Remove dependency on Hadoop

This pull request attempts to remove the dependency of the simulator on Apache Hadoop 
which is pulled in as a consequence of using parquet-mr. 
The reason for removal is that Apache Hadoop does not work natively on Windows 
without user intervention, which makes adoption on this platform more difficult.

* Add Windows as CI target for the OpenDC simulator
* Use `LocalInputFile` for Parquet reader usages
* Use `LocalOutputFile` for Parquet writer usages
* Remove Apache Hadoop as dependency of OpenDC.

17 files changed, 527 insertions, 83 deletions

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index b0e8cc38..28e5846b 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -12,6 +12,9 @@ jobs:
       matrix:
         os: [ubuntu-latest]
         java: [8, 16]
+        include:
+          - os: windows-latest
+            java: 16
     steps:
       - name: Checkout repository
         uses: actions/checkout@v2
diff --git a/gradle/libs.versions.toml b/gradle/libs.versions.toml
index 1d7fdd3e..22f713f6 100644
--- a/gradle/libs.versions.toml
+++ b/gradle/libs.versions.toml
@@ -5,6 +5,7 @@ slf4j = "1.7.30"
 log4j = "2.14.1"
 opentelemetry-main = "1.2.0"
 opentelemetry-metrics = "1.2.0-alpha"
+hadoop = "3.3.0"
 
 [libraries]
 kotlinx-coroutines = { module = "org.jetbrains.kotlinx:kotlinx-coroutines-core", version = "1.5.0" }
@@ -45,4 +46,5 @@ kotlinx-benchmark-runtime-jvm = { module = "org.jetbrains.kotlinx:kotlinx-benchm
 # Other
 mongodb = { module = "org.mongodb:mongodb-driver-sync", version = "4.2.3" }
 classgraph = { module = "io.github.classgraph:classgraph", version = "4.8.105" }
-hadoop-client = { module = "org.apache.hadoop:hadoop-client", version = "3.3.0" }
+hadoop-common = { module = "org.apache.hadoop:hadoop-common", version.ref = "hadoop" }
+hadoop-mapreduce-client-core = { module = "org.apache.hadoop:hadoop-mapreduce-client-core", version.ref = "hadoop" }
diff --git a/opendc-experiments/opendc-experiments-capelin/build.gradle.kts b/opendc-experiments/opendc-experiments-capelin/build.gradle.kts
index 0dade513..324cae3e 100644
--- a/opendc-experiments/opendc-experiments-capelin/build.gradle.kts
+++ b/opendc-experiments/opendc-experiments-capelin/build.gradle.kts
@@ -44,10 +44,5 @@ dependencies {
     implementation(libs.clikt)
 
     implementation(libs.parquet)
-    implementation(libs.hadoop.client) {
-        exclude(group = "org.slf4j", module = "slf4j-log4j12")
-        exclude(group = "log4j")
-    }
-
     testImplementation(libs.log4j.slf4j)
 }
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/telemetry/parquet/ParquetEventWriter.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/telemetry/parquet/ParquetEventWriter.kt
index 4fa6ae66..d8f7ff75 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/telemetry/parquet/ParquetEventWriter.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/telemetry/parquet/ParquetEventWriter.kt
@@ -25,10 +25,10 @@ package org.opendc.experiments.capelin.telemetry.parquet
 import mu.KotlinLogging
 import org.apache.avro.Schema
 import org.apache.avro.generic.GenericData
-import org.apache.hadoop.fs.Path
 import org.apache.parquet.avro.AvroParquetWriter
 import org.apache.parquet.hadoop.metadata.CompressionCodecName
 import org.opendc.experiments.capelin.telemetry.Event
+import org.opendc.format.util.LocalOutputFile
 import java.io.Closeable
 import java.io.File
 import java.util.concurrent.ArrayBlockingQueue
@@ -52,8 +52,7 @@ public open class ParquetEventWriter<in T : Event>(
     /**
      * The writer to write the Parquet file.
      */
-    @Suppress("DEPRECATION")
-    private val writer = AvroParquetWriter.builder<GenericData.Record>(Path(path.absolutePath))
+    private val writer = AvroParquetWriter.builder<GenericData.Record>(LocalOutputFile(path))
         .withSchema(schema)
         .withCompressionCodec(CompressionCodecName.SNAPPY)
         .withPageSize(4 * 1024 * 1024) // For compression
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20ParquetTraceReader.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20ParquetTraceReader.kt
index a8462a51..7f25137e 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20ParquetTraceReader.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20ParquetTraceReader.kt
@@ -38,7 +38,6 @@ import java.util.TreeSet
  * @param performanceInterferenceModel The performance model covering the workload in the VM trace.
  * @param run The run to which this reader belongs.
  */
-@OptIn(ExperimentalStdlibApi::class)
 public class Sc20ParquetTraceReader(
     rawReaders: List<Sc20RawParquetTraceReader>,
     performanceInterferenceModel: Map<String, PerformanceInterferenceModel>,
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20RawParquetTraceReader.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20RawParquetTraceReader.kt
index bd27cf02..54151c9f 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20RawParquetTraceReader.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20RawParquetTraceReader.kt
@@ -24,10 +24,9 @@ package org.opendc.experiments.capelin.trace
 
