Merge pull request #1948 from vjpai/poisson

Open loop sync/async multithreaded testing

8 files changed, 725 insertions, 63 deletions

diff --git a/test/cpp/qps/client.h b/test/cpp/qps/client.h
index dc3a9f2ac5..28cd32a197 100644
--- a/test/cpp/qps/client.h
+++ b/test/cpp/qps/client.h
@@ -35,6 +35,7 @@
 #define TEST_QPS_CLIENT_H
 
 #include "test/cpp/qps/histogram.h"
+#include "test/cpp/qps/interarrival.h"
 #include "test/cpp/qps/timer.h"
 #include "test/cpp/qps/qpstest.grpc.pb.h"
 
@@ -42,11 +43,31 @@
 #include <mutex>
 
 namespace grpc {
+
+#if defined(__APPLE__)
+// Specialize Timepoint for high res clock as we need that
+template <>
+class TimePoint<std::chrono::high_resolution_clock::time_point> {
+ public:
+  TimePoint(const std::chrono::high_resolution_clock::time_point& time) {
+    TimepointHR2Timespec(time, &time_);
+  }
+  gpr_timespec raw_time() const { return time_; }
+
+ private:
+  gpr_timespec time_;
+};
+#endif
+
 namespace testing {
 
+typedef std::chrono::high_resolution_clock grpc_time_source;
+typedef std::chrono::time_point<grpc_time_source> grpc_time;
+
 class Client {
  public:
-  explicit Client(const ClientConfig& config) : timer_(new Timer) {
+  explicit Client(const ClientConfig& config)
+      : timer_(new Timer), interarrival_timer_() {
     for (int i = 0; i < config.client_channels(); i++) {
       channels_.push_back(ClientChannelInfo(
           config.server_targets(i % config.server_targets_size()), config));
@@ -81,6 +102,7 @@ class Client {
 
  protected:
   SimpleRequest request_;
+  bool closed_loop_;
 
   class ClientChannelInfo {
    public:
@@ -106,6 +128,61 @@ class Client {
 
   virtual bool ThreadFunc(Histogram* histogram, size_t thread_idx) = 0;
 
+  void SetupLoadTest(const ClientConfig& config, size_t num_threads) {
+    // Set up the load distribution based on the number of threads
+    if (config.load_type() == CLOSED_LOOP) {
+      closed_loop_ = true;
+    } else {
+      closed_loop_ = false;
+
+      std::unique_ptr<RandomDist> random_dist;
+      const auto& load = config.load_params();
+      switch (config.load_type()) {
+        case POISSON:
+          random_dist.reset(
+              new ExpDist(load.poisson().offered_load() / num_threads));
+          break;
+        case UNIFORM:
+          random_dist.reset(
+              new UniformDist(load.uniform().interarrival_lo() * num_threads,
+                              load.uniform().interarrival_hi() * num_threads));
+          break;
+        case DETERMINISTIC:
+          random_dist.reset(
+              new DetDist(num_threads / load.determ().offered_load()));
+          break;
+        case PARETO:
+          random_dist.reset(
+              new ParetoDist(load.pareto().interarrival_base() * num_threads,
+                             load.pareto().alpha()));
+          break;
+        default:
+          GPR_ASSERT(false);
+          break;
+      }
+
+      interarrival_timer_.init(*random_dist, num_threads);
+      for (size_t i = 0; i < num_threads; i++) {
+        next_time_.push_back(
+            grpc_time_source::now() +
+            std::chrono::duration_cast<grpc_time_source::duration>(
+                interarrival_timer_(i)));
+      }
+    }
+  }
+
+  bool NextIssueTime(int thread_idx, grpc_time* time_delay) {
+    if (closed_loop_) {
+      return false;
+    } else {
+      *time_delay = next_time_[thread_idx];
+      next_time_[thread_idx] +=
+          std::chrono::duration_cast<grpc_time_source::duration>(
+              interarrival_timer_(thread_idx));
+      return true;
+    }
+  }
+
  private:
   class Thread {
    public:
@@ -168,6 +245,9 @@ class Client {
 
   std::vector<std::unique_ptr<Thread>> threads_;
   std::unique_ptr<Timer> timer_;
+
+  InterarrivalTimer interarrival_timer_;
+  std::vector<grpc_time> next_time_;
 };
 
 std::unique_ptr<Client> CreateSynchronousUnaryClient(const ClientConfig& args);
diff --git a/test/cpp/qps/client_async.cc b/test/cpp/qps/client_async.cc
index 00bbd8a8a0..a238b60a9c 100644
--- a/test/cpp/qps/client_async.cc
+++ b/test/cpp/qps/client_async.cc
@@ -32,8 +32,11 @@
  */
 
