[SPARK-2538] [PySpark] Hash based disk spilling aggregation

During aggregation in Python worker, if the memory usage is above spark.executor.memory, it will do disk spilling aggregation.

It will split the aggregation into multiple stage, in each stage, it will partition the aggregated data by hash and dump them into disks. After all the data are aggregated, it will merge all the stages together (partition by partition).

Author: Davies Liu <davies.liu@gmail.com>

Closes #1460 from davies/spill and squashes the following commits:

cad91bf [Davies Liu] call gc.collect() after data.clear() to release memory as much as possible.
37d71f7 [Davies Liu] balance the partitions
902f036 [Davies Liu] add shuffle.py into run-tests
dcf03a9 [Davies Liu] fix memory_info() of psutil
67e6eba [Davies Liu] comment for MAX_TOTAL_PARTITIONS
f6bd5d6 [Davies Liu] rollback next_limit() again, the performance difference is huge:
e74b785 [Davies Liu] fix code style and change next_limit to memory_limit
400be01 [Davies Liu] address all the comments
6178844 [Davies Liu] refactor and improve docs
fdd0a49 [Davies Liu] add long doc string for ExternalMerger
1a97ce4 [Davies Liu] limit used memory and size of objects in partitionBy()
e6cc7f9 [Davies Liu] Merge branch 'master' into spill
3652583 [Davies Liu] address comments
e78a0a0 [Davies Liu] fix style
24cec6a [Davies Liu] get local directory by SPARK_LOCAL_DIR
57ee7ef [Davies Liu] update docs
286aaff [Davies Liu] let spilled aggregation in Python configurable
e9a40f6 [Davies Liu] recursive merger
6edbd1f [Davies Liu] Hash based disk spilling aggregation