 import mu.KotlinLogging
 import org.apache.avro.generic.GenericData
-import org.apache.hadoop.fs.Path
-import org.apache.parquet.avro.AvroParquetReader
 import org.opendc.format.trace.TraceEntry
 import org.opendc.format.trace.TraceReader
+import org.opendc.format.util.LocalParquetReader
 import org.opendc.simulator.compute.workload.SimTraceWorkload
 import org.opendc.simulator.compute.workload.SimWorkload
 import java.io.File
@@ -40,16 +39,12 @@ private val logger = KotlinLogging.logger {}
  *
  * @param path The directory of the traces.
  */
-@OptIn(ExperimentalStdlibApi::class)
 public class Sc20RawParquetTraceReader(private val path: File) {
     /**
      * Read the fragments into memory.
      */
     private fun parseFragments(path: File): Map<String, List<SimTraceWorkload.Fragment>> {
-        @Suppress("DEPRECATION")
-        val reader = AvroParquetReader.builder<GenericData.Record>(Path(path.absolutePath, "trace.parquet"))
-            .disableCompatibility()
-            .build()
+        val reader = LocalParquetReader<GenericData.Record>(File(path, "trace.parquet"))
 
         val fragments = mutableMapOf<String, MutableList<SimTraceWorkload.Fragment>>()
 
@@ -81,10 +76,7 @@ public class Sc20RawParquetTraceReader(private val path: File) {
      * Read the metadata into a workload.
      */
     private fun parseMeta(path: File, fragments: Map<String, List<SimTraceWorkload.Fragment>>): List<TraceEntry<SimWorkload>> {
-        @Suppress("DEPRECATION")
-        val metaReader = AvroParquetReader.builder<GenericData.Record>(Path(path.absolutePath, "meta.parquet"))
-            .disableCompatibility()
-            .build()
+        val metaReader = LocalParquetReader<GenericData.Record>(File(path, "meta.parquet"))
 
         var counter = 0
         val entries = mutableListOf<TraceEntry<SimWorkload>>()
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20StreamingParquetTraceReader.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20StreamingParquetTraceReader.kt
index c5294b55..6792c2ab 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20StreamingParquetTraceReader.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20StreamingParquetTraceReader.kt
@@ -24,7 +24,6 @@ package org.opendc.experiments.capelin.trace
 