 #include <cassert>
+#include <forward_list>
 #include <functional>
+#include <list>
 #include <memory>
+#include <mutex>
 #include <string>
 #include <thread>
 #include <vector>
@@ -55,38 +58,55 @@
 namespace grpc {
 namespace testing {
 
+typedef std::list<grpc_time> deadline_list;
+
 class ClientRpcContext {
  public:
-  ClientRpcContext() {}
+  ClientRpcContext(int ch) : channel_id_(ch) {}
   virtual ~ClientRpcContext() {}
   // next state, return false if done. Collect stats when appropriate
   virtual bool RunNextState(bool, Histogram* hist) = 0;
-  virtual void StartNewClone() = 0;
+  virtual ClientRpcContext* StartNewClone() = 0;
   static void* tag(ClientRpcContext* c) { return reinterpret_cast<void*>(c); }
   static ClientRpcContext* detag(void* t) {
     return reinterpret_cast<ClientRpcContext*>(t);
   }
+
+  deadline_list::iterator deadline_posn() const { return deadline_posn_; }
+  void set_deadline_posn(const deadline_list::iterator& it) {
+    deadline_posn_ = it;
+  }
+  virtual void Start(CompletionQueue* cq) = 0;
+  int channel_id() const { return channel_id_; }
+
+ protected:
+  int channel_id_;
+
+ private:
+  deadline_list::iterator deadline_posn_;
 };
 
 template <class RequestType, class ResponseType>
 class ClientRpcContextUnaryImpl : public ClientRpcContext {
  public:
   ClientRpcContextUnaryImpl(
-      TestService::Stub* stub, const RequestType& req,
+      int channel_id, TestService::Stub* stub, const RequestType& req,
       std::function<
           std::unique_ptr<grpc::ClientAsyncResponseReader<ResponseType>>(
-              TestService::Stub*, grpc::ClientContext*, const RequestType&)>
-          start_req,
+              TestService::Stub*, grpc::ClientContext*, const RequestType&,
+              CompletionQueue*)> start_req,
       std::function<void(grpc::Status, ResponseType*)> on_done)
-      : context_(),
+      : ClientRpcContext(channel_id),
+        context_(),
         stub_(stub),
         req_(req),
         response_(),
         next_state_(&ClientRpcContextUnaryImpl::RespDone),
         callback_(on_done),
-        start_req_(start_req),
-        start_(Timer::Now()),
-        response_reader_(start_req(stub_, &context_, req_)) {
+        start_req_(start_req) {}
+  void Start(CompletionQueue* cq) GRPC_OVERRIDE {
+    start_ = Timer::Now();
+    response_reader_ = start_req_(stub_, &context_, req_, cq);
     response_reader_->Finish(&response_, &status_, ClientRpcContext::tag(this));
   }
   ~ClientRpcContextUnaryImpl() GRPC_OVERRIDE {}
@@ -98,8 +118,9 @@ class ClientRpcContextUnaryImpl : public ClientRpcContext {
     return ret;
   }
 
-  void StartNewClone() GRPC_OVERRIDE {
-    new ClientRpcContextUnaryImpl(stub_, req_, start_req_, callback_);
+  ClientRpcContext* StartNewClone() GRPC_OVERRIDE {
+    return new ClientRpcContextUnaryImpl(channel_id_, stub_, req_, start_req_,
+                                         callback_);
   }
 
