/* This file is part of VoltDB.
* Copyright (C) 2008-2014 VoltDB Inc.
*
* 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 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.
*/
/*
* This samples uses the native asynchronous request processing protocol
* to post requests to the VoltDB server, thus leveraging to the maximum
* VoltDB's ability to run requests in parallel on multiple database
* partitions, and multiple servers.
*
* While asynchronous processing is (marginally) more convoluted to work
* with and not adapted to all workloads, it is the preferred interaction
* model to VoltDB as it guarantees blazing performance.
*
* Because there is a risk of 'firehosing' a database cluster (if the
* cluster is too slow (slow or too few CPUs), this sample performs
* self-tuning to target a specific latency (10ms by default).
* This tuning process, as demonstrated here, is important and should be
* part of your pre-launch evalution so you can adequately provision your
* VoltDB cluster with the number of servers required for your needs.
*/
package eng1969;
import java.util.Random;
import java.util.Timer;
import java.util.TimerTask;
import java.util.concurrent.atomic.AtomicLongArray;
import org.voltdb.client.ClientResponse;
import org.voltdb.client.ProcedureCallback;
import org.voltdb.client.exampleutils.AppHelper;
import org.voltdb.client.exampleutils.ClientConnection;
import org.voltdb.client.exampleutils.ClientConnectionPool;
import org.voltdb.client.exampleutils.IRateLimiter;
import org.voltdb.client.exampleutils.LatencyLimiter;
import org.voltdb.client.exampleutils.RateLimiter;
public class AsyncBenchmark
{
// Initialize some common constants and variables
private static final AtomicLongArray TrackingResults = new AtomicLongArray(2);
// Reference to the database connection we will use
private static ClientConnection Con;
// Application entry point
public static void main(String[] args)
{
try
{
// Use the AppHelper utility class to retrieve command line application parameters
// Define parameters and pull from command line
AppHelper apph = new AppHelper(AsyncBenchmark.class.getCanonicalName())
.add("displayinterval", "display_interval_in_seconds", "Interval for performance feedback, in seconds.", 10)
.add("duration", "run_duration_in_seconds", "Benchmark duration, in seconds.", 120)
.add("servers", "comma_separated_server_list", "List of VoltDB servers to connect to.", "localhost")
.add("port", "port_number", "Client port to connect to on cluster nodes.", 21212)
.add("pool-size", "pool_size", "Size of the record pool to operate on - larger sizes will cause a higher insert/update-delete rate.", 100000)
.add("procedure", "procedure_name", "Procedure to call.", "UpdateKey")
.add("wait", "wait_duration", "Wait duration (only when calling one of the Wait procedures), in milliseconds.", 0)
.add("ratelimit", "rate_limit", "Rate limit to start from (number of transactions per second).", 100000)
.add("autotune", "auto_tune", "Flag indicating whether the benchmark should self-tune the transaction rate for a target execution latency (true|false).", "true")
.add("latency-target", "latency_target", "Execution latency to target to tune transaction rate (in milliseconds).", 10.0d)
.add("run-loader", "Run the leveldb loader", "true")
.setArguments(args)
;
// Retrieve parameters
final long displayInterval = apph.longValue("displayinterval");
final long duration = apph.longValue("duration");
final String servers = apph.stringValue("servers");
final int port = apph.intValue("port");
final int poolSize = apph.intValue("pool-size");
final String procedure = apph.stringValue("procedure");
final long wait = apph.intValue("wait");
final long rateLimit = apph.longValue("ratelimit");
final boolean autoTune = apph.booleanValue("autotune");
final double latencyTarget = apph.doubleValue("latency-target");
final String csv = apph.stringValue("stats");
final boolean runLoader = apph.booleanValue("run-loader");
// Validate parameters
apph.validate("duration", (duration > 0))
.validate("pool-size", (duration > 0))
.validate("wait", (wait >= 0))
.validate("ratelimit", (rateLimit > 0))
.validate("latency-target", (latencyTarget > 0))
;
// Display actual parameters, for reference
apph.printActualUsage();
// Get a client connection - we retry for a while in case the server hasn't started yet
Con = ClientConnectionPool.getWithRetry(servers, port);
// Create a Timer task to display performance data on the procedure
Timer timer = new Timer();
timer.scheduleAtFixedRate(new TimerTask()
{
@Override
public void run()
{
System.out.print(Con.getStatistics(procedure));
}
}
, displayInterval*1000l
, displayInterval*1000l
);
// Pick the transaction rate limiter helping object to use based on user request (rate limiting or latency targeting)
IRateLimiter limiter = autoTune ?
