Simply put: something I do, read, care about and hopefully might interest somebody else too.
|
posted Feb 13, 2012 9:01 AM by Sami Lehtinen
[
updated Feb 13, 2012 9:55 AM
]
I did some simple file system file allocation comparison. How files are allocated when allocation happens differently. I did this just for fun. Results are quite interesting. Especially if you don't have time or will to repeat tests.
Initialization:
I1: I formatted multiple 100 gigabyte volumes systems, one using ext4, one using ntfs and one using ntfs.
I2: I copied about 300k randomly sized files to file system, totalling about 6.9 gigabytes.
Tests:
T1: Copied 3.1 gigabyte file to volume. Final size was known and therefore pre-allocation could be used.
T2: Quickly created same file, final size unknown to system (pre-allocation unavailable), file allocated when it grows quickly. (about 80Mbytes/s)
T3: Slowly created same file, final size unknown to system (pre-allocation unavailable), file allocated when it grows slowly. (about 2Mbytes/s)
Results for BtrFS:
B1: Excellent result, 7 extents.
B2: Strange result, 46 extents, which some are very small.
B3: Very strange result, 531 extens, which almost 1/3rd contain only 1 block.
Results for Ext4:
E1: Excellent result, 29 extents. Please note that Ext4 got maximum extent size limit, only three of extents (excluding last one) are smaller than maximum.
E2: Very good result, 37 extents.
E3: Excellent result, 29 extents.
Results for Ntfs:
N1: Perfect results, 1 contiguous file.
N2: Skipped.
N3: Bad result, 5063 fragments. 614 kilobytes/fragment and for comparison it makes about 157 blocks(clusters)/fragment.
Conclusions:
Ext4 is very good what comes to avoiding fragmentation in general.
BtrFS still got clear delayed allocation flaws, slow allocation is actually quite bad.
NTFS, it behaves just it has behaved in all of my previous tests. When using pre-allocation results are perfect and with growing files, data is scattered all over disk.
Notes:
NTFS allocation was done using 64bit Windows 7, not Linux NTFS allocator, so can't claim that it's just Linux NTFS allocator that would cause fragmentation.
There was plenty of free disk space during these tests. It really outlines how bad BtrFS is now with slow allocation.
Same file content was used for all tests. BtrFS supports zlib and lzo. Also file content was already compressed using LZMA2.
If you want to use exactly same files for similar tests, I do have deliverable as one large bittorrent file.
Here's more detailed results if you're really interested.
BtrFS copy allocation:
File size of copy is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 3900207 40720
1 40720 4086804 3940926 34784
2 75504 4070400 4121587 2640
3 78144 4157441 4073039 147936
4 226080 4332544 4305376 248784
5 474864 4594688 4581327 232448
6 707312 4856832 4827135 108827 eof
copy: 7 extents found
BtrFS fast allocation:
File size of fast is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 4957187 1
1 1 4965660 4957187 77887
2 77888 4856832 5043546 70688
3 148576 4927521 4927519 480
4 149056 4957186 4928000 1
5 149057 4928002 4957186 29183
6 178240 5043548 4957184 9696
7 187936 5053245 5053243 40512
8 228448 5093758 5093756 2048
9 230496 3900207 5095805 44449
10 274945 3958348 3944655 18047
11 292992 4007289 3976394 29440
12 322432 4037177 4036728 18880
13 341312 4003250 4056056 3232
14 344544 3977514 4006481 6176
15 350720 3976396 3983689 320
16 351040 3977502 3976715 1
17 351041 3976717 3977502 671
18 351712 3983690 3977387 193
19 351905 3977388 3983882 31
20 351936 3983883 3977418 1248
21 353184 3985132 3985130 960
22 354144 3991493 3986091 1
23 354145 3986093 3991493 1823
24 355968 3991494 3987915 5184
25 361152 3987917 3996677 2112
26 363264 3977419 3990028 32
27 363296 3977503 3977450 1
28 363297 4056058 3977503 13343
29 376640 4467752 4069400 21184
30 397824 4520448 4488935 10016
31 407840 4538961 4530463 22400
32 430240 4594688 4561360 12672
33 442912 4638793 4607359 22432
34 465344 4661226 4661224 5632
35 470976 4703499 4666857 44160
36 515136 4797420 4747658 15968
37 531104 5118976 4813387 41280
38 572384 5160257 5160255 20768
39 593152 5181026 5181024 66656
40 659808 5310731 5247681 1
41 659809 5283659 5310731 17855
42 677664 5301515 5301513 9216
43 686880 5322247 5310730 58432
44 745312 5381120 5380678 48768
45 794080 5429889 5429887 22059 eof
fast: 46 extents found
BtrFS slow allocation:
File size of slow is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 2294997 1
1 1 2255636 2294997 463
2 464 2256100 2256098 128
3 592 2257205 2256227 1
4 593 2256229 2257205 367
5 960 2257206 2256595 1
6 961 2256597 2257206 319
7 1280 2257207 2256915 1
8 1281 2256917 2257207 287
9 1568 2257208 2257203 64
10 1632 2273468 2257271 1
11 1633 2257273 2273468 335
12 1968 2273469 2257607 1
13 1969 2257609 2273469 287
14 2256 2273470 2257895 1