1 files changed, 439 insertions, 0 deletions

diff --git a/python/pyspark/shuffle.py b/python/pyspark/shuffle.py
new file mode 100644
index 0000000000..e3923d1c36
--- /dev/null
+++ b/python/pyspark/shuffle.py
@@ -0,0 +1,439 @@
+#
+# Licensed to the Apache Software Foundation (ASF) under one or more
+# contributor license agreements.  See the NOTICE file distributed with
+# this work for additional information regarding copyright ownership.
+# The ASF licenses this file to You under the Apache License, Version 2.0
+# (the "License"); you may not use this file except in compliance with
+# the License.  You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+import os
+import sys
+import platform
+import shutil
+import warnings
+import gc
+
+from pyspark.serializers import BatchedSerializer, PickleSerializer
+
+try:
+    import psutil
+
+    def get_used_memory():
+        """ Return the used memory in MB """
+        process = psutil.Process(os.getpid())
+        if hasattr(process, "memory_info"):
+            info = process.memory_info()
+        else:
+            info = process.get_memory_info()
+        return info.rss >> 20
+except ImportError:
+
+    def get_used_memory():
+        """ Return the used memory in MB """
+        if platform.system() == 'Linux':
+            for line in open('/proc/self/status'):
+                if line.startswith('VmRSS:'):
+                    return int(line.split()[1]) >> 10
+        else:
+            warnings.warn("Please install psutil to have better "
+                    "support with spilling")
+            if platform.system() == "Darwin":
+                import resource
+                rss = resource.getrusage(resource.RUSAGE_SELF).ru_maxrss
+                return rss >> 20
+            # TODO: support windows
+        return 0
+
+
+class Aggregator(object):
+
+    """
+    Aggregator has tree functions to merge values into combiner.
+
+    createCombiner:  (value) -> combiner
+    mergeValue:      (combine, value) -> combiner
+    mergeCombiners:  (combiner, combiner) -> combiner
+    """
+
+    def __init__(self, createCombiner, mergeValue, mergeCombiners):
+        self.createCombiner = createCombiner
+        self.mergeValue = mergeValue
+        self.mergeCombiners = mergeCombiners
+
+
+class SimpleAggregator(Aggregator):
+
+    """
+    SimpleAggregator is useful for the cases that combiners have
+    same type with values
+    """
+
+    def __init__(self, combiner):
+        Aggregator.__init__(self, lambda x: x, combiner, combiner)
+
+
+class Merger(object):
+
+    """
+    Merge shuffled data together by aggregator
+    """
+
+    def __init__(self, aggregator):
+        self.agg = aggregator
+
+    def mergeValues(self, iterator):
+        """ Combine the items by creator and combiner """
+        raise NotImplementedError
+
+    def mergeCombiners(self, iterator):
+        """ Merge the combined items by mergeCombiner """
+        raise NotImplementedError
+
+    def iteritems(self):
+        """ Return the merged items ad iterator """
+        raise NotImplementedError
+
+
+class InMemoryMerger(Merger):
+
+    """
+    In memory merger based on in-memory dict.
+    """
+
+    def __init__(self, aggregator):
+        Merger.__init__(self, aggregator)
+        self.data = {}
+
+    def mergeValues(self, iterator):
+        """ Combine the items by creator and combiner """
+        # speed up attributes lookup
+        d, creator = self.data, self.agg.createCombiner
+        comb = self.agg.mergeValue
+        for k, v in iterator:
+            d[k] = comb(d[k], v) if k in d else creator(v)
+
+    def mergeCombiners(self, iterator):
+        """ Merge the combined items by mergeCombiner """
+        # speed up attributes lookup
+        d, comb = self.data, self.agg.mergeCombiners
+        for k, v in iterator:
+            d[k] = comb(d[k], v) if k in d else v
+
+    def iteritems(self):
+        """ Return the merged items ad iterator """
+        return self.data.iteritems()
+
+
+class ExternalMerger(Merger):
+
+    """
+    External merger will dump the aggregated data into disks when
+    memory usage goes above the limit, then merge them together.
+
+    This class works as follows:
+
+    - It repeatedly combine the items and save them in one dict in 
+      memory.
+
+    - When the used memory goes above memory limit, it will split
+      the combined data into partitions by hash code, dump them
+      into disk, one file per partition.
+
+    - Then it goes through the rest of the iterator, combine items
+      into different dict by hash. Until the used memory goes over
+      memory limit, it dump all the dicts into disks, one file per
+      dict. Repeat this again until combine all the items.
+
+    - Before return any items, it will load each partition and
+      combine them seperately. Yield them before loading next
+      partition.
+
+    - During loading a partition, if the memory goes over limit,
+      it will partition the loaded data and dump them into disks
+      and load them partition by partition again.
+
+    `data` and `pdata` are used to hold the merged items in memory.
+    At first, all the data are merged into `data`. Once the used