  private:
@@ -109,7 +130,7 @@ class ClientRpcContextUnaryImpl : public ClientRpcContext {
   }
   bool DoCallBack(bool) {
     callback_(status_, &response_);
-    return false;
+    return true;  // we're done, this'll be ignored
   }
   grpc::ClientContext context_;
   TestService::Stub* stub_;
@@ -118,29 +139,54 @@ class ClientRpcContextUnaryImpl : public ClientRpcContext {
   bool (ClientRpcContextUnaryImpl::*next_state_)(bool);
   std::function<void(grpc::Status, ResponseType*)> callback_;
   std::function<std::unique_ptr<grpc::ClientAsyncResponseReader<ResponseType>>(
-      TestService::Stub*, grpc::ClientContext*, const RequestType&)> start_req_;
+      TestService::Stub*, grpc::ClientContext*, const RequestType&,
+      CompletionQueue*)> start_req_;
   grpc::Status status_;
   double start_;
   std::unique_ptr<grpc::ClientAsyncResponseReader<ResponseType>>
       response_reader_;
 };
 
+typedef std::forward_list<ClientRpcContext*> context_list;
+
 class AsyncClient : public Client {
  public:
-  explicit AsyncClient(const ClientConfig& config,
-                       std::function<void(CompletionQueue*, TestService::Stub*,
-                                          const SimpleRequest&)> setup_ctx)
-      : Client(config) {
+  explicit AsyncClient(
+      const ClientConfig& config,
+      std::function<ClientRpcContext*(int, TestService::Stub*,
+                                      const SimpleRequest&)> setup_ctx)
+      : Client(config),
+        channel_lock_(config.client_channels()),
+        contexts_(config.client_channels()),
+        max_outstanding_per_channel_(config.outstanding_rpcs_per_channel()),
+        channel_count_(config.client_channels()),
+        pref_channel_inc_(config.async_client_threads()) {
+    SetupLoadTest(config, config.async_client_threads());
+
     for (int i = 0; i < config.async_client_threads(); i++) {
       cli_cqs_.emplace_back(new CompletionQueue);
+      if (!closed_loop_) {
+        rpc_deadlines_.emplace_back();
+        next_channel_.push_back(i % channel_count_);
+        issue_allowed_.push_back(true);
+
+        grpc_time next_issue;
+        NextIssueTime(i, &next_issue);
+        next_issue_.push_back(next_issue);
+      }
     }
+
     int t = 0;
     for (int i = 0; i < config.outstanding_rpcs_per_channel(); i++) {
-      for (auto channel = channels_.begin(); channel != channels_.end();
-           channel++) {
+      for (int ch = 0; ch < channel_count_; ch++) {
         auto* cq = cli_cqs_[t].get();
         t = (t + 1) % cli_cqs_.size();
-        setup_ctx(cq, channel->get_stub(), request_);
+        auto ctx = setup_ctx(ch, channels_[ch].get_stub(), request_);
+        if (closed_loop_) {
+          ctx->Start(cq);
+        } else {
+          contexts_[ch].push_front(ctx);
+        }
       }
     }
   }
@@ -159,30 +205,126 @@ class AsyncClient : public Client {
                   size_t thread_idx) GRPC_OVERRIDE GRPC_FINAL {
     void* got_tag;
     bool ok;
-    switch (cli_cqs_[thread_idx]->AsyncNext(
-        &got_tag, &ok,
-        std::chrono::system_clock::now() + std::chrono::seconds(1))) {
+    grpc_time deadline, short_deadline;
+    if (closed_loop_) {
+      deadline = grpc_time_source::now() + std::chrono::seconds(1);
+      short_deadline = deadline;
+    } else {
+      if (rpc_deadlines_[thread_idx].empty()) {
+        deadline = grpc_time_source::now() + std::chrono::seconds(1);
+      } else {
+        deadline = *(rpc_deadlines_[thread_idx].begin());
+      }
+      short_deadline =
+          issue_allowed_[thread_idx] ? next_issue_[thread_idx] : deadline;
+    }
+
+    bool got_event;
+
+    switch (cli_cqs_[thread_idx]->AsyncNext(&got_tag, &ok, short_deadline)) {
       case CompletionQueue::SHUTDOWN:
         return false;
       case CompletionQueue::TIMEOUT:
-        return true;
+        got_event = false;
+        break;
       case CompletionQueue::GOT_EVENT:
+        got_event = true;
+        break;
+      default:
+        GPR_ASSERT(false);
         break;
     }
-