 import mu.KotlinLogging
 import org.apache.avro.generic.GenericData
-import org.apache.hadoop.fs.Path
 import org.apache.parquet.avro.AvroParquetReader
 import org.apache.parquet.filter2.compat.FilterCompat
 import org.apache.parquet.filter2.predicate.FilterApi
@@ -33,6 +32,7 @@ import org.apache.parquet.filter2.predicate.UserDefinedPredicate
 import org.apache.parquet.io.api.Binary
 import org.opendc.format.trace.TraceEntry
 import org.opendc.format.trace.TraceReader
+import org.opendc.format.util.LocalInputFile
 import org.opendc.simulator.compute.interference.IMAGE_PERF_INTERFERENCE_MODEL
 import org.opendc.simulator.compute.interference.PerformanceInterferenceModel
 import org.opendc.simulator.compute.workload.SimTraceWorkload
@@ -54,7 +54,6 @@ private val logger = KotlinLogging.logger {}
  * @param traceFile The directory of the traces.
  * @param performanceInterferenceModel The performance model covering the workload in the VM trace.
  */
-@OptIn(ExperimentalStdlibApi::class)
 public class Sc20StreamingParquetTraceReader(
     traceFile: File,
     performanceInterferenceModel: PerformanceInterferenceModel? = null,
@@ -96,10 +95,10 @@ public class Sc20StreamingParquetTraceReader(
      * The thread to read the records in.
      */
     private val readerThread = thread(start = true, name = "sc20-reader") {
-        @Suppress("DEPRECATION")
-        val reader = AvroParquetReader.builder<GenericData.Record>(Path(traceFile.absolutePath, "trace.parquet"))
+        val reader = AvroParquetReader
+            .builder<GenericData.Record>(LocalInputFile(File(traceFile, "trace.parquet")))
             .disableCompatibility()
-            .run { if (filter != null) withFilter(filter) else this }
+            .withFilter(filter)
             .build()
 
         try {
@@ -164,10 +163,10 @@ public class Sc20StreamingParquetTraceReader(
         val entries = mutableMapOf<String, GenericData.Record>()
         val buffers = mutableMapOf<String, MutableList<MutableList<SimTraceWorkload.Fragment>>>()
 
-        @Suppress("DEPRECATION")
-        val metaReader = AvroParquetReader.builder<GenericData.Record>(Path(traceFile.absolutePath, "meta.parquet"))
+        val metaReader = AvroParquetReader
+            .builder<GenericData.Record>(LocalInputFile(File(traceFile, "meta.parquet")))
             .disableCompatibility()
-            .run { if (filter != null) withFilter(filter) else this }
+            .withFilter(filter)
             .build()
 
         while (true) {
@@ -178,7 +177,7 @@ public class Sc20StreamingParquetTraceReader(
 
         metaReader.close()
 
-        val selection = if (selectedVms.isEmpty()) entries.keys else selectedVms
+        val selection = selectedVms.ifEmpty { entries.keys }
 
         // Create the entry iterator
         iterator = selection.asSequence()
diff --git a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20TraceConverter.kt b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20TraceConverter.kt
index 1f9e289c..d0031a66 100644
--- a/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20TraceConverter.kt
+++ b/opendc-experiments/opendc-experiments-capelin/src/main/kotlin/org/opendc/experiments/capelin/trace/Sc20TraceConverter.kt
@@ -38,11 +38,11 @@ import me.tongfei.progressbar.ProgressBar
 import org.apache.avro.Schema
 import org.apache.avro.SchemaBuilder
 import org.apache.avro.generic.GenericData
-import org.apache.hadoop.fs.Path
 import org.apache.parquet.avro.AvroParquetWriter
 import org.apache.parquet.hadoop.ParquetWriter
 import org.apache.parquet.hadoop.metadata.CompressionCodecName
 import org.opendc.format.trace.sc20.Sc20VmPlacementReader
+import org.opendc.format.util.LocalOutputFile
 import java.io.BufferedReader
 import java.io.File
 import java.io.FileReader
@@ -109,16 +109,14 @@ public class TraceConverterCli : CliktCommand(name = "trace-converter") {
             traceParquet.delete()
         }
 
-        @Suppress("DEPRECATION")
-        val metaWriter = AvroParquetWriter.builder<GenericData.Record>(Path(metaParquet.toURI()))
+        val metaWriter = AvroParquetWriter.builder<GenericData.Record>(LocalOutputFile(metaParquet))
             .withSchema(metaSchema)
             .withCompressionCodec(CompressionCodecName.SNAPPY)
             .withPageSize(4 * 1024 * 1024) // For compression
             .withRowGroupSize(16 * 1024 * 1024) // For write buffering (Page size)
             .build()
 
