v3bwfile.py 36.6 KB
Newer Older
juga's avatar
juga committed
1
2
3
4
# -*- coding: utf-8 -*-
"""Classes and functions that create the bandwidth measurements document
(v3bw) used by bandwidth authorities."""

juga's avatar
juga committed
5
import copy
juga's avatar
juga committed
6
import logging
juga's avatar
juga committed
7
import os
juga's avatar
juga committed
8
from itertools import combinations
juga's avatar
juga committed
9
from statistics import median, mean
10
from stem.descriptor import parse_file
juga's avatar
juga committed
11

12
from sbws import __version__
13
from sbws.globals import (SPEC_VERSION, BW_LINE_SIZE, SBWS_SCALE_CONSTANT,
14
15
                          TORFLOW_SCALING, SBWS_SCALING, TORFLOW_BW_MARGIN,
                          TORFLOW_OBS_LAST, TORFLOW_OBS_MEAN,
16
                          TORFLOW_ROUND_DIG, MIN_REPORT)
juga's avatar
juga committed
17
from sbws.lib.resultdump import ResultSuccess, _ResultType
18
from sbws.util.filelock import DirectoryLock
juga's avatar
juga committed
19
20
from sbws.util.timestamp import (now_isodt_str, unixts_to_isodt_str,
                                 now_unixts)
21
from sbws.util.state import State
juga's avatar
juga committed
22
23
24

log = logging.getLogger(__name__)

25
LINE_SEP = '\n'
26
27
28
29
30
KEYVALUE_SEP_V110 = '='
KEYVALUE_SEP_V200 = ' '
# List of the extra KeyValues accepted by the class
EXTRA_ARG_KEYVALUES = ['software', 'software_version', 'file_created',
                       'earliest_bandwidth', 'generator_started']
31
32
33
34
STATS_KEYVALUES = ['num_measured_relays', 'num_target_relays',
                   'num_net_relays', 'perc_measured_relays',
                   'perc_measured_targed']
KEYVALUES_INT = STATS_KEYVALUES
35
# List of all unordered KeyValues currently being used to generate the file
36
37
UNORDERED_KEYVALUES = EXTRA_ARG_KEYVALUES + STATS_KEYVALUES + \
                      ['latest_bandwidth']
38
39
# List of all the KeyValues currently being used to generate the file
ALL_KEYVALUES = ['version'] + UNORDERED_KEYVALUES
40
TERMINATOR = '===='
juga's avatar
juga committed
41
42
# Num header lines in v1.1.0 using all the KeyValues
NUM_LINES_HEADER_V110 = len(ALL_KEYVALUES) + 2
43
44
LINE_TERMINATOR = TERMINATOR + LINE_SEP

juga's avatar
juga committed
45
46
# KeyValue separator in Bandwidth Lines
BW_KEYVALUE_SEP_V110 = ' '
47
48
49
50
51
# not inclding in the files the extra bws for now
BW_KEYVALUES_BASIC = ['node_id', 'bw']
BW_KEYVALUES_FILE = BW_KEYVALUES_BASIC + \
                    ['master_key_ed25519', 'nick', 'rtt', 'time',
                     'success', 'error_stream', 'error_circ', 'error_misc']
52
53
BW_KEYVALUES_EXTRA_BWS = ['bw_bs_median', 'bw_bs_mean', 'desc_avg_bw_bs',
                          'desc_obs_bw_bs_last', 'desc_obs_bw_bs_mean']
54
BW_KEYVALUES_EXTRA = BW_KEYVALUES_FILE + BW_KEYVALUES_EXTRA_BWS
55
BW_KEYVALUES_INT = ['bw', 'rtt', 'success', 'error_stream',
56
57
                    'error_circ', 'error_misc'] + BW_KEYVALUES_EXTRA_BWS
BW_KEYVALUES = BW_KEYVALUES_BASIC + BW_KEYVALUES_EXTRA
juga's avatar
juga committed
58
59


60
61
62
63
64
65
def kb_round_x_sig_dig(bw_bs, digits=TORFLOW_ROUND_DIG):
    """Convert bw to KB and round to x most significat digits."""
    bw_kb = bw_bs / 1000
    return max(int(round(bw_kb, -digits)), 1)


juga's avatar
juga committed
66
67
68
69
70
71
72
73
74
def num_results_of_type(results, type_str):
    return len([r for r in results if r.type == type_str])


# Better way to use enums?
def result_type_to_key(type_str):
    return type_str.replace('-', '_')


75
class V3BWHeader(object):
juga's avatar
juga committed
76
77
78
79
    """
    Create a bandwidth measurements (V3bw) header
    following bandwidth measurements document spec version 1.1.0.