new LatencyLimiter(Con, procedure, latencyTarget, rateLimit) :
new RateLimiter(rateLimit);
// Run the loader first.
if (runLoader) {
doLoader(poolSize);
}
// Run the benchmark loop for the requested duration
final long endTime = System.currentTimeMillis() + (1000l * duration);
Random rand = new Random();
while (endTime > System.currentTimeMillis())
{
doBenchmark(procedure, poolSize, rand, wait);
// Use the limiter to throttle client activity
limiter.throttle();
}
// We're done - stop the performance statistics display task
timer.cancel();
// Now print application results:
// 1. Tracking statistics
System.out.printf(
"-------------------------------------------------------------------------------------\n"
+ " Benchmark Results\n"
+ "-------------------------------------------------------------------------------------\n\n"
+ "A total of %d calls was received...\n"
+ " - %,9d Succeeded\n"
+ " - %,9d Failed (Transaction Error)\n"
+ "\n\n"
+ "-------------------------------------------------------------------------------------\n"
, TrackingResults.get(0)+TrackingResults.get(1)
, TrackingResults.get(0)
, TrackingResults.get(1)
);
// 3. Performance statistics (we only care about the procedure that we're benchmarking)
System.out.println(
"\n\n-------------------------------------------------------------------------------------\n"
+ " System Statistics\n"
+ "-------------------------------------------------------------------------------------\n\n");
System.out.print(Con.getStatistics(procedure).toString(false));
// Dump statistics to a CSV file
Con.saveStatistics(csv);
Con.close();
}
catch(Exception x)
{
System.out.println("Exception: " + x);
x.printStackTrace();
}
}
// There are poolSize rowid_groups in the database (default = 100,000)
// Each rowid_group has 1000 unique rowids such that (rowid_group, rowid) is unique.
// This gives 100M records in the DB by default.
private static void doBenchmark(String procedure, final int poolSize, final Random rand, long wait) throws Exception {
final int totalPartitions = 4;
final int maxGroupsPerPartition = 1000;
// ~1% chance of cold data
boolean coldData = (Math.abs(rand.nextLong()) % 100) <= 1;
long rowid_group = coldData ?
Math.abs(rand.nextLong()) % poolSize :
Math.abs(rand.nextLong()) % (totalPartitions * maxGroupsPerPartition);
long rowid = Math.abs(rand.nextLong()) % 1000;
Con.executeAsync(new ProcedureCallback()
{
@Override
public void clientCallback(ClientResponse response) throws Exception
{
// Track the result of the request (Success, Failure)
if (response.getStatus() == ClientResponse.SUCCESS)
TrackingResults.incrementAndGet(0);
else
TrackingResults.incrementAndGet(1);
}
}
, procedure
, rowid
, rowid_group
, new String("ABCDEFGHIJKLMNOPQRSTUVWXYZ").getBytes()
);
}
private static void doLoader(final int poolSize) throws Exception {
for (int i=0; i <= poolSize; i++) {
for (int rowid = 0; rowid < 1000; rowid++) {
Con.executeAsync(new ProcedureCallback()
{
@Override
public void clientCallback(ClientResponse response) throws Exception
{
if (response.getStatus() != ClientResponse.SUCCESS) {
System.out.println("Loader failed with response: " + response.getStatusString());
}
}
}
, "CreateKey"
, Long.valueOf(rowid)
, Long.valueOf(i)
, new String("ABCDEF").getBytes()
);
}
if (i % 10000 == 0) {
System.out.println("Loaded " + i + " groups.");
}
}
}
}