-        @Suppress("DEPRECATION")
-        val writer = AvroParquetWriter.builder<GenericData.Record>(Path(traceParquet.toURI()))
+        val writer = AvroParquetWriter.builder<GenericData.Record>(LocalOutputFile(traceParquet))
             .withSchema(schema)
             .withCompressionCodec(CompressionCodecName.SNAPPY)
             .withPageSize(4 * 1024 * 1024) // For compression
diff --git a/opendc-experiments/opendc-experiments-serverless20/build.gradle.kts b/opendc-experiments/opendc-experiments-serverless20/build.gradle.kts
index 88479765..7d68cb3a 100644
--- a/opendc-experiments/opendc-experiments-serverless20/build.gradle.kts
+++ b/opendc-experiments/opendc-experiments-serverless20/build.gradle.kts
@@ -37,10 +37,4 @@ dependencies {
     implementation(projects.opendcTelemetry.opendcTelemetrySdk)
     implementation(libs.kotlin.logging)
     implementation(libs.config)
-
-    implementation(libs.parquet)
-    implementation(libs.hadoop.client) {
-        exclude(group = "org.slf4j", module = "slf4j-log4j12")
-        exclude(group = "log4j")
-    }
 }
diff --git a/opendc-experiments/opendc-experiments-tf20/build.gradle.kts b/opendc-experiments/opendc-experiments-tf20/build.gradle.kts
index 64483bd4..b088045b 100644
--- a/opendc-experiments/opendc-experiments-tf20/build.gradle.kts
+++ b/opendc-experiments/opendc-experiments-tf20/build.gradle.kts
@@ -38,9 +38,4 @@ dependencies {
     implementation(projects.opendcUtils)
 
     implementation(libs.kotlin.logging)
-    implementation(libs.parquet)
-    implementation(libs.hadoop.client) {
-        exclude(group = "org.slf4j", module = "slf4j-log4j12")
-        exclude(group = "log4j")
-    }
 }
diff --git a/opendc-format/build.gradle.kts b/opendc-format/build.gradle.kts
index e95cb666..e19e0ec8 100644
--- a/opendc-format/build.gradle.kts
+++ b/opendc-format/build.gradle.kts
@@ -40,9 +40,29 @@ dependencies {
     }
     implementation(kotlin("reflect"))
 
-    implementation(libs.parquet)
-    implementation(libs.hadoop.client) {
+    /* This configuration is necessary for a slim dependency on Apache Parquet */
+    implementation(libs.parquet) {
+        exclude(group = "org.apache.hadoop")
+    }
+    runtimeOnly(libs.hadoop.common) {
         exclude(group = "org.slf4j", module = "slf4j-log4j12")
         exclude(group = "log4j")
+        exclude(group = "org.apache.hadoop")
+        exclude(group = "org.apache.curator")
+        exclude(group = "org.apache.zookeeper")
+        exclude(group = "org.apache.kerby")
+        exclude(group = "org.apache.httpcomponents")
+        exclude(group = "org.apache.htrace")
+        exclude(group = "commons-cli")
+        exclude(group = "javax.servlet")
+        exclude(group = "org.eclipse.jetty")
+        exclude(group = "com.sun.jersey")
+        exclude(group = "com.jcraft")
+        exclude(group = "dnsjava")
+    }
+    runtimeOnly(libs.hadoop.mapreduce.client.core) {
+        isTransitive = false
     }
+
+    testRuntimeOnly(libs.slf4j.simple)
 }
diff --git a/opendc-format/src/main/kotlin/org/opendc/format/trace/wtf/WtfTraceReader.kt b/opendc-format/src/main/kotlin/org/opendc/format/trace/wtf/WtfTraceReader.kt
index feadf61f..dde1b340 100644
--- a/opendc-format/src/main/kotlin/org/opendc/format/trace/wtf/WtfTraceReader.kt
+++ b/opendc-format/src/main/kotlin/org/opendc/format/trace/wtf/WtfTraceReader.kt
@@ -23,15 +23,16 @@
 package org.opendc.format.trace.wtf
 
 import org.apache.avro.generic.GenericRecord
-import org.apache.hadoop.fs.Path
-import org.apache.parquet.avro.AvroParquetReader
 import org.opendc.format.trace.TraceEntry
 import org.opendc.format.trace.TraceReader
+import org.opendc.format.util.LocalParquetReader
 import org.opendc.simulator.compute.workload.SimFlopsWorkload
 import org.opendc.workflow.api.Job
 import org.opendc.workflow.api.Task
 import org.opendc.workflow.api.WORKFLOW_TASK_CORES
 import org.opendc.workflow.api.WORKFLOW_TASK_DEADLINE
+import java.io.File
+import java.nio.file.Path
 import java.util.UUID
 import kotlin.math.min
 