80
    :param str timestamp: timestamp in Unix Epoch seconds of the most recent
81
        generator result.
juga's avatar
juga committed
82
83
84
    :param str version: the spec version
    :param str software: the name of the software that generates this
    :param str software_version: the version of the software
85
    :param dict kwargs: extra headers. Currently supported:
juga's avatar
juga committed
86

87
88
89
90
        - earliest_bandwidth: str, ISO 8601 timestamp in UTC time zone
          when the first bandwidth was obtained
        - generator_started: str, ISO 8601 timestamp in UTC time zone
          when the generator started
juga's avatar
juga committed
91
    """
92
    def __init__(self, timestamp, **kwargs):
juga's avatar
juga committed
93
94
95
96
        assert isinstance(timestamp, str)
        for v in kwargs.values():
            assert isinstance(v, str)
        self.timestamp = timestamp
97
98
99
100
        # KeyValues with default value when not given by kwargs
        self.version = kwargs.get('version', SPEC_VERSION)
        self.software = kwargs.get('software', 'sbws')
        self.software_version = kwargs.get('software_version', __version__)
101
        self.file_created = kwargs.get('file_created', now_isodt_str())
juga's avatar
juga committed
102
        # latest_bandwidth should not be in kwargs, since it MUST be the
103
        # same as timestamp
juga's avatar
juga committed
104
        self.latest_bandwidth = unixts_to_isodt_str(timestamp)
105
106
        [setattr(self, k, v) for k, v in kwargs.items()
         if k in EXTRA_ARG_KEYVALUES]
juga's avatar
juga committed
107

108
109
110
111
112
    def __str__(self):
        if self.version == '1.1.0':
            return self.strv110
        return self.strv200

juga's avatar
juga committed
113
    @classmethod
juga's avatar
juga committed
114
    def from_results(cls, results, state_fpath=''):
juga's avatar
juga committed
115
116
117
        kwargs = dict()
        latest_bandwidth = cls.latest_bandwidth_from_results(results)
        earliest_bandwidth = cls.earliest_bandwidth_from_results(results)
juga's avatar
juga committed
118
        generator_started = cls.generator_started_from_file(state_fpath)
juga's avatar
juga committed
119
120
121
122
123
124
125
126
        timestamp = str(latest_bandwidth)
        kwargs['latest_bandwidth'] = unixts_to_isodt_str(latest_bandwidth)
        kwargs['earliest_bandwidth'] = unixts_to_isodt_str(earliest_bandwidth)
        if generator_started is not None:
            kwargs['generator_started'] = generator_started
        h = cls(timestamp, **kwargs)
        return h

127
128
129
130
    @classmethod
    def from_lines_v110(cls, lines):
        """
        :param list lines: list of lines to parse
131
        :returns: tuple of V3BWHeader object and non-header lines
132
133
134
135
        """
        assert isinstance(lines, list)
        try:
            index_terminator = lines.index(TERMINATOR)
Matt Traudt's avatar
Matt Traudt committed
136
        except ValueError:
137
138
139
            # is not a bw file or is v100
            log.warn('Terminator is not in lines')
            return None
140
        ts = lines[0]
141
        kwargs = dict([l.split(KEYVALUE_SEP_V110)
142
                       for l in lines[:index_terminator]
143
                       if l.split(KEYVALUE_SEP_V110)[0] in ALL_KEYVALUES])
144
        h = cls(ts, **kwargs)
juga's avatar
juga committed
145
        # last line is new line
146
        return h, lines[index_terminator + 1:-1]
147
148
149
150

    @classmethod
    def from_text_v110(self, text):
        """
151
        :param str text: text to parse
152
        :returns: tuple of V3BWHeader object and non-header lines
153
154
155
        """
        assert isinstance(text, str)
        return self.from_lines_v110(text.split(LINE_SEP))
juga's avatar
juga committed
156

juga's avatar
juga committed
157
158
159
160
161
162
163
164
165
166
167
    @classmethod
    def from_lines_v100(cls, lines):
        """
        :param list lines: list of lines to parse
        :returns: tuple of V3BWHeader object and non-header lines
        """
        assert isinstance(lines, list)
        h = cls(lines[0])
        # last line is new line
        return h, lines[1:-1]

juga's avatar
juga committed
168
    @staticmethod
juga's avatar
juga committed
169
    def generator_started_from_file(state_fpath):
170
171
172
173
        '''
        ISO formatted timestamp for the time when the scanner process most
        recently started.
        '''
juga's avatar
juga committed
174
        state = State(state_fpath)
175
176
177
178
        if 'scanner_started' in state:
            return state['scanner_started']
        else:
            return None
juga's avatar
juga committed
179
180

    @staticmethod
juga's avatar
juga committed
181
    def latest_bandwidth_from_results(results):
juga's avatar
juga committed
182
        return round(max([r.time for fp in results for r in results[fp]]))
juga's avatar
juga committed
183
184
185

    @staticmethod
    def earliest_bandwidth_from_results(results):
juga's avatar
juga committed
186
        return round(min([r.time for fp in results for r in results[fp]]))
juga's avatar
juga committed
187

juga's avatar
juga committed
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
    @property
    def keyvalue_unordered_tuple_ls(self):
        """Return list of KeyValue tuples that do not have specific order."""
        # sort the list to generate determinist headers
        keyvalue_tuple_ls = sorted([(k, v) for k, v in self.__dict__.items()
                                    if k in UNORDERED_KEYVALUES])
        return keyvalue_tuple_ls

    @property
    def keyvalue_tuple_ls(self):
        """Return list of all KeyValue tuples"""
        return [('version', self.version)] + self.keyvalue_unordered_tuple_ls

    @property
    def keyvalue_v110str_ls(self):
        """Return KeyValue list of strings following spec v1.1.0."""
        keyvalues = [self.timestamp] + [KEYVALUE_SEP_V110.join([k, v])
                                        for k, v in self.keyvalue_tuple_ls]
        return keyvalues

    @property
    def strv110(self):
        """Return header string following spec v1.1.0."""
        header_str = LINE_SEP.join(self.keyvalue_v110str_ls) + LINE_SEP + \
            LINE_TERMINATOR
        return header_str