-    ClientRpcContext* ctx = ClientRpcContext::detag(got_tag);
-    if (ctx->RunNextState(ok, histogram) == false) {
-      // call the callback and then delete it
-      ctx->RunNextState(ok, histogram);
-      ctx->StartNewClone();
-      delete ctx;
+    if ((closed_loop_ || !rpc_deadlines_[thread_idx].empty()) &&
+        grpc_time_source::now() > deadline) {
+      // we have missed some 1-second deadline, which is too much
+      gpr_log(GPR_INFO, "Missed an RPC deadline, giving up");
+      return false;
+    }
+    if (got_event) {
+      ClientRpcContext* ctx = ClientRpcContext::detag(got_tag);
+      if (ctx->RunNextState(ok, histogram) == false) {
+        // call the callback and then clone the ctx
+        ctx->RunNextState(ok, histogram);
+        ClientRpcContext* clone_ctx = ctx->StartNewClone();
+        if (closed_loop_) {
+          clone_ctx->Start(cli_cqs_[thread_idx].get());
+        } else {
+          // Remove the entry from the rpc deadlines list
+          rpc_deadlines_[thread_idx].erase(ctx->deadline_posn());
+          // Put the clone_ctx in the list of idle contexts for this channel
+          // Under lock
+          int ch = clone_ctx->channel_id();
+          std::lock_guard<std::mutex> g(channel_lock_[ch]);
+          contexts_[ch].push_front(clone_ctx);
+        }
+        // delete the old version
+        delete ctx;
+      }
+      if (!closed_loop_)
+        issue_allowed_[thread_idx] =
+            true;  // may be ok now even if it hadn't been
+    }
+    if (!closed_loop_ && issue_allowed_[thread_idx] &&
+        grpc_time_source::now() >= next_issue_[thread_idx]) {
+      // Attempt to issue
+      bool issued = false;
+      for (int num_attempts = 0, channel_attempt = next_channel_[thread_idx];
+           num_attempts < channel_count_ && !issued; num_attempts++) {
+        bool can_issue = false;
+        ClientRpcContext* ctx = nullptr;
+        {
+          std::lock_guard<std::mutex> g(channel_lock_[channel_attempt]);
+          if (!contexts_[channel_attempt].empty()) {
+            // Get an idle context from the front of the list
+            ctx = *(contexts_[channel_attempt].begin());
+            contexts_[channel_attempt].pop_front();
+            can_issue = true;
+          }
+        }
+        if (can_issue) {
+          // do the work to issue
+          rpc_deadlines_[thread_idx].emplace_back(grpc_time_source::now() +
+                                                  std::chrono::seconds(1));
+          auto it = rpc_deadlines_[thread_idx].end();
+          --it;
+          ctx->set_deadline_posn(it);
+          ctx->Start(cli_cqs_[thread_idx].get());
+          issued = true;
+          // If we did issue, then next time, try our thread's next
+          // preferred channel
+          next_channel_[thread_idx] += pref_channel_inc_;
+          if (next_channel_[thread_idx] >= channel_count_)
+            next_channel_[thread_idx] = (thread_idx % channel_count_);
+        } else {
+          // Do a modular increment of channel attempt if we couldn't issue
+          channel_attempt = (channel_attempt + 1) % channel_count_;
+        }
+      }
+      if (issued) {
+        // We issued one; see when we can issue the next
+        grpc_time next_issue;
+        NextIssueTime(thread_idx, &next_issue);
+        next_issue_[thread_idx] = next_issue;
+      } else {
+        issue_allowed_[thread_idx] = false;
+      }
     }
-
     return true;
   }
 
  private:
   std::vector<std::unique_ptr<CompletionQueue>> cli_cqs_;
+
+  std::vector<deadline_list> rpc_deadlines_;  // per thread deadlines
+  std::vector<int> next_channel_;      // per thread round-robin channel ctr
+  std::vector<bool> issue_allowed_;    // may this thread attempt to issue
+  std::vector<grpc_time> next_issue_;  // when should it issue?
+
+  std::vector<std::mutex> channel_lock_;
+  std::vector<context_list> contexts_;  // per-channel list of idle contexts
+  int max_outstanding_per_channel_;
+  int channel_count_;
+  int pref_channel_inc_;
 };
 