@@ -41,13 +42,20 @@ import kotlin.math.min
  *
  * @param path The path to the trace.
  */
-public class WtfTraceReader(path: String) : TraceReader<Job> {
+public class WtfTraceReader(path: Path) : TraceReader<Job> {
     /**
      * The internal iterator to use for this reader.
      */
     private val iterator: Iterator<TraceEntry<Job>>
 
     /**
+     * Construct a [TraceReader] from the specified [path].
+     *
+     * @param path The path to the trace.
+     */
+    public constructor(path: File) : this(path.toPath())
+
+    /**
      * Initialize the reader.
      */
     init {
@@ -56,43 +64,43 @@ public class WtfTraceReader(path: String) : TraceReader<Job> {
         val tasks = mutableMapOf<Long, Task>()
         val taskDependencies = mutableMapOf<Task, List<Long>>()
 
-        @Suppress("DEPRECATION")
-        val reader = AvroParquetReader.builder<GenericRecord>(Path(path, "tasks/schema-1.0")).build()
+        LocalParquetReader<GenericRecord>(path.resolve("tasks/schema-1.0")).use { reader ->
+            while (true) {
+                val nextRecord = reader.read() ?: break
 
-        while (true) {
-            val nextRecord = reader.read() ?: break
+                val workflowId = nextRecord.get("workflow_id") as Long
+                val taskId = nextRecord.get("id") as Long
+                val submitTime = nextRecord.get("ts_submit") as Long
+                val runtime = nextRecord.get("runtime") as Long
+                val cores = (nextRecord.get("resource_amount_requested") as Double).toInt()
 
-            val workflowId = nextRecord.get("workflow_id") as Long
-            val taskId = nextRecord.get("id") as Long
-            val submitTime = nextRecord.get("ts_submit") as Long
-            val runtime = nextRecord.get("runtime") as Long
-            val cores = (nextRecord.get("resource_amount_requested") as Double).toInt()
-            @Suppress("UNCHECKED_CAST")
-            val dependencies = (nextRecord.get("parents") as ArrayList<GenericRecord>).map {
-                it.get("item") as Long
-            }
+                @Suppress("UNCHECKED_CAST")
+                val dependencies = (nextRecord.get("parents") as ArrayList<GenericRecord>).map {
+                    it.get("item") as Long
+                }
 
-            val flops: Long = 4100 * (runtime / 1000) * cores
+                val flops: Long = 4100 * (runtime / 1000) * cores
 
-            val workflow = workflows.getOrPut(workflowId) {
-                Job(UUID(0L, workflowId), "<unnamed>", HashSet())
-            }
-            val workload = SimFlopsWorkload(flops)
-            val task = Task(
-                UUID(0L, taskId),
-                "<unnamed>",
-                HashSet(),
-                mapOf(
-                    "workload" to workload,
-                    WORKFLOW_TASK_CORES to cores,
-                    WORKFLOW_TASK_DEADLINE to runtime
+                val workflow = workflows.getOrPut(workflowId) {
+                    Job(UUID(0L, workflowId), "<unnamed>", HashSet())
+                }
+                val workload = SimFlopsWorkload(flops)
+                val task = Task(
+                    UUID(0L, taskId),
+                    "<unnamed>",
+                    HashSet(),
+                    mapOf(
+                        "workload" to workload,
+                        WORKFLOW_TASK_CORES to cores,
+                        WORKFLOW_TASK_DEADLINE to runtime
+                    )
                 )
-            )
 
-            starts.merge(workflowId, submitTime, ::min)
-            (workflow.tasks as MutableSet<Task>).add(task)
-            tasks[taskId] = task
-            taskDependencies[task] = dependencies
+                starts.merge(workflowId, submitTime, ::min)
+                (workflow.tasks as MutableSet<Task>).add(task)
+                tasks[taskId] = task
+                taskDependencies[task] = dependencies
+            }
         }
 