    @property
    def keyvalue_v200_ls(self):
        """Return KeyValue list of strings following spec v2.0.0."""
        keyvalue = [self.timestamp] + [KEYVALUE_SEP_V200.join([k, v])
                                       for k, v in self.keyvalue_tuple_ls]
        return keyvalue

    @property
    def strv200(self):
        """Return header string following spec v2.0.0."""
        header_str = LINE_SEP.join(self.keyvalue_v200_ls) + LINE_SEP + \
            LINE_TERMINATOR
        return header_str

    @property
    def num_lines(self):
        return len(self.__str__().split(LINE_SEP))
juga's avatar
juga committed
232

juga's avatar
juga committed
233
234
235
236
237
    def add_stats(self, **kwargs):
        # Using kwargs because attributes might chage.
        [setattr(self, k, str(v)) for k, v in kwargs.items()
         if k in STATS_KEYVALUES]

juga's avatar
juga committed
238

juga's avatar
juga committed
239
class V3BWLine(object):
juga's avatar
juga committed
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
    """
    Create a Bandwidth List line following the spec version 1.1.0.

    :param str node_id:
    :param int bw:
    :param dict kwargs: extra headers. Currently supported:

        - nickname, str
        - master_key_ed25519, str
        - rtt, int
        - time, str
        - sucess, int
        - error_stream, int
        - error_circ, int
        - error_misc, int
    """
juga's avatar
juga committed
256
257
258
    def __init__(self, node_id, bw, **kwargs):
        assert isinstance(node_id, str)
        assert isinstance(bw, int)
259
        assert node_id.startswith('$')
juga's avatar
juga committed
260
261
262
        self.node_id = node_id
        self.bw = bw
        [setattr(self, k, v) for k, v in kwargs.items()
263
         if k in BW_KEYVALUES_EXTRA]
juga's avatar
juga committed
264

juga's avatar
juga committed
265
    def __str__(self):
juga's avatar
juga committed
266
267
        return self.bw_strv110

juga's avatar
juga committed
268
    @classmethod
juga's avatar
juga committed
269
270
    def from_results(cls, results, secs_recent=None, secs_away=None,
                     min_num=0):
271
272
273
274
275
276
277
278
279
        """Convert sbws results to relays' Bandwidth Lines

        ``bs`` stands for Bytes/seconds
        ``bw_bs_mean`` means the bw is obtained from the mean of the all the
        downloads' bandwidth.
        Downloads' bandwidth are calculated as the amount of data received
        divided by the the time it took to received.
        bw = data (Bytes) / time (seconds)
        """
juga's avatar
juga committed
280
        success_results = [r for r in results if isinstance(r, ResultSuccess)]
281
        # log.debug("Len success_results %s", len(success_results))
juga's avatar
juga committed
282
283
284
285
286
287
288
        node_id = '$' + results[0].fingerprint
        kwargs = dict()
        kwargs['nick'] = results[0].nickname
        if getattr(results[0], 'master_key_ed25519'):
            kwargs['master_key_ed25519'] = results[0].master_key_ed25519
        kwargs['time'] = cls.last_time_from_results(results)
        kwargs.update(cls.result_types_from_results(results))
289
290
        # useful args for scaling
        if success_results:
291
            if not len(success_results) >= min_num:
juga's avatar
juga committed
292
                # log.debug('The number of results is les than %s', min_num)
293
294
                return None
            results_away = \
295
296
297
                cls.results_away_each_other(success_results, secs_away)
            if not results_away:
                return None
298
299
            # log.debug("Results away from each other: %s",
            #           [unixts_to_isodt_str(r.time) for r in results_away])
300
301
302
            results_recent = cls.results_recent_than(results_away, secs_recent)
            if not results_recent:
                return None
303
            kwargs['desc_avg_bw_bs'] = \
304
305
306
307
                results_recent[-1].relay_average_bandwidth
            kwargs['rtt'] = cls.rtt_from_results(results_recent)
            bw = cls.bw_bs_median_from_results(results_recent)
            kwargs['bw_bs_mean'] = cls.bw_bs_mean_from_results(results_recent)
308
            kwargs['bw_bs_median'] = cls.bw_bs_median_from_results(
309
                results_recent)
310
            kwargs['desc_obs_bw_bs_last'] = \
311
                cls.desc_obs_bw_bs_last_from_results(results_recent)
312
            kwargs['desc_obs_bw_bs_mean'] = \
313
                cls.desc_obs_bw_bs_mean_from_results(results_recent)
314
315
316
            bwl = cls(node_id, bw, **kwargs)
            return bwl
        return None
juga's avatar
juga committed
317
318
319
320
321
322