15 2257 2257897 2273470 335
16 2592 2296777 2258231 1
17 2593 2258233 2296777 287
18 2880 2296778 2258519 1
19 2881 2258521 2296778 287
20 3168 2273289 2258807 1
21 3169 2258809 2273289 287
22 3456 2273290 2259095 1
23 3457 2259097 2273290 287
24 3744 2273291 2259383 1
25 3745 2259385 2273291 287
26 4032 2273292 2259671 1
27 4033 2259673 2273292 303
28 4336 2273293 2259975 1
29 4337 2259977 2273293 303
30 4640 2273294 2260279 1
31 4641 2260281 2273294 287
32 4928 2273295 2260567 1
33 4929 2260569 2273295 271
34 5200 2273296 2260839 1
35 5201 2260841 2273296 303
36 5504 2273297 2261143 1
37 5505 2261145 2273297 255
38 5760 2273298 2261399 1
39 5761 2261401 2273298 255
40 6016 2273299 2261655 1
41 6017 2261657 2273299 271
42 6288 2273300 2261927 1
43 6289 2261929 2273300 287
44 6576 2273301 2262215 1
45 6577 2262217 2273301 255
46 6832 2273302 2262471 1
47 6833 2262473 2273302 271
48 7104 2273303 2262743 1
49 7105 2262745 2273303 287
50 7392 2273304 2263031 1
51 7393 2263033 2273304 271
52 7664 2273305 2263303 1
53 7665 2263305 2273305 287
54 7952 2273306 2263591 1
55 7953 2263593 2273306 271
56 8224 2273307 2263863 1
57 8225 2263865 2273307 271
58 8496 2273308 2264135 1
59 8497 2264137 2273308 287
60 8784 2273309 2264423 1
61 8785 2264425 2273309 255
62 9040 2273310 2264679 1
63 9041 2264681 2273310 287
64 9328 2273311 2264967 1
65 9329 2264969 2273311 303
66 9632 2273312 2265271 1
67 9633 2265273 2273312 271
68 9904 2273313 2265543 1
69 9905 2265545 2273313 271
70 10176 2273314 2265815 1
71 10177 2265817 2273314 287
72 10464 2273315 2266103 1
73 10465 2266105 2273315 271
74 10736 2273316 2266375 1
75 10737 2266377 2273316 271
76 11008 2273317 2266647 1
77 11009 2266649 2273317 271
78 11280 2273318 2266919 1
79 11281 2266921 2273318 271
80 11552 2273319 2267191 1
81 11553 2267193 2273319 271
82 11824 2273320 2267463 1
83 11825 2267465 2273320 287
84 12112 2273321 2267751 1
85 12113 2267753 2273321 287
86 12400 2273322 2268039 1
87 12401 2268041 2273322 303
88 12704 2273323 2268343 1
89 12705 2268345 2273323 271
90 12976 2273324 2268615 1
91 12977 2268617 2273324 223
92 13200 2273325 2268839 1
93 13201 2268841 2273325 255
94 13456 2273326 2269095 1
95 13457 2269097 2273326 255
96 13712 2273327 2269351 1
97 13713 2269353 2273327 271
98 13984 2273328 2269623 1
99 13985 2269625 2273328 287
100 14272 2273329 2269911 1
101 14273 2269913 2273329 271
102 14544 2273330 2270183 1
103 14545 2270185 2273330 271
104 14816 2273331 2270455 1
105 14817 2270457 2273331 319
106 15136 2273332 2270775 1
107 15137 2270777 2273332 271
108 15408 2273333 2271047 1
109 15409 2271049 2273333 271
110 15680 2273334 2271319 1
111 15681 2271321 2273334 287
112 15968 2273335 2271607 1
113 15969 2271609 2273335 287
114 16256 2273336 2271895 1
115 16257 2271897 2273336 239
116 16496 2273337 2272135 1
117 16497 2272137 2273337 303
118 16800 2273338 2272439 1
119 16801 2272441 2273338 287
120 17088 2273339 2272727 1
121 17089 2272729 2273339 287
122 17376 2273340 2273015 1
123 17377 2273017 2273340 271
124 17648 2273341 2273287 1
125 17649 2273471 2273341 287
126 17936 2273342 2273757 1
127 17937 2273759 2273342 303
128 18240 2273343 2274061 1
129 18241 2274063 2273343 271
130 18512 2273344 2274333 1
131 18513 2274335 2273344 255
132 18768 2274591 2274589 160
133 18928 2274752 2274750 144
134 19072 2273345 2274895 1
135 19073 2274897 2273345 319
136 19392 2273346 2275215 1
137 19393 2275217 2273346 287
138 19680 2273347 2275503 1
139 19681 2275505 2273347 303
140 19984 2273348 2275807 1
141 19985 2275809 2273348 271
142 20256 2273349 2276079 1
143 20257 2276081 2273349 287
144 20544 2273350 2276367 1
145 20545 2276369 2273350 287
146 20832 2273351 2276655 1
147 20833 2276657 2273351 271
148 21104 2273352 2276927 1
149 21105 2276929 2273352 303
150 21408 2273353 2277231 1
151 21409 2277233 2273353 271
152 21680 2273354 2277503 1
153 21681 2277505 2273354 255
154 21936 2273355 2277759 1
155 21937 2277761 2273355 255
156 22192 2273356 2278015 1
157 22193 2278017 2273356 431
158 22624 2273357 2278447 16
159 22640 2273374 2273372 1
160 22641 2278449 2273374 207
161 22848 2273375 2278655 1
162 22849 2278657 2273375 1023
163 23872 2279681 2279679 352
164 24224 2273376 2280032 48
165 24272 2273425 2273423 1
166 24273 2280034 2273425 303
167 24576 2273426 2280336 1
168 24577 2280338 2273426 271
169 24848 2273427 2280608 1
170 24849 2280610 2273427 255
171 25104 2273428 2280864 1
172 25105 2280866 2273428 271
173 25376 2273429 2281136 1
174 25377 2281138 2273429 271
175 25648 2273430 2281408 1
176 25649 2281410 2273430 287
177 25936 2273431 2281696 1
178 25937 2281698 2273431 271
179 26208 2273432 2281968 1
180 26209 2281970 2273432 287
181 26496 2273433 2282256 1
182 26497 2282258 2273433 639
183 27136 2273434 2282896 1
184 27137 2282898 2273434 287
185 27424 2273435 2283184 1
186 27425 2283186 2273435 303
187 27728 2273436 2283488 1