         // Fix dependencies and dependents for all tasks
diff --git a/opendc-format/src/main/kotlin/org/opendc/format/util/LocalInputFile.kt b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalInputFile.kt
new file mode 100644
index 00000000..92319ace
--- /dev/null
+++ b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalInputFile.kt
@@ -0,0 +1,107 @@
+/*
+ * Copyright (c) 2021 AtLarge Research
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+package org.opendc.format.util
+
+import org.apache.parquet.io.InputFile
+import org.apache.parquet.io.SeekableInputStream
+import java.io.EOFException
+import java.io.File
+import java.nio.ByteBuffer
+import java.nio.channels.FileChannel
+import java.nio.file.Path
+import java.nio.file.StandardOpenOption
+
+/**
+ * An [InputFile] on the local filesystem.
+ */
+public class LocalInputFile(private val path: Path) : InputFile {
+    /**
+     * The [FileChannel] used for accessing the input path.
+     */
+    private val channel = FileChannel.open(path, StandardOpenOption.READ)
+
+    /**
+     * Construct a [LocalInputFile] for the specified [file].
+     */
+    public constructor(file: File) : this(file.toPath())
+
+    override fun getLength(): Long = channel.size()
+
+    override fun newStream(): SeekableInputStream = object : SeekableInputStream() {
+        override fun read(buf: ByteBuffer): Int {
+            return channel.read(buf)
+        }
+
+        override fun read(): Int {
+            val single = ByteBuffer.allocate(1)
+            var read: Int
+
+            // ReadableByteChannel#read might read zero bytes so continue until we read at least one byte
+            do {
+                read = channel.read(single)
+            } while (read == 0)
+
+            return if (read == -1) {
+                read
+            } else {
+                single.get(0).toInt() and 0xff
+            }
+        }
+
+        override fun getPos(): Long {
+            return channel.position()
+        }
+
+        override fun seek(newPos: Long) {
+            channel.position(newPos)
+        }
+
+        override fun readFully(bytes: ByteArray) {
+            readFully(ByteBuffer.wrap(bytes))
+        }
+
+        override fun readFully(bytes: ByteArray, start: Int, len: Int) {
+            readFully(ByteBuffer.wrap(bytes, start, len))
+        }
+
+        override fun readFully(buf: ByteBuffer) {
+            var remainder = buf.remaining()
+            while (remainder > 0) {
+                val read = channel.read(buf)
+                remainder -= read
+
+                if (read == -1 && remainder > 0) {
+                    throw EOFException()
+                }
+            }
+        }
+
+        override fun close() {
+            channel.close()
+        }
+
+        override fun toString(): String = "NioSeekableInputStream"
+    }
+
+    override fun toString(): String = "LocalInputFile[path=$path]"
+}
diff --git a/opendc-format/src/main/kotlin/org/opendc/format/util/LocalOutputFile.kt b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalOutputFile.kt
new file mode 100644
index 00000000..657bca5a
--- /dev/null
+++ b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalOutputFile.kt
@@ -0,0 +1,95 @@
+/*
+ * Copyright (c) 2021 AtLarge Research
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+package org.opendc.format.util
+
+import org.apache.parquet.io.OutputFile
+import org.apache.parquet.io.PositionOutputStream
+import java.io.File
+import java.io.OutputStream
+import java.nio.file.Files
+import java.nio.file.Path
+import java.nio.file.StandardOpenOption
+
+/**
+ * An [OutputFile] on the local filesystem.
+ */
+public class LocalOutputFile(private val path: Path) : OutputFile {
+    /**
+     * Construct a [LocalOutputFile] from the specified [file]
+     */
+    public constructor(file: File) : this(file.toPath())
+
+    override fun create(blockSizeHint: Long): PositionOutputStream {
+        val output = Files.newOutputStream(path, StandardOpenOption.CREATE_NEW, StandardOpenOption.WRITE)
+        return NioPositionOutputStream(output)
+    }
+
+    override fun createOrOverwrite(blockSizeHint: Long): PositionOutputStream {
+        val output = Files.newOutputStream(path, StandardOpenOption.CREATE, StandardOpenOption.WRITE, StandardOpenOption.TRUNCATE_EXISTING)