    @classmethod
    def from_data(cls, data, fingerprint):
        assert fingerprint in data
        return cls.from_results(data[fingerprint])

juga's avatar
juga committed
323
324
325
326
327
328
    @classmethod
    def from_bw_line_v110(cls, line):
        assert isinstance(line, str)
        kwargs = dict([kv.split(KEYVALUE_SEP_V110)
                       for kv in line.split(BW_KEYVALUE_SEP_V110)
                       if kv.split(KEYVALUE_SEP_V110)[0] in BW_KEYVALUES])
juga's avatar
juga committed
329
330
331
        for k, v in kwargs.items():
            if k in BW_KEYVALUES_INT:
                kwargs[k] = int(v)
332
333
334
335
336
        node_id = kwargs['node_id']
        bw = kwargs['bw']
        del kwargs['node_id']
        del kwargs['bw']
        bw_line = cls(node_id, bw, **kwargs)
juga's avatar
juga committed
337
338
        return bw_line

juga's avatar
juga committed
339
340
    @staticmethod
    def results_away_each_other(results, secs_away=None):
juga's avatar
juga committed
341
342
        # log.debug("Checking whether results are away from each other in %s "
        #           "secs.", secs_away)
juga's avatar
juga committed
343
344
        if secs_away is None or len(results) < 2:
            return results
juga's avatar
juga committed
345
346
347
348
349
350
        for a, b in combinations(results, 2):
            if abs(a.time - b.time) > secs_away:
                return results
        # log.debug("Results are NOT away from each other in at least %ss: %s",
        #           secs_away, [unixts_to_isodt_str(r.time) for r in results])
        return None
juga's avatar
juga committed
351
352
353
354
355

    @staticmethod
    def results_recent_than(results, secs_recent=None):
        if secs_recent is None:
            return results
juga's avatar
juga committed
356
        results_recent = list(filter(
juga's avatar
juga committed
357
                            lambda x: (now_unixts() - x.time) < secs_recent,
juga's avatar
juga committed
358
359
360
361
362
363
                            results))
        # if not results_recent:
        #     log.debug("Results are NOT more recent than %ss: %s",
        #               secs_recent,
        #               [unixts_to_isodt_str(r.time) for r in results])
        return results_recent
juga's avatar
juga committed
364

365
366
367
368
369
370
371
372
373
374
    @staticmethod
    def bw_bs_median_from_results(results):
        return max(round(median([dl['amount'] / dl['duration']
                                 for r in results for dl in r.downloads])), 1)

    @staticmethod
    def bw_bs_mean_from_results(results):
        return max(round(mean([dl['amount'] / dl['duration']
                               for r in results for dl in r.downloads])), 1)

juga's avatar
juga committed
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
    @staticmethod
    def last_time_from_results(results):
        return unixts_to_isodt_str(round(max([r.time for r in results])))

    @staticmethod
    def rtt_from_results(results):
        # convert from miliseconds to seconds
        rtts = [(round(rtt * 1000)) for r in results for rtt in r.rtts]
        rtt = round(median(rtts))
        return rtt

    @staticmethod
    def result_types_from_results(results):
        rt_dict = dict([(result_type_to_key(rt.value),
                         num_results_of_type(results, rt.value))
                        for rt in _ResultType])
        return rt_dict

393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
    @staticmethod
    def desc_obs_bw_bs_mean_from_results(results):
        desc_obs_bws = []
        for r in results:
            if r.relay_observed_bandwidth is not None:
                desc_obs_bws.append(r.relay_observed_bandwidth)
        if desc_obs_bws:
            return max(round(mean(desc_obs_bws)), 1)
        return None

    @staticmethod
    def desc_obs_bw_bs_last_from_results(results):
        # the last is at the end of the list
        for r in reversed(results):
            if r.relay_observed_bandwidth is not None:
                return r.relay_observed_bandwidth
        return None

juga's avatar
juga committed
411
412
413
414
415
416
417
    @property
    def bw_keyvalue_tuple_ls(self):
        """Return list of KeyValue Bandwidth Line tuples."""
        # sort the list to generate determinist headers
        keyvalue_tuple_ls = sorted([(k, v) for k, v in self.__dict__.items()
                                    if k in BW_KEYVALUES])
        return keyvalue_tuple_ls
juga's avatar
juga committed
418

juga's avatar
juga committed
419
420
421
422
423
424
425
426
    @property
    def bw_keyvalue_v110str_ls(self):
        """Return list of KeyValue Bandwidth Line strings following
        spec v1.1.0.
        """
        bw_keyvalue_str = [KEYVALUE_SEP_V110 .join([k, str(v)])
                           for k, v in self.bw_keyvalue_tuple_ls]
        return bw_keyvalue_str
juga's avatar
juga committed
427

juga's avatar
juga committed
428
429
430
431
432
433
434
435
436
437
438
    @property
    def bw_strv110(self):
        """Return Bandwidth Line string following spec v1.1.0."""
        bw_line_str = BW_KEYVALUE_SEP_V110.join(
                        self.bw_keyvalue_v110str_ls) + LINE_SEP
        if len(bw_line_str) > BW_LINE_SIZE:
            # if this is the case, probably there are too many KeyValues,
            # or the limit needs to be changed in Tor
            log.warn("The bandwidth line %s is longer than %s",
                     len(bw_line_str), BW_LINE_SIZE)
        return bw_line_str
juga's avatar
juga committed
439

juga's avatar
juga committed
440

441
class V3BWFile(object):
juga's avatar
juga committed
442
443
444
445
446
447
    """
    Create a Bandwidth List file following spec version 1.1.0

    :param V3BWHeader v3bwheader: header
    :param list v3bwlines: V3BWLines
    """
juga's avatar
juga committed
448
449
450
451
452
    def __init__(self, v3bwheader, v3bwlines):
        self.header = v3bwheader
        self.bw_lines = v3bwlines

    def __str__(self):
453
        return str(self.header) + ''.join([str(bw_line) or ''
juga's avatar
juga committed
454
455
                                           for bw_line in self.bw_lines])

juga's avatar
juga committed
456
    @classmethod
juga's avatar
juga committed
457
    def from_results(cls, results, state_fpath='',
458
459
                     scale_constant=SBWS_SCALE_CONSTANT,
                     scaling_method=None, torflow_obs=TORFLOW_OBS_LAST,
juga's avatar
juga committed
460
                     torflow_cap=TORFLOW_BW_MARGIN,
461
                     torflow_round_digs=TORFLOW_ROUND_DIG,
462
                     secs_recent=None, secs_away=None, min_num=0,
juga's avatar
juga committed
463
                     consensus_path=None, reverse=False):
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
        """Create V3BWFile class from sbws Results.