188 27729 2283490 2273436 255
189 27984 2273437 2283744 1
190 27985 2283746 2273437 223
191 28208 2273438 2283968 1
192 28209 2283970 2273438 175
193 28384 2273439 2284144 1
194 28385 2284146 2273439 191
195 28576 2273440 2284336 1
196 28577 2284338 2273440 191
197 28768 2273441 2284528 1
198 28769 2284530 2273441 207
199 28976 2273442 2284736 1
200 28977 2284738 2273442 175
201 29152 2273443 2284912 16
202 29168 2284914 2273458 32
203 29200 2284947 2284945 48
204 29248 2273460 2284994 1
205 29249 2284996 2273460 191
206 29440 2273461 2285186 1
207 29441 2285188 2273461 255
208 29696 2273462 2285442 1
209 29697 2285444 2273462 255
210 29952 2273463 2285698 1
211 29953 2285700 2273463 223
212 30176 2273464 2285922 1
213 30177 2285924 2273464 255
214 30432 2273465 2286178 1
215 30433 2286180 2273465 255
216 30688 2273466 2286434 1
217 30689 2286436 2273466 3279
218 33968 2289716 2289714 5280
219 39248 2273467 2294995 1
220 39249 2296779 2273467 4207
221 43456 2300987 2300985 3664
222 47120 2317067 2304650 1
223 47121 2304652 2317067 3775
224 50896 2308428 2308426 4032
225 54928 2317068 2312459 1
226 54929 2312461 2317068 3903
227 58832 2317070 2316363 4112
228 62944 2316365 2321181 1
229 62945 2321183 2316365 4127
230 67072 2325311 2325309 3984
231 71056 2316366 2329294 1
232 71057 2329296 2316366 3983
233 75040 2333280 2333278 3840
234 78880 2316367 2337119 1
235 78881 2337121 2316367 3919
236 82800 2344050 2341039 4032
237 86832 2363669 2348081 1
238 86833 2348083 2363669 3791
239 90624 2351875 2351873 3872
240 94496 2341041 2355746 1
241 94497 2355748 2341041 3839
242 98336 2363670 2359586 4160
243 102496 2341042 2367829 1
244 102497 2367831 2341042 4271
245 106768 2372103 2372101 4352
246 111120 2380662 2376454 1
247 111121 2376456 2380662 3887
248 115008 2380665 2380342 3984
249 118992 2397106 2384648 1
250 118993 2384650 2397106 3711
251 122704 2388362 2388360 3984
252 126688 2397107 2392345 1
253 126689 2392347 2397107 3791
254 130480 2397110 2396137 3840
255 134320 2380344 2400949 1
256 134321 2400951 2380344 3999
257 138320 2406842 2404949 4016
258 142336 2419383 2410857 1
259 142337 2410859 2419383 4031
260 146368 2414891 2414889 3888
261 150256 2419384 2418778 1
262 150257 2419386 2419384 3823
263 154080 2423210 2423208 4032
264 158112 2404951 2427241 1
265 158113 2427243 2404951 3903
266 162016 2431147 2431145 4016
267 166032 2404952 2435162 1
268 166033 2438119 2404952 3615
269 169648 2441735 2441733 4160
270 173808 2453941 2445894 1
271 173809 2445896 2453941 4063
272 177872 2449960 2449958 3728
273 181600 2458915 2453687 1
274 181601 2453942 2458915 4031
275 185632 2458918 2457972 3776
276 189408 2435164 2462693 1
277 189409 2462695 2435164 3887
278 193296 2469315 2466581 3952
279 197248 2495798 2473266 1
280 197249 2473268 2495798 4063
281 201312 2477332 2477330 4048
282 205360 2466583 2481379 1
283 205361 2481381 2466583 3903
284 209264 2485285 2485283 3920
285 213184 2466584 2489204 1
286 213185 2489206 2466584 3791
287 216976 2497536 2492996 4000
288 220976 2505409 2501535 1
289 220977 2501537 2505409 3871
290 224848 2505410 2505407 4000
291 228848 2513283 2509409 1
292 228849 2509411 2513283 3871
293 232720 2513284 2513281 3696
294 236416 2520885 2516979 1
295 236417 2516981 2520885 3903
296 240320 2520886 2520883 3680
297 244000 2528423 2524565 1
298 244001 2524567 2528423 3855
299 247856 2530304 2528421 3920
300 251776 2538113 2534223 1
301 251777 2534225 2538113 3887
302 255664 2538114 2538111 4096
303 259760 2546243 2542209 1
304 259761 2542211 2546243 4031
305 263792 2546244 2546241 3712
306 267504 2554325 2549955 1
307 267505 2549957 2554325 4367
308 271872 2554326 2554323 3792
309 275664 2562023 2558117 1
310 275665 2558119 2562023 3903
311 279568 2563072 2562021 3808
312 283376 2570737 2566879 1
313 283377 2566881 2570737 3855
314 287232 2570738 2570735 3952
315 291184 2578563 2574689 1
316 291185 2574691 2578563 3871
317 295056 2578564 2578561 3984
318 299040 2586405 2582547 1
319 299041 2582549 2586405 3855
320 302896 2586406 2586403 3904
321 306800 2594391 2590309 1
322 306801 2590311 2594391 4079
323 310880 2595840 2594389 3760
324 314640 2603457 2599599 1
325 314641 2599601 2603457 3855
326 318496 2603458 2603455 4016
327 322512 2611427 2607473 1
328 322513 2607475 2611427 3951
329 326464 2611428 2611425 3888
330 330352 2619189 2615315 1
331 330353 2615317 2619189 3871
332 334224 2619190 2619187 4096
333 338320 2623287 2623285 1
334 338321 2359588 2623287 3775
335 342096 2623288 2363362 3936
336 346032 2627225 2627223 1
337 346033 2628608 2627225 3871
338 349904 2632480 2632478 3888
339 353792 2640065 2636367 1
340 353793 2636369 2640065 3695
341 357488 2640066 2640063 3808
342 361296 2648051 2643873 1
343 361297 2643875 2648051 4175
344 365472 2648052 2648049 3728
345 369200 2655845 2651779 1
346 369201 2651781 2655845 4063
347 373264 2655846 2655843 4624