 class AsyncUnaryClient GRPC_FINAL : public AsyncClient {
@@ -194,15 +336,15 @@ class AsyncUnaryClient GRPC_FINAL : public AsyncClient {
   ~AsyncUnaryClient() GRPC_OVERRIDE { EndThreads(); }
 
  private:
-  static void SetupCtx(CompletionQueue* cq, TestService::Stub* stub,
-                       const SimpleRequest& req) {
+  static ClientRpcContext* SetupCtx(int channel_id, TestService::Stub* stub,
+                                    const SimpleRequest& req) {
     auto check_done = [](grpc::Status s, SimpleResponse* response) {};
-    auto start_req = [cq](TestService::Stub* stub, grpc::ClientContext* ctx,
-                          const SimpleRequest& request) {
+    auto start_req = [](TestService::Stub* stub, grpc::ClientContext* ctx,
+                        const SimpleRequest& request, CompletionQueue* cq) {
       return stub->AsyncUnaryCall(ctx, request, cq);
     };
-    new ClientRpcContextUnaryImpl<SimpleRequest, SimpleResponse>(
-        stub, req, start_req, check_done);
+    return new ClientRpcContextUnaryImpl<SimpleRequest, SimpleResponse>(
+        channel_id, stub, req, start_req, check_done);
   }
 };
 
@@ -210,26 +352,30 @@ template <class RequestType, class ResponseType>
 class ClientRpcContextStreamingImpl : public ClientRpcContext {
  public:
   ClientRpcContextStreamingImpl(
-      TestService::Stub* stub, const RequestType& req,
-      std::function<std::unique_ptr<
-          grpc::ClientAsyncReaderWriter<RequestType, ResponseType>>(
-          TestService::Stub*, grpc::ClientContext*, void*)> start_req,
+      int channel_id, TestService::Stub* stub, const RequestType& req,
+      std::function<std::unique_ptr<grpc::ClientAsyncReaderWriter<
+          RequestType, ResponseType>>(TestService::Stub*, grpc::ClientContext*,
+                                      CompletionQueue*, void*)> start_req,
       std::function<void(grpc::Status, ResponseType*)> on_done)
-      : context_(),
+      : ClientRpcContext(channel_id),
+        context_(),
         stub_(stub),
         req_(req),
         response_(),
         next_state_(&ClientRpcContextStreamingImpl::ReqSent),
         callback_(on_done),
         start_req_(start_req),
-        start_(Timer::Now()),
-        stream_(start_req_(stub_, &context_, ClientRpcContext::tag(this))) {}
+        start_(Timer::Now()) {}
   ~ClientRpcContextStreamingImpl() GRPC_OVERRIDE {}
   bool RunNextState(bool ok, Histogram* hist) GRPC_OVERRIDE {
     return (this->*next_state_)(ok, hist);
   }
-  void StartNewClone() GRPC_OVERRIDE {
-    new ClientRpcContextStreamingImpl(stub_, req_, start_req_, callback_);
+  ClientRpcContext* StartNewClone() GRPC_OVERRIDE {
+    return new ClientRpcContextStreamingImpl(channel_id_, stub_, req_,
+                                             start_req_, callback_);
+  }
+  void Start(CompletionQueue* cq) GRPC_OVERRIDE {
+    stream_ = start_req_(stub_, &context_, cq, ClientRpcContext::tag(this));
   }
 
  private:
@@ -263,7 +409,8 @@ class ClientRpcContextStreamingImpl : public ClientRpcContext {
   std::function<void(grpc::Status, ResponseType*)> callback_;
   std::function<
       std::unique_ptr<grpc::ClientAsyncReaderWriter<RequestType, ResponseType>>(
-          TestService::Stub*, grpc::ClientContext*, void*)> start_req_;
+          TestService::Stub*, grpc::ClientContext*, CompletionQueue*, void*)>
+      start_req_;
   grpc::Status status_;
   double start_;
   std::unique_ptr<grpc::ClientAsyncReaderWriter<RequestType, ResponseType>>
@@ -274,22 +421,25 @@ class AsyncStreamingClient GRPC_FINAL : public AsyncClient {
  public:
   explicit AsyncStreamingClient(const ClientConfig& config)
       : AsyncClient(config, SetupCtx) {
+    // async streaming currently only supported closed loop
+    GPR_ASSERT(config.load_type() == CLOSED_LOOP);
+
     StartThreads(config.async_client_threads());
   }
 