        :param dict results: see below
        :param str state_fpath: path to the state file
        :param int scaling_method:
            Scaling method to obtain the bandwidth
            Posiable values: {NONE, SBWS_SCALING, TORFLOW_SCALING} = {0, 1, 2}
        :param int scale_constant: sbws scaling constant
        :param int torflow_obs: method to choose descriptor observed bandwidth
        :param bool reverse: whether to sort the bw lines descending or not

        Results are in the form::

            {'relay_fp1': [Result1, Result2, ...],
             'relay_fp2': [Result1, Result2, ...]}

        """
        # TODO: change scaling_method to TORFLOW_SCALING before getting this
        # in production
        log.info('Processing results to generate a bandwidth list file.')
        header = V3BWHeader.from_results(results, state_fpath)
        bw_lines_raw = []
486
        num_net_relays = cls.read_num_net_relays(consensus_path)
487
        state = State(state_fpath)
juga's avatar
juga committed
488
        for fp, values in results.items():
489
            # log.debug("Relay fp %s", fp)
juga's avatar
juga committed
490
491
            line = V3BWLine.from_results(values, secs_recent, secs_away,
                                         min_num)
juga's avatar
juga committed
492
493
            if line is not None:
                bw_lines_raw.append(line)
494
        if not bw_lines_raw:
495
496
            log.info("After applying restrictions to the raw results, "
                     "there is not any. Scaling can not be applied.")
497
498
            cls.update_progress(cls, bw_lines_raw, header, num_net_relays,
                                state)
499
500
501
            return cls(header, [])
        if scaling_method == SBWS_SCALING:
            bw_lines = cls.bw_sbws_scale(bw_lines_raw, scale_constant)
juga's avatar
juga committed
502
            cls.warn_if_not_accurate_enough(bw_lines, scale_constant)
503
504
            # log.debug(bw_lines[-1])
        elif scaling_method == TORFLOW_SCALING:
juga's avatar
juga committed
505
            bw_lines = cls.bw_torflow_scale(bw_lines_raw, torflow_obs,
506
                                            torflow_cap, torflow_round_digs)
507
            # log.debug(bw_lines[-1])
508
            cls.update_progress(cls, bw_lines, header, num_net_relays, state)
juga's avatar
juga committed
509
        else:
510
511
            bw_lines = cls.bw_kb(bw_lines_raw)
            # log.debug(bw_lines[-1])
juga's avatar
juga committed
512
513
514
        f = cls(header, bw_lines)
        return f

515
516
517
518
519
520
521
522
523
524
    @classmethod
    def from_v110_fpath(cls, fpath):
        log.info('Parsing bandwidth file %s', fpath)
        with open(fpath) as fd:
            text = fd.read()
        all_lines = text.split(LINE_SEP)
        header, lines = V3BWHeader.from_lines_v110(all_lines)
        bw_lines = [V3BWLine.from_bw_line_v110(line) for line in lines]
        return cls(header, bw_lines)

525
526
527
528
529
530
531
532
533
534
535
    @classmethod
    def from_v100_fpath(cls, fpath):
        log.info('Parsing bandwidth file %s', fpath)
        with open(fpath) as fd:
            text = fd.read()
        all_lines = text.split(LINE_SEP)
        header, lines = V3BWHeader.from_lines_v100(all_lines)
        bw_lines = sorted([V3BWLine.from_bw_line_v110(l) for l in lines],
                          key=lambda l: l.bw)
        return cls(header, bw_lines)

juga's avatar
juga committed
536
537
538
539
540
541
542
    @staticmethod
    def bw_kb(bw_lines, reverse=False):
        bw_lines_scaled = copy.deepcopy(bw_lines)
        for l in bw_lines_scaled:
            l.bw = max(round(l.bw / 1000), 1)
        return sorted(bw_lines_scaled, key=lambda x: x.bw, reverse=reverse)

juga's avatar
juga committed
543
    @staticmethod
544
    def bw_sbws_scale(bw_lines, scale_constant=SBWS_SCALE_CONSTANT,
juga's avatar
juga committed
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
                      reverse=False):
        """Return a new V3BwLine list scaled using sbws method.