348 377888 2660471 2660469 1
349 377889 2661376 2660471 3839
350 381728 2665216 2665214 3840
351 385568 2672945 2669055 1
352 385569 2669057 2672945 3887
353 389456 2672946 2672943 3984
354 393440 2681411 2676929 1
355 393441 2676931 2681411 4479
356 397920 2681412 2681409 4016
357 401936 2689317 2685427 1
358 401937 2685429 2689317 3887
359 405824 2689318 2689315 4128
360 409952 2693447 2693445 1
361 409953 2694144 2693447 3903
362 413856 2698048 2698046 3968
363 417824 2706529 2702015 1
364 417825 2702017 2706529 4511
365 422336 2706530 2706527 3952
366 426288 2714195 2710481 1
367 426289 2710483 2714195 3711
368 430000 2714196 2714193 3824
369 433824 2721893 2718019 1
370 433825 2718021 2721893 3871
371 437696 2721894 2721891 3984
372 441680 2725879 2725877 1
373 441681 2726912 2725879 3951
374 445632 2730864 2730862 4176
375 449808 2739105 2735039 1
376 449809 2735041 2739105 4063
377 453872 2739106 2739103 3520
378 457392 2746355 2742625 1
379 457393 2742627 2746355 3727
380 461120 2750118 2746353 4240
381 465360 2758631 2754357 1
382 465361 2754359 2758631 4271
383 469632 2759680 2758629 3920
384 473552 2768065 2763599 1
385 473553 2763601 2768065 4463
386 478016 2768066 2768063 4560
387 482576 2772627 2772625 1
388 482577 2746356 2772627 3743
389 486320 2772628 2750098 3840
390 490160 2780437 2776467 1
391 490161 2776469 2780437 3967
392 494128 2780438 2780435 4000
393 498128 2788599 2784437 1
394 498129 2784439 2788599 4159
395 502288 2788600 2788597 3968
396 506256 2795033 2792567 1
397 506257 2810855 2795033 3631
398 509888 2814487 2814485 3280
399 513168 2820952 2817766 1
400 513169 2817768 2820952 3183
401 516352 2820953 2820950 3840
402 520192 2824794 2824792 1
403 520193 2796313 2824794 4175
404 524368 2800489 2800487 3744
405 528112 2824795 2804232 1
406 528113 2804234 2824795 3487
407 531600 2825584 2807720 4208
408 535808 2839729 2829791 1
409 535809 2829793 2839729 3839
410 539648 2833633 2833631 3824
411 543472 2839730 2837456 1
412 543473 2842425 2839730 3775
413 547248 2846201 2846199 3792
414 551040 2837458 2849992 1
415 551041 2849994 2837458 3999
416 555040 2853994 2853992 4128
417 559168 2837459 2858121 1
418 559169 2858123 2837459 3871
419 563040 2867391 2861993 3936
420 566976 2880529 2871326 1
421 566977 2871328 2880529 3855
422 570832 2875184 2875182 3968
423 574800 2880530 2879151 1
424 574801 2861995 2880530 3807
425 578608 2881987 2865801 4032
426 582640 2865803 2886018 1
427 582641 2887116 2865803 3807
428 586448 2890924 2890922 4128
429 590576 2918872 2895051 1
430 590577 2895053 2918872 3935
431 594512 2899516 2898987 3904
432 598416 2918873 2903419 1
433 598417 2903421 2918873 3839
434 602256 2907261 2907259 3872
435 606128 2918874 2911132 1
436 606129 2911134 2918874 4031
437 610160 2920347 2915164 4048
438 614208 2898989 2924394 1
439 614209 2926197 2898989 4319
440 618528 2930517 2930515 3728
441 622256 2944908 2934244 1
442 622257 2934246 2944908 3727
443 625984 2937974 2937972 4080
444 630064 2944909 2942053 1
445 630065 2946430 2944909 3791
446 633856 2950222 2950220 3808
447 637664 2924396 2954029 1
448 637665 2979066 2924396 3983
449 641648 2959587 2983048 4144
450 645792 2954031 2963730 1
451 645793 2963732 2954031 4255
452 650048 2967988 2967986 4096
453 654144 2983050 2972083 1
454 654145 2954032 2983050 3663
455 657808 2972085 2957694 3776
456 661584 2983051 2975860 1
457 661585 2985732 2983051 3855
458 665440 3008302 2989586 4112
459 669552 3016431 3012413 1
460 669553 3012415 3016431 4015
461 673568 2991465 3016429 4112
462 677680 3000706 2995576 1
463 677681 2995578 3000706 4207
464 681888 3000709 2999784 3872
465 685760 3016432 3004580 1
466 685761 3021824 3016432 3983
467 689744 3025808 3025806 3984
468 693728 3033745 3029791 1
469 693729 3029793 3033745 3951
470 697680 3033746 3033743 4000
471 701680 3041827 3037745 1
472 701681 3037747 3041827 4079
473 705760 3041828 3041825 4032
474 709792 3050005 3045859 1
475 709793 3045861 3050005 4143
476 713936 3050006 3050003 3968
477 717904 3053975 3053973 1
478 717905 3054592 3053975 3871
479 721776 3058464 3058462 4112
480 725888 3066529 3062575 1
481 725889 3062577 3066529 3951
482 729840 3066530 3066527 4112
483 733952 3074579 3070641 1
484 733953 3070643 3074579 3935
485 737888 3074580 3074577 3808
486 741696 3082437 3078387 1
487 741697 3078389 3082437 4047
488 745744 3082438 3082435 3696
489 749440 3086135 3086133 1
490 749441 3087360 3086135 3679
491 753120 3093064 3091038 3664
492 756784 3100697 3096727 1
493 756785 3096729 3100697 3967
494 760752 3103733 3100695 4128
495 764880 3093055 3107860 1
496 764881 3107862 3093055 3903
497 768784 3111766 3111764 3648
498 772432 3093056 3115413 1
499 772433 3115415 3093056 3791
500 776224 3119207 3119205 3680
501 779904 3093057 3122886 1
502 779905 3123607 3093057 4031
503 783936 3127639 3127637 4064
504 788000 3131704 3131702 1
