1 .. -*- coding: utf-8-with-signature -*-
3 ============================================
4 Performance costs for some common operations
5 ============================================
7 1. `Publishing an A-byte immutable file`_
8 2. `Publishing an A-byte mutable file`_
9 3. `Downloading B bytes of an A-byte immutable file`_
10 4. `Downloading B bytes of an A-byte mutable file`_
11 5. `Modifying B bytes of an A-byte mutable file`_
12 6. `Inserting/Removing B bytes in an A-byte mutable file`_
13 7. `Adding an entry to an A-entry directory`_
14 8. `Listing an A entry directory`_
15 9. `Checking an A-byte file`_
16 10. `Verifying an A-byte file (immutable)`_
17 11. `Repairing an A-byte file (mutable or immutable)`_
19 ``K`` indicates the number of shares required to reconstruct the file
22 ``N`` indicates the total number of shares produced (default: 10)
24 ``S`` indicates the segment size (default: 128 KiB)
26 ``A`` indicates the number of bytes in a file
28 ``B`` indicates the number of bytes of a file that are being read or
31 ``G`` indicates the number of storage servers on your grid
33 Most of these cost estimates may have a further constant multiplier: when a
34 formula says ``N/K*S``, the cost may actually be ``2*N/K*S`` or ``3*N/K*S``.
35 Also note that all references to mutable files are for SDMF-formatted files;
36 this document has not yet been updated to describe the MDMF format.
38 Publishing an ``A``-byte immutable file
39 =======================================
41 when the file is already uploaded
42 ---------------------------------
44 If the file is already uploaded with the exact same contents, same
45 erasure coding parameters (K, N), and same added convergence secret,
46 then it reads the whole file from disk one time while hashing it to
47 compute the storage index, then contacts about N servers to ask each
48 one to store a share. All of the servers reply that they already have
49 a copy of that share, and the upload is done.
59 when the file is not already uploaded
60 -------------------------------------
62 If the file is not already uploaded with the exact same contents, same
63 erasure coding parameters (K, N), and same added convergence secret,
64 then it reads the whole file from disk one time while hashing it to
65 compute the storage index, then contacts about N servers to ask each
66 one to store a share. Then it uploads each share to a storage server.
74 memory footprint: N/K*S
76 Publishing an ``A``-byte mutable file
77 =====================================
79 cpu: ~A + a large constant for RSA keypair generation
83 memory footprint: N/K*A
85 notes: Tahoe-LAFS generates a new RSA keypair for each mutable file that it
86 publishes to a grid. This takes up to 1 or 2 seconds on a typical desktop PC.
88 Part of the process of encrypting, encoding, and uploading a mutable file to a
89 Tahoe-LAFS grid requires that the entire file be in memory at once. For larger
90 files, this may cause Tahoe-LAFS to have an unacceptably large memory footprint
91 (at least when uploading a mutable file).
93 Downloading ``B`` bytes of an ``A``-byte immutable file
94 =======================================================
100 notes: When Tahoe-LAFS 1.8.0 or later is asked to read an arbitrary
101 range of an immutable file, only the S-byte segments that overlap the
102 requested range will be downloaded.
104 (Earlier versions would download from the beginning of the file up
105 until the end of the requested range, and then continue to download
106 the rest of the file even after the request was satisfied.)
108 Downloading ``B`` bytes of an ``A``-byte mutable file
109 =====================================================
117 notes: As currently implemented, mutable files must be downloaded in
118 their entirety before any part of them can be read. We are
119 exploring fixes for this; see ticket #393 for more information.
121 Modifying ``B`` bytes of an ``A``-byte mutable file
122 ===================================================
128 memory footprint: N/K*A
130 notes: If you upload a changed version of a mutable file that you
131 earlier put onto your grid with, say, 'tahoe put --mutable',
132 Tahoe-LAFS will replace the old file with the new file on the
133 grid, rather than attempting to modify only those portions of the
134 file that have changed. Modifying a file in this manner is
135 essentially uploading the file over again, except that it re-uses
136 the existing RSA keypair instead of generating a new one.
138 Inserting/Removing ``B`` bytes in an ``A``-byte mutable file
139 ============================================================
145 memory footprint: N/K*A
147 notes: Modifying any part of a mutable file in Tahoe-LAFS requires that
148 the entire file be downloaded, modified, held in memory while it is
149 encrypted and encoded, and then re-uploaded. A future version of the
150 mutable file layout ("LDMF") may provide efficient inserts and
151 deletes. Note that this sort of modification is mostly used internally
152 for directories, and isn't something that the WUI, CLI, or other
153 interfaces will do -- instead, they will simply overwrite the file to
154 be modified, as described in "Modifying B bytes of an A-byte mutable
157 Adding an entry to an ``A``-entry directory
158 ===========================================
164 memory footprint: N/K*~A
166 notes: In Tahoe-LAFS, directories are implemented as specialized mutable
167 files. So adding an entry to a directory is essentially adding B
168 (actually, 300-330) bytes somewhere in an existing mutable file.
170 Listing an ``A`` entry directory
171 ================================
177 memory footprint: N/K*~A
179 notes: Listing a directory requires that the mutable file storing the
180 directory be downloaded from the grid. So listing an A entry
181 directory requires downloading a (roughly) 330 * A byte mutable
182 file, since each directory entry is about 300-330 bytes in size.
184 Checking an ``A``-byte file
185 ===========================
191 memory footprint: negligible
193 notes: To check a file, Tahoe-LAFS queries all the servers that it knows
194 about. Note that neither of these values directly depend on the size
195 of the file. This is relatively inexpensive, compared to the verify
196 and repair operations.
198 Verifying an A-byte file (immutable)
199 ====================================
205 memory footprint: N/K*S
207 notes: To verify a file, Tahoe-LAFS downloads all of the ciphertext
208 shares that were originally uploaded to the grid and integrity checks
209 them. This is (for grids with good redundancy) more expensive than
210 downloading an A-byte file, since only a fraction of these shares would
211 be necessary to recover the file.
213 Verifying an A-byte file (mutable)
214 ==================================
220 memory footprint: N/K*A
222 notes: To verify a file, Tahoe-LAFS downloads all of the ciphertext
223 shares that were originally uploaded to the grid and integrity checks
224 them. This is (for grids with good redundancy) more expensive than
225 downloading an A-byte file, since only a fraction of these shares would
226 be necessary to recover the file.
228 Repairing an ``A``-byte file (mutable or immutable)
229 ===================================================
231 cpu: variable, between ~A and ~N/K*A
233 network: variable; between A and N/K*A
235 memory footprint (immutable): (1+N/K)*S
236 (SDMF mutable): (1+N/K)*A
238 notes: To repair a file, Tahoe-LAFS downloads the file, and
239 generates/uploads missing shares in the same way as when it initially
240 uploads the file. So, depending on how many shares are missing, this
241 can cost as little as a download or as much as a download followed by
244 Since SDMF files have only one segment, which must be processed in its
245 entirety, repair requires a full-file download followed by a full-file