   ~AsyncStreamingClient() GRPC_OVERRIDE { EndThreads(); }
 
  private:
-  static void SetupCtx(CompletionQueue* cq, TestService::Stub* stub,
-                       const SimpleRequest& req) {
+  static ClientRpcContext* SetupCtx(int channel_id, TestService::Stub* stub,
+                                    const SimpleRequest& req) {
     auto check_done = [](grpc::Status s, SimpleResponse* response) {};
-    auto start_req = [cq](TestService::Stub* stub, grpc::ClientContext* ctx,
-                          void* tag) {
+    auto start_req = [](TestService::Stub* stub, grpc::ClientContext* ctx,
+                        CompletionQueue* cq, void* tag) {
       auto stream = stub->AsyncStreamingCall(ctx, cq, tag);
       return stream;
     };
-    new ClientRpcContextStreamingImpl<SimpleRequest, SimpleResponse>(
-        stub, req, start_req, check_done);
+    return new ClientRpcContextStreamingImpl<SimpleRequest, SimpleResponse>(
+        channel_id, stub, req, start_req, check_done);
   }
 };
 
diff --git a/test/cpp/qps/client_sync.cc b/test/cpp/qps/client_sync.cc
index c28dc91321..d1682caf06 100644
--- a/test/cpp/qps/client_sync.cc
+++ b/test/cpp/qps/client_sync.cc
@@ -32,6 +32,7 @@
  */
 
 #include <cassert>
+#include <chrono>
 #include <memory>
 #include <mutex>
 #include <string>
@@ -57,6 +58,7 @@
 #include "test/cpp/qps/client.h"
 #include "test/cpp/qps/qpstest.grpc.pb.h"
 #include "test/cpp/qps/histogram.h"
+#include "test/cpp/qps/interarrival.h"
 #include "test/cpp/qps/timer.h"
 
 namespace grpc {
@@ -68,11 +70,19 @@ class SynchronousClient : public Client {
     num_threads_ =
         config.outstanding_rpcs_per_channel() * config.client_channels();
     responses_.resize(num_threads_);
+    SetupLoadTest(config, num_threads_);
   }
 
   virtual ~SynchronousClient(){};
 
  protected:
+  void WaitToIssue(int thread_idx) {
+    grpc_time next_time;
+    if (NextIssueTime(thread_idx, &next_time)) {
+      std::this_thread::sleep_until(next_time);
+    }
+  }
+
   size_t num_threads_;
   std::vector<SimpleResponse> responses_;
 };
@@ -86,6 +96,7 @@ class SynchronousUnaryClient GRPC_FINAL : public SynchronousClient {
   ~SynchronousUnaryClient() { EndThreads(); }
 
   bool ThreadFunc(Histogram* histogram, size_t thread_idx) GRPC_OVERRIDE {
+    WaitToIssue(thread_idx);
     auto* stub = channels_[thread_idx % channels_.size()].get_stub();
     double start = Timer::Now();
     grpc::ClientContext context;
@@ -119,6 +130,7 @@ class SynchronousStreamingClient GRPC_FINAL : public SynchronousClient {
   }
 