505 788001 3140351 3131704 3215
506 791216 3143567 3143565 3968
507 795184 3151472 3147534 1
508 795185 3147536 3151472 3935
509 799120 3132254 3151470 3888
510 803008 3151473 3136141 1
511 803009 3136143 3151473 3967
512 806976 3152896 3140109 4400
513 811376 3158465 3157295 1
514 811377 3157297 3158465 1167
515 812544 3158466 3158463 224
516 812768 3160976 3158689 1
517 812769 3158691 3160976 239
518 813008 3160977 3158929 1
519 813009 3158931 3160977 223
520 813232 3160978 3159153 1
521 813233 3159155 3160978 271
522 813504 3160979 3159425 1
523 813505 3159427 3160979 239
524 813744 3160980 3159665 1
525 813745 3159667 3160980 223
526 813968 3160981 3159889 1
527 813969 3159891 3160981 1084
528 815053 3160982 3160974 1072
529 816125 3162068 3162053 1
530 816126 3162055 3162068 13 eof
slow: 531 extents found
Ext4 copy allocation:
File size of copy.7z is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 374784 18432
1 18432 395264 393215 30720
2 49152 428032 425983 30720
3 79872 460800 458751 30720
4 110592 493568 491519 30720
5 141312 565248 524287 24576
6 165888 591872 589823 30720
7 196608 624640 622591 30720
8 227328 657408 655359 30720
9 258048 690176 688127 30720
10 288768 722944 720895 30720
11 319488 755712 753663 30720
12 350208 788480 786431 30720
13 380928 821248 819199 30720
14 411648 854016 851967 30720
15 442368 886784 884735 30720
16 473088 919552 917503 30720
17 503808 952320 950271 30720
18 534528 985088 983039 30720
19 565248 1017856 1015807 30720
20 595968 1083392 1048575 30720
21 626688 1116160 1114111 30720
22 657408 1148928 1146879 30720
23 688128 1181696 1179647 30720
24 718848 1214464 1212415 30720
25 749568 1247232 1245183 30720
26 780288 1280000 1277951 30720
27 811008 1312768 1310719 4096
28 815104 1311233 1316863 1035 eof
copy.7z: 29 extents found
Ext4 fast allocation:
File size of fast is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 2230272 30720
1 30720 2263040 2260991 30720
2 61440 2295808 2293759 30720
3 92160 2328576 2326527 30720
4 122880 2361344 2359295 30720
5 153600 2394112 2392063 30720
6 184320 2426880 2424831 30720
7 215040 2459648 2457599 30720
8 245760 2492416 2490367 30720
9 276480 2525184 2523135 30720
10 307200 2557952 2555903 30720
11 337920 2590720 2588671 28672
12 366592 253952 2619391 2048
13 368640 258048 255999 2048
14 370688 280576 260095 2048
15 372736 286720 282623 2048
16 374784 270336 288767 4096
17 378880 290816 274431 4096
18 382976 301056 294911 2048
19 385024 313344 303103 2048
20 387072 321536 315391 4096
21 391168 374784 325631 18432
22 409600 395264 393215 30720
23 440320 428032 425983 30720
24 471040 460800 458751 30720
25 501760 493568 491519 30720
26 532480 565248 524287 24576
27 557056 591872 589823 30720
28 587776 624640 622591 30720
29 618496 657408 655359 30720
30 649216 690176 688127 30720
31 679936 722944 720895 30720
32 710656 755712 753663 30720
33 741376 788480 786431 30720
34 772096 821248 819199 30720
35 802816 854016 851967 24576
36 827392 852481 878591 45 eof
fast: 37 extents found
Ext4 slow allocation:
File size of slow is 3342903522 (816139 blocks, blocksize 4096)
ext logical physical expected length flags
0 0 374784 18432
1 18432 395264 393215 30720
2 49152 428032 425983 30720
3 79872 460800 458751 30720
4 110592 493568 491519 30720
5 141312 565248 524287 24576
6 165888 591872 589823 30720
7 196608 624640 622591 30720
8 227328 657408 655359 30720
9 258048 690176 688127 30720
10 288768 722944 720895 30720
11 319488 755712 753663 30720
12 350208 788480 786431 30720
13 380928 821248 819199 30720
14 411648 854016 851967 30720
15 442368 886784 884735 30720
16 473088 919552 917503 30720
17 503808 952320 950271 30720
18 534528 985088 983039 30720
19 565248 1017856 1015807 30720
20 595968 1083392 1048575 30720
21 626688 1116160 1114111 30720
22 657408 1148928 1146879 30720
23 688128 1181696 1179647 30720
24 718848 1214464 1212415 30720
25 749568 1247232 1245183 30720
26 780288 1280000 1277951 30720
27 811008 1312768 1310719 16384
28 827392 1311233 1329151 116 eof
slow: 29 extents found
NTFS, incremental allocation:
File is 816139 physical clusters in length.
File is in 5063 fragment(s).
NTFS, copy with pre-allocation:
File is 816139 physical clusters in length.
File is in 1 fragment(s).
Using pre-allocation with NTFS file system is very important. It should be used when ever it is possible (when final file size is known). Unfortunately many applications aren't using this important feature.
Linux kernel used: 2.6.38-8-server
|
posted Feb 12, 2012 8:22 AM by Sami Lehtinen
[
updated Feb 12, 2012 8:23 AM
]
- Tested and installed RFID end client identification system.
- Tested and compared BtrFS none / lzo / zlib compression algorithms and disk performance on virtual servers.
- e4defrag is finally becoming ready, just when BtrFS is becoming slowly ready for production use? BtrFS provides autodefrag mount option.
