all of the ciphertext
201 shares that were originally uploaded to the grid and integrity checks
202 them. This is (for well-behaved grids) more expensive than downloading
203 an A-byte file, since only a fraction of these shares are necessary to
206 Repairing an ``A``-byte file (mutable or immutable)
207 ===================================================
209 cpu: variable, between ~A and ~N/K*A
211 network: variable; between A and N/K*A
213 memory footprint: (1+N/K)*S
215 notes: To repair a file, Tahoe-LAFS downloads the file, and
216 generates/uploads missing shares in the same way as when it initially
217 uploads the file. So, depending on how many shares are missing, this
218 can cost as little as a download or as much as a download followed by