diff options
Diffstat (limited to 'docs/client_examples/benchmarks')
3 files changed, 361 insertions, 0 deletions
diff --git a/docs/client_examples/benchmarks/get_sample.sh b/docs/client_examples/benchmarks/get_sample.sh new file mode 100755 index 00000000..1995eee1 --- /dev/null +++ b/docs/client_examples/benchmarks/get_sample.sh @@ -0,0 +1,3 @@ +#!/bin/sh +mkdir tmp +wget http://www.gutenberg.org/cache/epub/101/pg101.txt -O hacker_crackdown.txt diff --git a/docs/client_examples/benchmarks/measure_index_times.py b/docs/client_examples/benchmarks/measure_index_times.py new file mode 100644 index 00000000..f9349758 --- /dev/null +++ b/docs/client_examples/benchmarks/measure_index_times.py @@ -0,0 +1,179 @@ +# -*- coding: utf-8 -*- +# measure_index_times.py +# Copyright (C) 2014 LEAP +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +""" +Measure u1db retrieval times for different u1db index situations. +""" +from __future__ import print_function +from functools import partial +import datetime +import hashlib +import os +import sys + +from twisted.internet import defer, reactor + +from leap.soledad.common import l2db +from leap.soledad.client import adbapi +from leap.soledad.client._db.sqlcipher import SQLCipherOptions + + +folder = os.environ.get("TMPDIR", "tmp") +numdocs = int(os.environ.get("DOCS", "1000")) +silent = os.environ.get("SILENT", False) +tmpdb = os.path.join(folder, "test.soledad") + + +sample_file = os.environ.get("SAMPLE", "hacker_crackdown.txt") +sample_path = os.path.join(os.curdir, sample_file) + +try: + with open(sample_file) as f: + SAMPLE = f.readlines() +except Exception: + print("[!] Problem opening sample file. Did you download " + "the sample, or correctly set 'SAMPLE' env var?") + sys.exit(1) + +if numdocs > len(SAMPLE): + print("[!] Sorry! The requested DOCS number is larger than " + "the num of lines in our sample file") + sys.exit(1) + + +def debug(*args): + if not silent: + print(*args) + + +debug("[+] db path:", tmpdb) +debug("[+] num docs", numdocs) + +if os.path.isfile(tmpdb): + debug("[+] Removing existing db file...") + os.remove(tmpdb) + +start_time = datetime.datetime.now() + +opts = SQLCipherOptions(tmpdb, "secret", create=True) +dbpool = adbapi.getConnectionPool(opts) + + +def createDoc(doc): + return dbpool.runU1DBQuery("create_doc", doc) + + +db_indexes = { + 'by-chash': ['chash'], + 'by-number': ['number']} + + +def create_indexes(_): + deferreds = [] + for index, definition in db_indexes.items(): + d = dbpool.runU1DBQuery("create_index", index, *definition) + deferreds.append(d) + return defer.gatherResults(deferreds) + + +class TimeWitness(object): + def __init__(self, init_time): + self.init_time = init_time + + def get_time_count(self): + return datetime.datetime.now() - self.init_time + + +def get_from_index(_): + init_time = datetime.datetime.now() + debug("GETTING FROM INDEX...", init_time) + + def printValue(res, time): + print("RESULT->", res) + print("Index Query Took: ", time.get_time_count()) + return res + + d = dbpool.runU1DBQuery( + "get_from_index", "by-chash", + # "1150c7f10fabce0a57ce13071349fc5064f15bdb0cc1bf2852f74ef3f103aff5") + # XXX this is line 89 from the hacker crackdown... + # Should accept any other optional hash as an enviroment variable. + "57793320d4997a673fc7062652da0596c36a4e9fbe31310d2281e67d56d82469") + d.addCallback(printValue, TimeWitness(init_time)) + return d + + +def getAllDocs(): + return dbpool.runU1DBQuery("get_all_docs") + + +def errBack(e): + debug("[!] ERROR FOUND!!!") + e.printTraceback() + reactor.stop() + + +def countDocs(_): + debug("counting docs...") + d = getAllDocs() + d.addCallbacks(printResult, errBack) + d.addCallbacks(allDone, errBack) + return d + + +def printResult(r, **kwargs): + if kwargs: + debug(*kwargs.values()) + elif isinstance(r, l2db.Document): + debug(r.doc_id, r.content['number']) + else: + len_results = len(r[1]) + debug("GOT %s results" % len(r[1])) + + if len_results == numdocs: + debug("ALL GOOD") + else: + debug("[!] MISSING DOCS!!!!!") + raise ValueError("We didn't expect this result len") + + +def allDone(_): + debug("ALL DONE!") + + end_time = datetime.datetime.now() + print((end_time - start_time).total_seconds()) + reactor.stop() + + +def insert_docs(_): + deferreds = [] + for i in range(numdocs): + payload = SAMPLE[i] + chash = hashlib.sha256(payload).hexdigest() + doc = {"number": i, "payload": payload, 'chash': chash} + d = createDoc(doc) + d.addCallbacks(partial(printResult, i=i, chash=chash, payload=payload), + lambda e: e.printTraceback()) + deferreds.append(d) + return