- Closed case with Clonzeilla and TFTPDS32 DHCP protocol issue. It's now clear that Clonezilla sends invalid request packets and are being ignored by DHCP server application. To find out all the details, tried both Wireshark and (MS) Network Monitor.
- There is also strange issue with Clonezilla, Linux, pxelinux network driver. When using one Realtek Gigabit Ethernet adapter with pxelinux, it seems that something goes so badly wrong that network card is "permanently disabled". Even booting computer to Windows, where it did work before this boot attempt doesn't fix the situation. Some engineers thought that it's better to send those computers for warranty repair, but before they were able to send those. I told them that have they really rebooted those computers? They said yes. Then I unplugged the power cord, pressed power button a few times, and plugged cable back in. - Yup, now you got working network adapter again. No, it's not real cold reboot, if you keep motherboard powered all the time. Unfortunately even latest pxelinux.0 file won't fix this issue. Yes, I have reported it.
- Compiled and installed obfsproxy to virtual server. I think information flow should be free and available to everyone. There are now over 20 users using it all the time.
- Created new (git) branch for 9oxnet project. Plan is to make required changes, so I can use python27 runtime with Google App Engine. After very quick testing I found out that upgrading from 2.5 runtime to 2.7 requires more changes than I initially thought. I also need to confirm that application is truly threadsafe, otherwise interesting issues could arise.
- I had to produce one report, there were multiple options how to collect data for it. But I decided that it's not bad idea to be able to fully master pivot queries, so I used a few hours to learn it throught and managed to pull all required information with one query including all required multitable joins etc.
- Listened just a few too many Social Engineering podcasts, those weren't that good. Won't listen more.
- Checked what are the actual fragmenting effects of sparsely allocated files on ext4 filesystem. It seems that even if large file is created as sparse file, and filled up in fully random order, ext4 manages to create in average 2 megabyte extents. Files are of course fragment up between extents, but that's not excessive. Without extent feature, results could be very much worse. Like they are when doing similar test is executed using NTFS file system. It leads to incredibly many fragments, just a few contiguous blocks here and there. If you want even worse results, fill up compressed NTFS file in random order. It seems that 7-zip doesn't pre-allocate files even when extracting data. It would be smart to pre-allocate space for file because final file size is of naturally known.
- During doing that I found out that there is currently failure in Ubuntu power management. It sets ext4 commit time always to 0 if AC power is being used. I had to modify /usr/lib/pm-utils/power.d/journal-commit file to get 60 second commit time. That's what though I would have been using all the time. When I checked mounts I found out that there was commit=60,commit=0 entries on ext4 mount rows. Unfortunately I haven't been able to enable NCQ (and AHCI) with my Intel P35 Express motherboard and Linux kernel. I have tried all basic tricks.
- Checked out (again) latest info about SSD drives and issues realated to those: file systems designed for SSD drives (log filesystem), flash file system, wear leveling, write amplification, data remanence, trim. Low level wear leveling can caused serious drive fragmentation which isn't visible at all to operating system. Performance of some Intel drives suffer very badly. Running OS file system level defrag only makes this situation often worse. (ubifs, logfs, jffs2, ahci, ncq, trim, lba)
|
posted Feb 5, 2012 3:31 AM by Sami Lehtinen
[
updated Feb 5, 2012 4:27 AM
]
- WPS is absolutely broken by bad design. It's very important to disable WPS completely from your WLAN / Wi-Fi router / station. Check out Reaver cracking tool. PIN and Push Button features are highly vulnerable, future NFC or USB based modes could be better.
- APE Payment: They instruct users to globally enable all 3rd party cookies. Just to allow use of their services. Why? After quick testing at least Firefox easily allows to enable 3rd party cookies just for one specified domain. Which is enough in this case. I IE worked just fine too. But with Google Chrome there was strange thing. It was possible to block 3rd party cookies from certain domains, but there wasn't option to allow 3rd party cookies only from specified domain(s). That's quite a problem in this case. So 3rd party cookies must be globally enabled to use APE Payment.
- Nordnet Bank doesn't use 2FA/SMS authentication. Their system is already capable of sending SMS messages and they could even charge for authentication messages. But they still refuse to do this. It's interesting to see how banking sector handles security suggestions from customers. They're hadling large sums of money, because they're major nordic stock broker. But customer security isn't that important, right?
- Updated to latest Google App Engine SDK 1.6.2 and tested that my projects work perfectly with it. I still should upgrade from Python 2.5 runtime to Python 2.7, but might be too early.
- Checked latest file system develoment. What are the problems with larger ext4 and XFS file systems and how those issues can be solved.
- I'm often checking routing, AS numbers and stuff. Nice and easy to use tools for anyone are route.im and robtex.com. But this TPB trace is really strange.
- Tried HAProxy, it's very exellent load balancing tool.
- Checked out iLOQ S10 locking system. I got installation offer from them. It seems that this solution got hardware design flaw, that can be used by attacker using several different exploits. Unfortunately I didn't have time to check out the digital part of this lock. Keys are using DS2432 1Kb 1-Wire chips with SHA512 engine. I'm pretty sure that electric part of the lock could be also exploitable, but its propably much more complex and time consuming than those simple physical attacks. I personally would like to try those physical attacks, because those seem to pretty simple to do.
- Installed port knocking software knockd to my private communication server.
Of course I also did lots of less interesting stuff.
|
posted Jan 29, 2012 8:38 AM by Sami Lehtinen
[
updated Jan 31, 2012 7:51 AM
]
I have been glued to computer and my mobile terminal. And I have managed to to following stuff during this week.
- Refresing memory and catching up with latest development around: mimo, ofdm, ssb, qam, lte, 3g, 4g, hspa+, umts, dual carrier (dual cell)
- Read about possibly upcoming new EU privacy / data protection law. It might be very hard hit for small web sites. Would anyone really expect those to follow this kind of legalization?
- Checked out Google's new privacy policy & terms of service.