+    memory goes over limit, the items in `data` are dumped indo
+    disks, `data` will be cleared, all rest of items will be merged
+    into `pdata` and then dumped into disks. Before returning, all
+    the items in `pdata` will be dumped into disks.
+
+    Finally, if any items were spilled into disks, each partition
+    will be merged into `data` and be yielded, then cleared.
+
+    >>> agg = SimpleAggregator(lambda x, y: x + y)
+    >>> merger = ExternalMerger(agg, 10)
+    >>> N = 10000
+    >>> merger.mergeValues(zip(xrange(N), xrange(N)) * 10)
+    >>> assert merger.spills > 0
+    >>> sum(v for k,v in merger.iteritems())
+    499950000
+
+    >>> merger = ExternalMerger(agg, 10)
+    >>> merger.mergeCombiners(zip(xrange(N), xrange(N)) * 10)
+    >>> assert merger.spills > 0
+    >>> sum(v for k,v in merger.iteritems())
+    499950000
+    """
+
+    # the max total partitions created recursively
+    MAX_TOTAL_PARTITIONS = 4096
+
+    def __init__(self, aggregator, memory_limit=512, serializer=None,
+            localdirs=None, scale=1, partitions=59, batch=1000):
+        Merger.__init__(self, aggregator)
+        self.memory_limit = memory_limit
+        # default serializer is only used for tests
+        self.serializer = serializer or \
+                BatchedSerializer(PickleSerializer(), 1024)
+        self.localdirs = localdirs or self._get_dirs()
+        # number of partitions when spill data into disks
+        self.partitions = partitions
+        # check the memory after # of items merged
+        self.batch = batch
+        # scale is used to scale down the hash of key for recursive hash map
+        self.scale = scale
+        # unpartitioned merged data
+        self.data = {}
+        # partitioned merged data, list of dicts
+        self.pdata = []
+        # number of chunks dumped into disks
+        self.spills = 0
+        # randomize the hash of key, id(o) is the address of o (aligned by 8)
+        self._seed = id(self) + 7
+
+    def _get_dirs(self):
+        """ Get all the directories """
+        path = os.environ.get("SPARK_LOCAL_DIR", "/tmp")
+        dirs = path.split(",")
+        return [os.path.join(d, "python", str(os.getpid()), str(id(self)))
+                for d in dirs]
+
+    def _get_spill_dir(self, n):
+        """ Choose one directory for spill by number n """
+        return os.path.join(self.localdirs[n % len(self.localdirs)], str(n))
+
+    def _next_limit(self):
+        """
+        Return the next memory limit. If the memory is not released
+        after spilling, it will dump the data only when the used memory
+        starts to increase.
+        """
+        return max(self.memory_limit, get_used_memory() * 1.05)
+
+    def mergeValues(self, iterator):
+        """ Combine the items by creator and combiner """
+        iterator = iter(iterator)
+        # speedup attribute lookup
+        creator, comb = self.agg.createCombiner, self.agg.mergeValue
+        d, c, batch = self.data, 0, self.batch
+
+        for k, v in iterator:
+            d[k] = comb(d[k], v) if k in d else creator(v)
+
+            c += 1
+            if c % batch == 0 and get_used_memory() > self.memory_limit:
+                self._spill()
+                self._partitioned_mergeValues(iterator, self._next_limit())
+                break
+
+    def _partition(self, key):
+        """ Return the partition for key """
+        return hash((key, self._seed)) % self.partitions
+
+    def _partitioned_mergeValues(self, iterator, limit=0):
+        """ Partition the items by key, then combine them """
+        # speedup attribute lookup
+        creator, comb = self.agg.createCombiner, self.agg.mergeValue
+        c, pdata, hfun, batch = 0, self.pdata, self._partition, self.batch
+
+        for k, v in iterator:
+            d = pdata[hfun(k)]
+            d[k] = comb(d[k], v) if k in d else creator(v)
+            if not limit:
+                continue
+
+            c += 1
+            if c % batch == 0 and get_used_memory() > limit:
+                self._spill()
+                limit = self._next_limit()
+
+    def mergeCombiners(self, iterator, check=True):
+        """ Merge (K,V) pair by mergeCombiner """
+        iterator = iter(iterator)
+        # speedup attribute lookup
+        d, comb, batch = self.data, self.agg.mergeCombiners, self.batch
+        c = 0
+        for k, v in iterator:
+            d[k] = comb(d[k], v) if k in d else v
+            if not check:
+                continue
+
+            c += 1
+            if c % batch == 0 and get_used_memory() > self.memory_limit:
+                self._spill()
+                self._partitioned_mergeCombiners(iterator, self._next_limit())
+                break
+
+    def _partitioned_mergeCombiners(self, iterator, limit=0):
+        """ Partition the items by key, then merge them """
+        comb, pdata = self.agg.mergeCombiners, self.pdata
+        c, hfun = 0, self._partition
+        for k, v in iterator:
+            d = pdata[hfun(k)]
+            d[k] = comb(d[k], v) if k in d else v
+            if not limit:
+                continue
+
+            c += 1