defer.gatherResults(deferreds, consumeErrors=True) + + +d = create_indexes(None) +d.addCallback(insert_docs) +d.addCallback(get_from_index) +d.addCallback(countDocs) + +reactor.run() diff --git a/docs/client_examples/benchmarks/measure_index_times_custom_docid.py b/docs/client_examples/benchmarks/measure_index_times_custom_docid.py new file mode 100644 index 00000000..4f273c64 --- /dev/null +++ b/docs/client_examples/benchmarks/measure_index_times_custom_docid.py @@ -0,0 +1,179 @@ +# -*- coding: utf-8 -*- +# measure_index_times.py +# Copyright (C) 2014 LEAP +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see <http://www.gnu.org/licenses/>. +""" +Measure u1db retrieval times for different u1db index situations. +""" +from __future__ import print_function +from functools import partial +import datetime +import hashlib +import os +import sys + +from twisted.internet import defer, reactor + +from leap.soledad.client import adbapi +from leap.soledad.client._db.sqlcipher import SQLCipherOptions +from leap.soledad.common import l2db + + +folder = os.environ.get("TMPDIR", "tmp") +numdocs = int(os.environ.get("DOCS", "1000")) +silent = os.environ.get("SILENT", False) +tmpdb = os.path.join(folder, "test.soledad") + + +sample_file = os.environ.get("SAMPLE", "hacker_crackdown.txt") +sample_path = os.path.join(os.curdir, sample_file) + +try: + with open(sample_file) as f: + SAMPLE = f.readlines() +except Exception: + print("[!] Problem opening sample file. Did you download " + "the sample, or correctly set 'SAMPLE' env var?") + sys.exit(1) + +if numdocs > len(SAMPLE): + print("[!] Sorry! The requested DOCS number is larger than " + "the num of lines in our sample file") + sys.exit(1) + + +def debug(*args): + if not silent: + print(*args) + + +debug("[+] db path:", tmpdb) +debug("[+] num docs", numdocs) + +if os.path.isfile(tmpdb): + debug("[+] Removing existing db file...") + os.remove(tmpdb) + +start_time = datetime.datetime.now() + +opts = SQLCipherOptions(tmpdb, "secret", create=True) +dbpool = adbapi.getConnectionPool(opts) + + +def createDoc(doc, doc_id): + return dbpool.runU1DBQuery("create_doc", doc, doc_id=doc_id) + + +db_indexes = { + 'by-chash': ['chash'], + 'by-number': ['number']} + + +def create_indexes(_): + deferreds = [] + for index, definition in db_indexes.items(): + d = dbpool.runU1DBQuery("create_index", index, *definition) + deferreds.append(d) + return defer.gatherResults(deferreds) + + +class TimeWitness(object): + def __init__(self, init_time): + self.init_time = init_time + + def get_time_count(self): + return datetime.datetime.now() - self.init_time + + +def get_from_index(_): + init_time = datetime.datetime.now() + debug("GETTING FROM INDEX...", init_time) + + def printValue(res, time): + print("RESULT->", res) + print("Index Query Took: ", time.get_time_count()) + return res + + d = dbpool.runU1DBQuery( + "get_doc", + # "1150c7f10fabce0a57ce13071349fc5064f15bdb0cc1bf2852f74ef3f103aff5") + # XXX this is line 89 from the hacker crackdown... + # Should accept any other optional hash as an enviroment variable. + "57793320d4997a673fc7062652da0596c36a4e9fbe31310d2281e67d56d82469") + d.addCallback(printValue, TimeWitness(init_time)) + return d + + +def getAllDocs(): + return dbpool.runU1DBQuery("get_all_docs") + + +def errBack(e): + debug("[!] ERROR FOUND!!!") + e.printTraceback() + reactor.stop() + + +def countDocs(_): + debug("counting docs...") + d = getAllDocs() + d.addCallbacks(printResult, errBack) + d.addCallbacks(allDone, errBack) + return d + + +def printResult(r, **kwargs): + if kwargs: + debug(*kwargs.values()) + elif isinstance(r, l2db.Document): + debug(r.doc_id, r.content['number']) + else: + len_results = len(r[1]) + debug("GOT %s results" % len(r[1])) + + if len_results == numdocs: + debug("ALL GOOD") + else: + debug("[!] MISSING DOCS!!!!!") + raise ValueError("We didn't expect this result len") + + +def allDone(_): + debug("ALL DONE!") + + end_time = datetime.datetime.now() + print((end_time - start_time).total_seconds()) + reactor.stop() + + +def insert_docs(_): + deferreds = [] + for i in range(numdocs): + payload = SAMPLE[i] + chash = hashlib.sha256(payload).hexdigest() + doc = {"number": i, "payload": payload, 'chash': chash} + d = createDoc(doc, doc_id=chash) + d.addCallbacks(partial(printResult, i=i, chash=chash, payload=payload), + lambda e: e.printTraceback()) + deferreds.append(d) + return defer.gatherResults(deferreds, consumeErrors=True) + + +d = create_indexes(None) +d.addCallback(insert_docs) +d.addCallback(get_from_index) +d.addCallback(countDocs) + +reactor.run() |