- I have always loved optimizing TCP protocol by tweaking parameters. Now there is nice article about making Internet faster by tweaking TCP functionality and parameters.
- Tried out UpCloud, new cloud service. Details of the service are not yet public, therefore I can't tell too much about it. But let's hope service is soon available for everyone.
- Quickly tested load testing service LoadStorm.
- Lots of discussion in LinkedIn IT Security, Startup, SaaS, Cloud Computing groups.
- Read tons of news, and magazines about tech stuff, developers tools and startups.
|
posted Jan 29, 2012 3:51 AM by Sami Lehtinen
[
updated Feb 6, 2012 9:00 AM
]
Here's my little rant about Yahoo! Mail. It's clearly not a good choice as your primary email provider. - There is daily message quota, when you have sent enough messages, it says that you can send more messages tomorrow. Limit is quite low. Because I have set it often when having my "email day", when I take care of all my queued mails. I assume that limit is something like ~200 mails. +/- 50 mails. Maybe number of recipients also affects it. But limit is very annoying for active email user.When daily send quota is full, it seems that some messages just disappear and aren't left in outbox to be sent tomorrow. This is really really bad thing.
- When receiving larger quantities of email, service starts blocking mail traffic.
Diagnostic-Code: X-Postfix; delivery temporarily suspended: host
mx1.mail.eu.yahoo.com[77.238.177.9] refused to talk to me: 421 4.7.0 [TS02]
Messages from 205.251.134.191 temporarily deferred - 4.16.56.1; see - IMAP+ (imap/idle) often times out or doesn't respond to ping. Connecting
simply dies. Simple tasks like moving message to another folder, can
take several minutes.
- SMTP servers do not allow parallel SMTP sessions. Making delivering larger email batches slow.
- Asks way too often to solve CAPTCHA code, if email contains link(s).
- Several encryption / security issues:
- Web interface doesn't use HTTPS (after password is given) or IP-address
change checkes. Stealing session cookie is absolutely possible and
using it from other IP allowing session hijacking attack.
- Yahoo's SMTPS server and IMAP servers got at times invalid TLS certificates. If I
can't verify certificate everytime, it means that I have to accept
invalid certificates, making MITM attack very easy. This kind of MITM attack allows easy way to steal passwords when travelling or using unknown WLAN.
- MTA's do not use TLS at all. So all email is actually being transported in clear for inbound and outbound traffic.
- No 2FA support.
- Even if I have configured my Yahoo sessions to expire in 24 hours, my mobile sessions last longer. This is quite interesting. It shoulnd't be unclear that 24 hour session should last max 24 hours.
Are there any positive things? - Yep, a few. - Web interface is pretty much working one.
- European services are run from European based servers.
- Temporary email addresses, I just love those. Of course you can get similar service from TrashMail or SpamGourmet. But using these services adds proxies, complexity, unreliability to incoming and outgoing email handling.
Based on these issues, I wouldn't recommend Yahoo Mail as email provider for anyone. Except if you use it for secondary purposes, and do not transfer any private information using it. |
posted Jan 22, 2012 10:26 AM by Sami Lehtinen
[
updated Jan 22, 2012 10:28 AM
]
I have been studying every day, several hours a day. Study subjects like: - Encryption & Hashing: Threefish, OCB mode, Skein, Jenkins, Pearson - Nice article about hashing
- Compression: LZ77, LZMA2, LZMA, Huffman, Burrows Wheeler
- File system: Refreshed my memory about BtrFS and it's latest development
- Cloud hosting: IBM SmartCloud, I also tried Nodester - Didn't like it, beacause I'm not familiar enough with JavaScript
- Web casts: Risky Business and plenty of stuff from TED
- Networking: DCCP, ECN, NAT66 / NPT66, shim6, IPv6 multihoming using NAT without BGP.
World IPv6 launch is here!
IPv6 + NAT, please no! NAT breaks Internet connectivity!
SCTP: At least for my apps SCTP would be far superior compared to TCP and also would make communication faster. Packets also can be received out of order if order doesn't matter further saving bandwidth. If ECN (removes packet loss) is used with SCTP it's real killer solution compared to plain old TCP. Less data to be sent faster and more reliable delivery. SCTP keeps message boundaries clear and allows multiple parallel streams with one negotiation. TCP's sliding window adds overhead by delivering partitial messages. Also SCTP uses CRC32 instead of CRC16. I have experienced issues with bad links and huge files while using TCP, received file / data is simply corrupted. Not completely, but even one corrupted block is enough to completely destroy compressed file.
- Reading: Good to know about data security and Google. Good to know information in easy to understand package. - I really recommend reading this.
- Other stuff: Squid cache eviction algorithms and configuration parameters, more SaaS and Virtualization stuff, Magnet URI scheme
- Writings: I wrote long article about two factor authentication (2FA) and practical ways to implement it. I recommended that the authorization key should be directly related to the transaction being confirmed. Using OTP passwords without payload parameters is just bad design. In those cases user does not know what they're actually confirming. Any MITM or Trojan could modify messages in real-time. So tokens aren't good solution, SMS messages with enough information to confirm transaction and single use code for that is pretty good. But we shouldn't forget that there are many technical and not so technical ways to attack that implementation too. GSM/2G networks aren't so secure, if attacker got high value target. Also phone software could be modified and can be rogue applications or whole operating system can be replaced to further deceive user.