+        return NioPositionOutputStream(output)
+    }
+
+    override fun supportsBlockSize(): Boolean = false
+
+    override fun defaultBlockSize(): Long =
+        throw UnsupportedOperationException("Local filesystem does not have default block size")
+
+    override fun getPath(): String = path.toString()
+
+    /**
+     * Implementation of [PositionOutputStream] for an [OutputStream].
+     */
+    private class NioPositionOutputStream(private val output: OutputStream) : PositionOutputStream() {
+        /**
+         * The current position in the file.
+         */
+        private var _pos = 0L
+
+        override fun getPos(): Long = _pos
+
+        override fun write(b: Int) {
+            output.write(b)
+            _pos++
+        }
+
+        override fun write(b: ByteArray) {
+            output.write(b)
+            _pos += b.size
+        }
+
+        override fun write(b: ByteArray, off: Int, len: Int) {
+            output.write(b, off, len)
+            _pos += len
+        }
+
+        override fun flush() {
+            output.flush()
+        }
+
+        override fun close() {
+            output.close()
+        }
+
+        override fun toString(): String = "NioPositionOutputStream[output=$output]"
+    }
+}
diff --git a/opendc-format/src/main/kotlin/org/opendc/format/util/LocalParquetReader.kt b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalParquetReader.kt
new file mode 100644
index 00000000..5083f3e1
--- /dev/null
+++ b/opendc-format/src/main/kotlin/org/opendc/format/util/LocalParquetReader.kt
@@ -0,0 +1,112 @@
+/*
+ * Copyright (c) 2021 AtLarge Research
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+package org.opendc.format.util
+
+import org.apache.parquet.avro.AvroParquetReader
+import org.apache.parquet.hadoop.ParquetReader
+import org.apache.parquet.io.InputFile
+import java.io.File
+import java.io.IOException
+import java.nio.file.Files
+import java.nio.file.Path
+import kotlin.io.path.isDirectory
+
+/**
+ * A helper class to read Parquet files.
+ *
+ * @param path The path to the Parquet file or directory to read.
+ */
+public class LocalParquetReader<out T>(path: Path) : AutoCloseable {
+    /**
+     * The input files to process.
+     */
+    private val filesIterator = if (path.isDirectory())
+        Files.list(path)
+            .filter { !it.isDirectory() }
+            .sorted()
+            .map { LocalInputFile(it) }
+            .iterator()
+    else
+        listOf(LocalInputFile(path)).iterator()
+
+    /**
+     * The Parquet reader to use.
+     */
+    private var reader: ParquetReader<T>? = null
+
+    /**
+     * Construct a [LocalParquetReader] for the specified [file].
+     */
+    public constructor(file: File) : this(file.toPath())
+
+    /**
+     * Read a single entry in the Parquet file.
+     */
+    public fun read(): T? {
+        return try {
+            val next = reader?.read()
+            if (next != null) {
+                next
+            } else {
+                initReader()
+
+                if (reader == null)
+                    null
+                else
+                    read()
+            }
+        } catch (e: InterruptedException) {
+            throw IOException(e)
+        }
+    }
+
+    /**
+     * Close the Parquet reader.
+     */
+    override fun close() {
+        reader?.close()
+    }
+
+    /**
+     * Initialize the next reader.
+     */
+    private fun initReader() {
+        reader?.close()
+
+        this.reader = if (filesIterator.hasNext()) {
+            createReader(filesIterator.next())
+        } else {
+            null
+        }
+    }
+
+    /**
+     * Create a Parquet reader for the specified file.
+     */
+    private fun createReader(input: InputFile): ParquetReader<T> {
+        return AvroParquetReader
+            .builder<T>(input)
+            .disableCompatibility()
+            .build()
+    }
+}
diff --git a/opendc-format/src/test/kotlin/org/opendc/format/trace/wtf/WtfTraceReaderTest.kt b/opendc-format/src/test/kotlin/org/opendc/format/trace/wtf/WtfTraceReaderTest.kt
index bcfa7553..31ae03e0 100644
--- a/opendc-format/src/test/kotlin/org/opendc/format/trace/wtf/WtfTraceReaderTest.kt
+++ b/opendc-format/src/test/kotlin/org/opendc/format/trace/wtf/WtfTraceReaderTest.kt
@@ -24,6 +24,7 @@ package org.opendc.format.trace.wtf
 