        :param list bw_lines:
            bw lines to scale, not self.bw_lines,
            since this method will be before self.bw_lines have been
            initialized.
        :param int scale_constant:
            the constant to multiply by the ratio and
            the bandwidth to obtain the new bandwidth
        :returns list: V3BwLine list
        """
        # If a relay has MaxAdvertisedBandwidth set, they may be capable of
        # some large amount of bandwidth but prefer if they didn't receive it.
        # We also could have managed to measure them faster than their
        # {,Relay}BandwidthRate somehow.
        #
        # See https://github.com/pastly/simple-bw-scanner/issues/155 and
        # https://trac.torproject.org/projects/tor/ticket/8494
        #
        # Note how this isn't some measured-by-us average of bandwidth. It's
        # the first value on the 'bandwidth' line in the relay's server
        # descriptor.
        log.debug('Scaling bandwidth using sbws method.')
        m = median([l.bw for l in bw_lines])
        bw_lines_scaled = copy.deepcopy(bw_lines)
        for l in bw_lines_scaled:
            # min is to limit the bw to descriptor average-bandwidth
            # max to avoid bandwidth with 0 value
            l.bw = max(round(min(l.desc_avg_bw_bs,
575
                                 l.bw * scale_constant / m)
juga's avatar
juga committed
576
577
578
579
580
                             / 1000), 1)
        return sorted(bw_lines_scaled, key=lambda x: x.bw, reverse=reverse)

    @staticmethod
    def warn_if_not_accurate_enough(bw_lines,
581
                                    scale_constant=SBWS_SCALE_CONSTANT):
juga's avatar
juga committed
582
583
584
585
586
587
588
589
        margin = 0.001
        accuracy_ratio = median([l.bw for l in bw_lines]) / scale_constant
        log.info('The generated lines are within {:.5}% of what they should '
                 'be'.format((1 - accuracy_ratio) * 100))
        if accuracy_ratio < 1 - margin or accuracy_ratio > 1 + margin:
            log.warning('There was %f%% error and only +/- %f%% is '
                        'allowed', (1 - accuracy_ratio) * 100, margin * 100)

590
    @staticmethod
591
    def bw_torflow_scale(bw_lines, desc_obs_bws=TORFLOW_OBS_MEAN,
592
593
                         cap=TORFLOW_BW_MARGIN,
                         num_round_dig=TORFLOW_ROUND_DIG, reverse=False):
juga's avatar
juga committed
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
        """
        Obtain final bandwidth measurements applying Torflow's scaling
        method.

        From Torflow's README.spec.txt (section 2.2)::

            In this way, the resulting network status consensus bandwidth values  # NOQA
            are effectively re-weighted proportional to how much faster the node  # NOQA
            was as compared to the rest of the network.

        The variables and steps used in Torflow:

        **strm_bw**::

            The strm_bw field is the average (mean) of all the streams for the relay  # NOQA
            identified by the fingerprint field.
            strm_bw = sum(bw stream x)/|n stream|

        **filt_bw**::

            The filt_bw field is computed similarly, but only the streams equal to  # NOQA
            or greater than the strm_bw are counted in order to filter very slow  # NOQA
            streams due to slow node pairings.

        **filt_sbw and strm_sbw**::

            for rs in RouterStats.query.filter(stats_clause).\
                  options(eagerload_all('router.streams.circuit.routers')).all():  # NOQA
              tot_sbw = 0
              sbw_cnt = 0
              for s in rs.router.streams:
                if isinstance(s, ClosedStream):
                  skip = False
                  #for br in badrouters:
                  #  if br != rs:
                  #    if br.router in s.circuit.routers:
                  #      skip = True
                  if not skip:
                    # Throw out outliers < mean
                    # (too much variance for stddev to filter much)
                    if rs.strm_closed == 1 or s.bandwidth() >= rs.sbw:
                      tot_sbw += s.bandwidth()
                      sbw_cnt += 1

            if sbw_cnt: rs.filt_sbw = tot_sbw/sbw_cnt
            else: rs.filt_sbw = None

        **filt_avg, and strm_avg**::

            Once we have determined the most recent measurements for each node, we  # NOQA
            compute an average of the filt_bw fields over all nodes we have measured.  # NOQA

        ::

            filt_avg = sum(map(lambda n: n.filt_bw, nodes.itervalues()))/float(len(nodes))  # NOQA
            strm_avg = sum(map(lambda n: n.strm_bw, nodes.itervalues()))/float(len(nodes))  # NOQA

        **true_filt_avg and true_strm_avg**::

            for cl in ["Guard+Exit", "Guard", "Exit", "Middle"]:
                true_filt_avg[cl] = filt_avg
                true_strm_avg[cl] = strm_avg

        In the non-pid case, all types of nodes get the same avg

        **n.fbw_ratio and n.fsw_ratio**::

            for n in nodes.itervalues():
                n.fbw_ratio = n.filt_bw/true_filt_avg[n.node_class()]
                n.sbw_ratio = n.strm_bw/true_strm_avg[n.node_class()]

        **n.ratio**::

            These averages are used to produce ratios for each node by dividing the  # NOQA
            measured value for that node by the network average.

        ::

            # Choose the larger between sbw and fbw
              if n.sbw_ratio > n.fbw_ratio:
                n.ratio = n.sbw_ratio
              else:
                n.ratio = n.fbw_ratio

        **desc_bw**:

        It is the ``observed bandwidth`` in the descriptor, NOT the ``average
        bandwidth``::

            return Router(ns.idhex, ns.nickname, bw_observed, dead, exitpolicy,
            ns.flags, ip, version, os, uptime, published, contact, rate_limited,  # NOQA
            ns.orhash, ns.bandwidth, extra_info_digest, ns.unmeasured)
            self.desc_bw = max(bw,1) # Avoid div by 0

        **new_bw**::

            These ratios are then multiplied by the most recent observed descriptor  # NOQA
            bandwidth we have available for each node, to produce a new value for  # NOQA
            the network status consensus process.

        ::

            n.new_bw = n.desc_bw*n.ratio

        The descriptor observed bandwidth is multiplied by the ratio.