I couldn't list everything, but that's small part of stuff I have done. There is also long rant coming about Yahoo Mail, listing all the dirty facts. |
posted Jan 21, 2012 11:04 PM by Sami Lehtinen
[
updated Jan 21, 2012 11:31 PM
]
Today some protocol discussion in one forum reminded my self about old good times as SysOp of BBS system. Systems and protocols like: PCBoard, FidoCom, Rcom, Opus, SBBS, QBBS and file transfer protocols like Kermit, Xmodem, Ymodem, Zmodem, Smodem, HS/Link, BiModem and Hydracom came right back to my mind. Also with all their design issues. BBS systemsPCBoard used index and message data files per board, it was pretty good design. FidoCom and Rcom stored each message in separate file, which was really bad thing for larger systems because of fat 16 limited number of files on disk. This also made collecting messages for QWK or BlueWave off-line message reader formats very slow. SBBS and QBBS used one message database file for all messages. It caused huge problems when systems were connected to global fidonet, which replicated messages to systems all over the world. I assume that the designers of the system never realized how many messages there actually could be and therefore they used 16 bit message id. Maximum number of messages in database could be only 65536 messages. That caused message backlog to be really short for very active systems. Of course that was huge number of messages if you thought to have local system, with only one phoneline and single user at time. That is exactly how times change. When you design something, design it so that it won't run out. Eh, like IPv4. Benefits of having all messages in single file with index was that picking required messages for off-line reading was really super fast compared to other designs I mentioned ealier which required opening large number of individual files instead of using fast seeks in file. File transfer protocolsHS/Link was near perfect with bidirectional transfers and full streaming. Except it didn not use adaptive block size. This critical issue caused big block size to discard almost all data on noisy line. Of course smaller block size could have been chosen when starting transfer. But if line was good using small block size was really inefficient. So lack of automatic adaption to line conditions made this protocol unusable at least for me. Zmodem was nice, but it didn't have fully streaming error correction (causing every error to stop transfer and discard lots of perfectly good data.) BiModem just wasn't supported by any systems as well as Hydracom was poorly supported. But then came Smodem which was simply great. With adaptive blocksize and multiplexed simple ack/nak channels it was clearly superior to any other protocol implemented that far. Other older protocols like Xmodem were inefficient and also slow, no sliding window, simple single ack/nak syled transfer causing latency to cause wasted idle time. Smodem was actually so extremely good compared to competition that I was really bit sad that we didn't have it years before. Modem protocolsOther protocols which were integrated to modem software directly came very familiar. v.42 error correction, v.42bis data compression and older MNP4 (EC) and MNP5 (DC). Not forgetting SREJ (Selective Reject) which allowed only corrupted frames to be resent. Working efficiently and very similarly (as end result) to Smodem, although technical approach and implementation was very different. Some terminal programs like Transend were able to implement MNP4 and MNP5 error correction and compression in terminal software, there fore you didn't need more expensive modem having those features. (1200 bps / 2400 bps era) Terminal emulationVT102 and ANSI were ruling terminal protocols for long time. But after that came Avatar, it was drastically faster and more efficient than those protocols. Which was especially needed for colorful SBBS systems. Avatar used one escape char and binary payload after that to give commands, so it was much more compact and efficient compared to ansicodes which were easily 5-10 characters long. Some terminal apps I personally used: Kermit, ProComm, Transend, Telemate, Telix, Terminate. End of BBS eraWell well, that was nice nostalgic moment. It was year 1995 when I completely stopped using BBS systems and Internet had fully replaced those for me as communication method. Some Finnish BBS continued running for a very long time after that. Most notable one was MikroBitti magazines PCBoard based system called MBnet. Also some BBS systems allowed connecting those over internet using telnet, but for me there was no reason to use those "closed and slow" systems anymore. Want to read more? Check out: FidoNet, BBS, Zmodem, Smodem want broader description in form of documentary? Watch BBC's BBS documentary. |
posted Jan 21, 2012 11:06 AM by Sami Lehtinen
[
updated Jan 21, 2012 11:08 AM
]
Here's my Opera Mini wishlist (S60 platform) and comments: 1. IDNA whitelist doesn't contain fi tld-domain, oh why?
2. Some times left marging is completely missing and it requires screen to me scrolled to left constantly to see left most letters. If I stop scrolling to left, half or full first letter of each row is outside screen. It's also very annoying. Mostly this feature occures with mobile Wikipedia.
3. Opera often returns empty pages. Hitting reload several times fixes the problem. But why it's required.
4. If screen orientation has been switched, Opera doesn't recognize touch screen clicks correctly. This is old, but very annoying when you have device which turns screen orientation when using keyboard. One way to deal with this, is opening keyboard and using landscape mode all the time. Which naturally isn't what I want to do.
5. Some times links or page requests and responces are mixed between tabs. That's actually pretty strange. But all pages still are pages that I have visited or have links on pages I have visited, so at least it's not mixing data between users as far as I know. Some times system gets really messed up, when using 10+ tabs and Opera Mini for longer time. After that wrong pages are returned almost every time. Only way to fix that nasty situation is to restart Opera.
6. Opera should have internal PDF reader. It also also should be able to re-wrap text to make it well readable on phone screen. Technically it could preprocess PDF files on server and transfer and present those using Opera Mini's standard transport and UI. It would result as best possible UX. Therefore PDF files should not be any different from any other web-content. Now downloading files and opening those using slow native application isn't user friendly at all.
7. Option to clear privacy information on application exit.
8. IPv6 support (very low priority) - http://www.worldipv6launch.org/
9. Positive feedback: It was really nice to be able to disable that annyoing automatic phone number detection.
I'll be posing more stuff soon. I have just been really really busy with new technologies, testing, studying, etc. |
posted Dec 25, 2011 6:40 AM by Sami Lehtinen
[
updated Dec 25, 2011 6:41 AM
]
First of all, I have been enjoying great Xmas food and sleeping. But I have managed to get something else done too.
|
posted Dec 18, 2011 5:18 AM by Sami Lehtinen
[
updated Dec 18, 2011 5:24 AM
]
9ox.net now uses Google Safe Browsing API to reject phishing and malware sites. It's already using SURBL and WOT. I guess I don't do any changes before converting to Python 2.7. All changes this far which I have made, haven't affected database structure. Therefore it has been relatively safe to do these changes. When I'll implement Python 2.7 and multithreading, that's right time to fix the database structure. I have one known performance bottlenecks which is caused by bit bad implementation, and it's directly related to bad database structure design. Now it's good time to study something else for a while. I think I'll check out some compression and hashing & encryption algorithms. |
|
