Github user nickwallen commented on a diff in the pull request:
https://github.com/apache/metron/pull/1150#discussion_r210664479
--- Diff:
metron-analytics/metron-profiler-spark/src/main/java/org/apache/metron/profiler/spark/function/ProfileBuilderFunction.java
---
@@ -0,0 +1,107 @@
+/*
+ *
+ * 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.
+ *
+ */
+package org.apache.metron.profiler.spark.function;
+
+import org.apache.metron.profiler.DefaultMessageDistributor;
+import org.apache.metron.profiler.MessageDistributor;
+import org.apache.metron.profiler.MessageRoute;
+import org.apache.metron.profiler.ProfileMeasurement;
+import org.apache.metron.profiler.spark.ProfileMeasurementAdapter;
+import org.apache.metron.stellar.dsl.Context;
+import org.apache.spark.api.java.function.MapGroupsFunction;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.lang.invoke.MethodHandles;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Map;
+import java.util.Properties;
+import java.util.concurrent.TimeUnit;
+import java.util.stream.Collectors;
+import java.util.stream.Stream;
+import java.util.stream.StreamSupport;
+
+import static java.util.Comparator.comparing;
+import static
org.apache.metron.profiler.spark.BatchProfilerConfig.PERIOD_DURATION;
+import static
org.apache.metron.profiler.spark.BatchProfilerConfig.PERIOD_DURATION_UNITS;
+
+/**
+ * The function responsible for building profiles in Spark.
+ */
+public class ProfileBuilderFunction implements MapGroupsFunction<String,
MessageRoute, ProfileMeasurementAdapter> {
+
+ protected static final Logger LOG =
LoggerFactory.getLogger(MethodHandles.lookup().lookupClass());
+
+ private long periodDurationMillis;
+ private Map<String, String> globals;
+
+ public ProfileBuilderFunction(Properties properties, Map<String, String>
globals) {
+ TimeUnit periodDurationUnits =
TimeUnit.valueOf(PERIOD_DURATION_UNITS.get(properties, String.class));
+ int periodDuration = PERIOD_DURATION.get(properties, Integer.class);
+ this.periodDurationMillis =
periodDurationUnits.toMillis(periodDuration);
+ this.globals = globals;
+ }
+
+ /**
+ * Build a profile from a set of message routes.
+ *
+ * <p>This assumes that all of the necessary routes have been provided
+ *
+ * @param group The group identifier.
+ * @param iterator The message routes.
+ * @return
+ */
+ @Override
+ public ProfileMeasurementAdapter call(String group,
Iterator<MessageRoute> iterator) throws Exception {
+ // create the distributor; some settings are unnecessary because it is
cleaned-up immediately after processing the batch
+ int maxRoutes = Integer.MAX_VALUE;
+ long profileTTLMillis = Long.MAX_VALUE;
+ MessageDistributor distributor = new
DefaultMessageDistributor(periodDurationMillis, profileTTLMillis, maxRoutes);
+ Context context = TaskUtils.getContext(globals);
+
+ // sort the messages/routes
+ List<MessageRoute> routes = toStream(iterator)
+ .sorted(comparing(rt -> rt.getTimestamp()))
+ .collect(Collectors.toList());
+ LOG.debug("Building a profile for group '{}' from {} message(s)",
group, routes.size());
+
+ // apply each message/route to build the profile
+ for(MessageRoute route: routes) {
+ distributor.distribute(route, context);
+ }
--- End diff --
What we have (which maintains ordering) is probably good enough for a first
pass. Please tell me if you disagree.
--
> @simonellistonball I can see how people would want ordered timestamps.
If you're using profiles to find sequences of events that indicate state of an
attack, you do need order.
My gut tells me not necessarily, but we probably need to dig into specific
examples to see who is right. ;) Definitely worthy of digging into though.
Trying to explicitly support ordering complicates matters when streaming;
no closed data set. If you haven't set your time lag sufficiently large enough
then an out-of-order messages could break your use case.
Storm's event time processing is simple, so if an out-of-order event
arrives beyond your time lag, it gets dropped on the floor. And increasing
your time lag, equally increases latency which is a tough trade-off.
Other platforms like Spark support recalculation, so we could use a higher
time lag without suffering equally increased latency. Even then, it only
supports recalculation so long after the fact and does come at a cost.
All that being said (whether Storm, Spark, etc) a user's time lag setting
would impact the accuracy of their results if they depend on strict ordering.
If the user doesn't expect strict ordering, then we eliminate most of the
problem.
I am also thinking that our goal here is to make the output produced in
batch indistinguishable from that produced when streaming. Its much easier to
meet that goal if there is no expectation of ordering.
---