+            if c % self.batch == 0 and get_used_memory() > limit:
+                self._spill()
+                limit = self._next_limit()
+
+    def _spill(self):
+        """
+        dump already partitioned data into disks.
+
+        It will dump the data in batch for better performance.
+        """
+        path = self._get_spill_dir(self.spills)
+        if not os.path.exists(path):
+            os.makedirs(path)
+
+        if not self.pdata:
+            # The data has not been partitioned, it will iterator the
+            # dataset once, write them into different files, has no
+            # additional memory. It only called when the memory goes
+            # above limit at the first time.
+
+            # open all the files for writing
+            streams = [open(os.path.join(path, str(i)), 'w')
+                       for i in range(self.partitions)]
+
+            for k, v in self.data.iteritems():
+                h = self._partition(k)
+                # put one item in batch, make it compatitable with load_stream
+                # it will increase the memory if dump them in batch
+                self.serializer.dump_stream([(k, v)], streams[h])
+
+            for s in streams:
+                s.close()
+
+            self.data.clear()
+            self.pdata = [{} for i in range(self.partitions)]
+
+        else:
+            for i in range(self.partitions):
+                p = os.path.join(path, str(i))
+                with open(p, "w") as f:
+                    # dump items in batch
+                    self.serializer.dump_stream(self.pdata[i].iteritems(), f)
+                self.pdata[i].clear()
+
+        self.spills += 1
+        gc.collect() # release the memory as much as possible
+
+    def iteritems(self):
+        """ Return all merged items as iterator """
+        if not self.pdata and not self.spills:
+            return self.data.iteritems()
+        return self._external_items()
+
+    def _external_items(self):
+        """ Return all partitioned items as iterator """
+        assert not self.data
+        if any(self.pdata):
+            self._spill()
+        hard_limit = self._next_limit()
+
+        try:
+            for i in range(self.partitions):
+                self.data = {}
+                for j in range(self.spills):
+                    path = self._get_spill_dir(j)
+                    p = os.path.join(path, str(i))
+                    # do not check memory during merging
+                    self.mergeCombiners(self.serializer.load_stream(open(p)),
+                                        False)
+
+                    # limit the total partitions
+                    if (self.scale * self.partitions < self.MAX_TOTAL_PARTITIONS
+                            and j < self.spills - 1
+                            and get_used_memory() > hard_limit):
+                        self.data.clear() # will read from disk again
+                        gc.collect() # release the memory as much as possible
+                        for v in self._recursive_merged_items(i):
+                            yield v
+                        return
+
+                for v in self.data.iteritems():
+                    yield v
+                self.data.clear()
+                gc.collect()
+
+                # remove the merged partition
+                for j in range(self.spills):
+                    path = self._get_spill_dir(j)
+                    os.remove(os.path.join(path, str(i)))
+
+        finally:
+            self._cleanup()
+
+    def _cleanup(self):
+        """ Clean up all the files in disks """
+        for d in self.localdirs:
+            shutil.rmtree(d, True)
+
+    def _recursive_merged_items(self, start):
+        """
+        merge the partitioned items and return the as iterator
+
+        If one partition can not be fit in memory, then them will be
+        partitioned and merged recursively.
+        """
+        # make sure all the data are dumps into disks.
+        assert not self.data
+        if any(self.pdata):
+            self._spill()
+        assert self.spills > 0
+
+        for i in range(start, self.partitions):
+            subdirs = [os.path.join(d, "parts", str(i))
+                            for d in self.localdirs]
+            m = ExternalMerger(self.agg, self.memory_limit, self.serializer,
+                    subdirs, self.scale * self.partitions)
+            m.pdata = [{} for _ in range(self.partitions)]
+            limit = self._next_limit()
+
+            for j in range(self.spills):
+                path = self._get_spill_dir(j)
+                p = os.path.join(path, str(i))
+                m._partitioned_mergeCombiners(
+                        self.serializer.load_stream(open(p)))
+
+                if get_used_memory() > limit:
+                    m._spill()
+                    limit = self._next_limit()
+
+            for v in m._external_items():
+                yield v
+
+            # remove the merged partition
+            for j in range(self.spills):
+                path = self._get_spill_dir(j)
+                os.remove(os.path.join(path, str(i)))
+
+
+if __name__ == "__main__":
+    import doctest
+    doctest.testmod()