 import org.junit.jupiter.api.Assertions.*
 import org.junit.jupiter.api.Test
+import java.io.File
 
 /**
  * Test suite for the [WtfTraceReader] class.
@@ -34,7 +35,7 @@ class WtfTraceReaderTest {
      */
     @Test
     fun testParseWtf() {
-        val reader = WtfTraceReader("src/test/resources/wtf-trace")
+        val reader = WtfTraceReader(File("src/test/resources/wtf-trace"))
         var entry = reader.next()
         assertEquals(0, entry.start)
         assertEquals(23, entry.workload.tasks.size)
diff --git a/opendc-format/src/test/kotlin/org/opendc/format/util/ParquetTest.kt b/opendc-format/src/test/kotlin/org/opendc/format/util/ParquetTest.kt
new file mode 100644
index 00000000..e496dd96
--- /dev/null
+++ b/opendc-format/src/test/kotlin/org/opendc/format/util/ParquetTest.kt
@@ -0,0 +1,125 @@
+/*
+ * Copyright (c) 2021 AtLarge Research
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+package org.opendc.format.util
+
+import org.apache.avro.SchemaBuilder
+import org.apache.avro.generic.GenericData
+import org.apache.parquet.avro.AvroParquetReader
+import org.apache.parquet.avro.AvroParquetWriter
+import org.apache.parquet.hadoop.ParquetFileWriter
+import org.junit.jupiter.api.*
+import org.junit.jupiter.api.Assertions.assertEquals
+import java.io.File
+import java.nio.file.FileAlreadyExistsException
+import java.nio.file.NoSuchFileException
+
+/**
+ * Test suite for the Parquet helper classes.
+ */
+internal class ParquetTest {
+    private val schema = SchemaBuilder
+        .record("test")
+        .namespace("org.opendc.format.util")
+        .fields()
+        .name("field").type().intType().noDefault()
+        .endRecord()
+
+    private lateinit var file: File
+
+    /**
+     * Setup the test
+     */
+    @BeforeEach
+    fun setUp() {
+        file = File.createTempFile("opendc", "parquet")
+    }
+
+    /**
+     * Tear down the test.
+     */
+    @AfterEach
+    fun tearDown() {
+        file.delete()
+    }
+
+    /**
+     * Initial test to verify whether the Parquet writer works.
+     */
+    @Test
+    fun testSmoke() {
+        val n = 4
+        val writer = AvroParquetWriter.builder<GenericData.Record>(LocalOutputFile(file))
+            .withSchema(schema)
+            .withWriteMode(ParquetFileWriter.Mode.OVERWRITE)
+            .build()
+
+        try {
+            repeat(n) { i ->
+                val record = GenericData.Record(schema)
+                record.put("field", i)
+                writer.write(record)
+            }
+        } finally {
+            writer.close()
+        }
+
+        val reader = AvroParquetReader.builder<GenericData.Record>(LocalInputFile(file))
+            .build()
+
+        var counter = 0
+        try {
+            while (true) {
+                val record = reader.read() ?: break
+                assertEquals(counter++, record.get("field"))
+            }
+        } finally {
+            reader.close()
+        }
+
+        assertEquals(n, counter)
+    }
+
+    /**
+     * Test if overwriting fails if not specified.
+     */
+    @Test
+    fun testOverwrite() {
+        assertThrows<FileAlreadyExistsException> {
+            AvroParquetWriter.builder<GenericData.Record>(LocalOutputFile(file))
+                .withSchema(schema)
+                .build()
+        }
+    }
+
+    /**
+     * Test non-existent file.
+     */
+    @Test
+    fun testNonExistent() {
+        file.delete()
+        assertThrows<NoSuchFileException> {
+            AvroParquetReader.builder<GenericData.Record>(LocalInputFile(file))
+                .build()
+        }
+    }
+}