        **Limit the bandwidth to a maximum**::

            NODE_CAP = 0.05

        ::

            if n.new_bw > tot_net_bw*NODE_CAP:
              plog("INFO", "Clipping extremely fast "+n.node_class()+" node "+n.idhex+"="+n.nick+  # NOQA
                   " at "+str(100*NODE_CAP)+"% of network capacity ("+
                   str(n.new_bw)+"->"+str(int(tot_net_bw*NODE_CAP))+") "+
                   " pid_error="+str(n.pid_error)+
                   " pid_error_sum="+str(n.pid_error_sum))
              n.new_bw = int(tot_net_bw*NODE_CAP)

        However, tot_net_bw does not seems to be updated when not using pid.
        This clipping would make faster relays to all have the same value.

        All of that can be expressed as:

        .. math::
720

juga's avatar
juga committed
721
722
723
724
725
726
            bwn_i &=
                max\\left(
                    \\frac{bw_i}{\\mu},
                    \\frac{bwf_i}{\\mu_{bwf}}
                    \\right)
                \\times bwobs_i
juga's avatar
juga committed
727
728
729

        .. math::

juga's avatar
juga committed
730
731
732
733
734
735
736
737
738
739
740
741
742
743
             bwn_i &=
                max\\left(
                    \\frac{bw_i}{\\mu},
                    min \\left(
                        bw_i,
                        bw_i \\times \\mu
                        \\right)
                            \\times
                            \\frac{bw_i}{\\sum_{i=1}^{n}
                            min \\left(bw_i,
                                bw_i \\times \\mu
                            \\right)}
                    \\right)
                \\times bwobs_i \\
744

juga's avatar
juga committed
745
746
747
748
             &=
                max\\left(
                    \\frac{bw_i}{\\frac{\\sum_{i=1}^{n}bw_i}{n}},
                    min \\left(
juga's avatar
juga committed
749
                        bw_i,
juga's avatar
juga committed
750
751
752
753
754
755
756
757
758
                        bw_i \\times \\frac{\\sum_{i=1}^{n}bw_i}{n}
                        \\right)
                            \\times
                            \\frac{bw_i}{\\sum_{i=1}^{n}
                            min \\left(bw_i,
                                bw_i \\times \\frac{\\sum_{i=1}^{n}bw_i}{n}
                            \\right)}
                    \\right)
                \\times bwobs_i
juga's avatar
juga committed
759
760
761
762
763
764
765

        """
        log.info("Calculating relays' bandwidth using Torflow method.")
        bw_lines_tf = copy.deepcopy(bw_lines)
        # mean (Torflow's strm_avg)
        mu = mean([l.bw_bs_mean for l in bw_lines])
        # filtered mean (Torflow's filt_avg)
766
        muf = mean([max(l.bw_bs_mean, mu) for l in bw_lines])
juga's avatar
juga committed
767
768
        # bw sum (Torflow's tot_net_bw or tot_sbw)
        sum_bw = sum([l.bw_bs_mean for l in bw_lines])
juga's avatar
juga committed
769
770
        # Torflow's clipping
        hlimit = sum_bw * TORFLOW_BW_MARGIN
juga's avatar
juga committed
771
772
773
        log.debug('sum %s', sum_bw)
        log.debug('mu %s', mu)
        log.debug('muf %s', muf)
juga's avatar
juga committed
774
        log.debug('hlimit %s', hlimit)
juga's avatar
juga committed
775
776
777
778
779
780
        for l in bw_lines_tf:
            if desc_obs_bws == TORFLOW_OBS_LAST:
                desc_obs_bw = l.desc_obs_bw_bs_last
            elif desc_obs_bws == TORFLOW_OBS_MEAN:
                desc_obs_bw = l.desc_obs_bw_bs_mean
            # just applying the formula above:
781
782
            bw_new = kb_round_x_sig_dig(
                max(
juga's avatar
juga committed
783
                    l.bw_bs_mean / mu,  # ratio
784
                    max(l.bw_bs_mean, mu) / muf  # ratio filtered
785
786
                    ) * desc_obs_bw, \
                digits=num_round_dig)  # convert to KB
juga's avatar
juga committed
787
788
789
790
791
            # Cap maximum bw
            if cap is not None:
                bw_new = min(hlimit, bw_new)
            # remove decimals and avoid 0
            l.bw = max(round(bw_new), 1)
juga's avatar
juga committed
792
        return sorted(bw_lines_tf, key=lambda x: x.bw, reverse=reverse)
793

794
    @staticmethod
795
796
797
798
799
800
    def read_num_net_relays(consensus_path):
        """Read the number of relays in the Network from the cached consensus
        file."""
        num = None
        try:
            num = len(list(parse_file(consensus_path)))
801
        except (FileNotFoundError, AttributeError):
802
803
804
805
806
807
            log.info("It is not possible to obtain statistics about the "
                     "percentage of measured relays because the cached "
                     "consensus file is not found.")
        log.debug("Number of relays in the network %s", num)
        return num

808
    @staticmethod
809
810
    def measured_progress_stats(bw_lines, num_net_relays,
                                min_perc_reached_before):
811
812
813
814
815
        """ Statistics about measurements progress,
        to be included in the header.

        :param list bw_lines: the bw_lines after scaling and applying filters.
        :param str consensus_path: the path to the cached consensus file.
816
        :param str state_fpath: the path to the state file
817
818
819
820
821
822
823
824
825
        :returns dict, bool: Statistics about the progress made with
            measurements and whether the percentage of measured relays has been
            reached.

        """
        # cached-consensus should be updated every time that scanner get the
        # network status or descriptors?
        # It will not be updated to the last consensus, but the list of
        # measured relays is not either.
826
        assert isinstance(num_net_relays, int)
827
828
829
        assert isinstance(bw_lines, list)
        statsd = {}
        statsd['num_measured_relays'] = len(bw_lines)
830
        statsd['num_net_relays'] = num_net_relays
831
832
833
        statsd['num_target_relays'] = round(statsd['num_net_relays']
                                            * MIN_REPORT / 100)
        statsd['perc_measured_relays'] = round(len(bw_lines) * 100
834
                                               / statsd['num_net_relays'])
835
836
        statsd['perc_measured_targed'] = MIN_REPORT
        if statsd['num_measured_relays'] < statsd['num_target_relays']:
837
            # if min percent was was reached before, warn
838
839
840
841
842
843
844
845
846
            # otherwise, debug
            if min_perc_reached_before is not None:
                log.warning('The percentage of the measured relays is less '
                            'than the %s%% of the relays in the network (%s).',
                            MIN_REPORT, statsd['num_net_relays'])
            else:
                log.info('The percentage of the measured relays is less '
                         'than the %s%% of the relays in the network (%s).',
                         MIN_REPORT, statsd['num_net_relays'])
847
848
849
850
851
852
853
854
855
856
            return statsd, False
        return statsd, True

    @property
    def is_min_perc(self):
        if getattr(self.header, 'num_measured_relays', 0) \
                < getattr(self.header, 'num_target_relays', 0):
            return False
        return True

juga's avatar
juga committed
857
    @property
juga's avatar
juga committed
858
859
    def sum_bw(self):
        return sum([l.bw for l in self.bw_lines])
juga's avatar
juga committed
860
861

    @property
juga's avatar
juga committed
862
    def num(self):
juga's avatar
juga committed
863
864
865
        return len(self.bw_lines)

    @property
juga's avatar
juga committed
866
867
868
869
870
871
    def mean_bw(self):
        return mean([l.bw for l in self.bw_lines])

    @property
    def median_bw(self):
        return median([l.bw for l in self.bw_lines])
juga's avatar
juga committed
872

juga's avatar
juga committed
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
    @property
    def max_bw(self):
        return max([l.bw for l in self.bw_lines])

    @property
    def min_bw(self):
        return min([l.bw for l in self.bw_lines])

    @property
    def info_stats(self):
        if not self.bw_lines:
            return
        [log.info(': '.join([attr, str(getattr(self, attr))])) for attr in
         ['sum_bw', 'mean_bw', 'median_bw', 'num',
          'max_bw', 'min_bw']]

889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
    def update_progress(self, bw_lines, header, num_net_relays, state):
        min_perc_reached_before = state.get('min_perc_reached')
        if num_net_relays is not None:
            statsd, success = self.measured_progress_stats(
                bw_lines, num_net_relays, min_perc_reached_before)
            # add statistics about progress only when there are not enough
            # measured relays. Should some stats be added always?
            if not success:
                header.add_stats(**statsd)
                bw_lines = []
                state['min_perc_reached'] = None
            else:
                state['min_perc_reached'] = now_isodt_str()
        return bw_lines

juga's avatar
juga committed
904
905
906
907
908
909
910
911
912
913
    def bw_line_for_node_id(self, node_id):
        """Returns the bandwidth line for a given node fingerprint.

        Used to combine data when plotting.
        """
        bwl = [l for l in self.bw_lines if l.node_id == node_id]
        if bwl:
            return bwl[0]
        return None

juga's avatar
juga committed
914
915
916
917
918
919
920
921
922
    def to_plt(self, attrs=['bw'], sorted_by=None):
        """Return bandwidth data in a format useful for matplotlib.

        Used from external tool to plot.
        """
        x = [i for i in range(0, self.num)]
        ys = [[getattr(l, k) for l in self.bw_lines] for k in attrs]
        return x, ys, attrs

juga's avatar
juga committed
923
924
925
926
927
    def write(self, output):
        if output == '/dev/stdout':
            log.info("Writing to stdout is not supported.")
            return
        log.info('Writing v3bw file to %s', output)
928
929
        # To avoid inconsistent reads, the bandwidth data is written to an
        # archive path, then atomically symlinked to 'latest.v3bw'
juga's avatar
juga committed
930
931
        out_dir = os.path.dirname(output)
        out_link = os.path.join(out_dir, 'latest.v3bw')
932
        out_link_tmp = out_link + '.tmp'
juga's avatar
juga committed
933
934
935
936
937
938
        with DirectoryLock(out_dir):
            with open(output, 'wt') as fd:
                fd.write(str(self.header))
                for line in self.bw_lines:
                    fd.write(str(line))
            output_basename = os.path.basename(output)
939
940
941
942
943
944
945
946
947
948
            # To atomically symlink a file, we need to create a temporary link,
            # then rename it to the final link name. (POSIX guarantees that
            # rename is atomic.)
            log.debug('Creating symlink {} -> {}.'
                      .format(out_link_tmp, output_basename))
            os.symlink(output_basename, out_link_tmp)
            log.debug('Renaming symlink {} -> {} to {} -> {}.'
                      .format(out_link_tmp, output_basename,
                              out_link, output_basename))
            os.rename(out_link_tmp, out_link)