   bool ThreadFunc(Histogram* histogram, size_t thread_idx) GRPC_OVERRIDE {
+    WaitToIssue(thread_idx);
     double start = Timer::Now();
     if (stream_[thread_idx]->Write(request_) &&
         stream_[thread_idx]->Read(&responses_[thread_idx])) {
diff --git a/test/cpp/qps/interarrival.h b/test/cpp/qps/interarrival.h
new file mode 100644
index 0000000000..f90a17a894
--- /dev/null
+++ b/test/cpp/qps/interarrival.h
@@ -0,0 +1,178 @@
+/*
+ *
+ * Copyright 2015, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef TEST_QPS_INTERARRIVAL_H
+#define TEST_QPS_INTERARRIVAL_H
+
+#include <chrono>
+#include <cmath>
+#include <random>
+
+#include <grpc++/config.h>
+
+namespace grpc {
+namespace testing {
+
+// First create classes that define a random distribution
+// Note that this code does not include C++-specific random distribution
+// features supported in std::random. Although this would make this code easier,
+// this code is required to serve as the template code for other language
+// stacks. Thus, this code only uses a uniform distribution of doubles [0,1)
+// and then provides the distribution functions itself.
+
+class RandomDist {
+ public:
+  RandomDist() {}
+  virtual ~RandomDist() = 0;
+  // Argument to operator() is a uniform double in the range [0,1)
+  virtual double operator()(double uni) const = 0;
+};
+
+inline RandomDist::~RandomDist() {}
+
+// ExpDist implements an exponential distribution, which is the
+// interarrival distribution for a Poisson process. The parameter
+// lambda is the mean rate of arrivals. This is the
+// most useful distribution since it is actually additive and
+// memoryless. It is a good representation of activity coming in from
+// independent identical stationary sources. For more information,
+// see http://en.wikipedia.org/wiki/Exponential_distribution
+
+class ExpDist GRPC_FINAL : public RandomDist {
+ public:
+  explicit ExpDist(double lambda) : lambda_recip_(1.0 / lambda) {}
+  ~ExpDist() GRPC_OVERRIDE {}
+  double operator()(double uni) const GRPC_OVERRIDE {
+    // Note: Use 1.0-uni above to avoid NaN if uni is 0
+    return lambda_recip_ * (-log(1.0 - uni));
+  }
+
+ private:
+  double lambda_recip_;
+};
+
+// UniformDist implements a random distribution that has
+// interarrival time uniformly spread between [lo,hi). The
+// mean interarrival time is (lo+hi)/2. For more information,
+// see http://en.wikipedia.org/wiki/Uniform_distribution_%28continuous%29
+
+class UniformDist GRPC_FINAL : public RandomDist {
+ public:
+  UniformDist(double lo, double hi) : lo_(lo), range_(hi - lo) {}
+  ~UniformDist() GRPC_OVERRIDE {}
+  double operator()(double uni) const GRPC_OVERRIDE {
+    return uni * range_ + lo_;
+  }
+
+ private:
+  double lo_;
+  double range_;
+};
+
+// DetDist provides a random distribution with interarrival time
+// of val. Note that this is not additive, so using this on multiple
+// flows of control (threads within the same client or separate
+// clients) will not preserve any deterministic interarrival gap across
+// requests.
+
+class DetDist GRPC_FINAL : public RandomDist {
+ public:
+  explicit DetDist(double val) : val_(val) {}
+  ~DetDist() GRPC_OVERRIDE {}
+  double operator()(double uni) const GRPC_OVERRIDE { return val_; }
+
+ private:
+  double val_;
+};
+
+// ParetoDist provides a random distribution with interarrival time
+// spread according to a Pareto (heavy-tailed) distribution. In this
+// model, many interarrival times are close to the base, but a sufficient
+// number will be high (up to infinity) as to disturb the mean. It is a
+// good representation of the response times of data center jobs. See
+// http://en.wikipedia.org/wiki/Pareto_distribution
+
+class ParetoDist GRPC_FINAL : public RandomDist {
+ public:
+  ParetoDist(double base, double alpha)
+      : base_(base), alpha_recip_(1.0 / alpha) {}
+  ~ParetoDist() GRPC_OVERRIDE {}
+  double operator()(double uni) const GRPC_OVERRIDE {
+    // Note: Use 1.0-uni above to avoid div by zero if uni is 0
+    return base_ / pow(1.0 - uni, alpha_recip_);
+  }
+
+ private:
+  double base_;
+  double alpha_recip_;
+};
+
+// A class library for generating pseudo-random interarrival times
+// in an efficient re-entrant way. The random table is built at construction
+// time, and each call must include the thread id of the invoker
+
+typedef std::default_random_engine qps_random_engine;
+
+class InterarrivalTimer {
+ public:
+  InterarrivalTimer() {}
+  void init(const RandomDist& r, int threads, int entries = 1000000) {
+    qps_random_engine gen;
+    std::uniform_real_distribution<double> uniform(0.0, 1.0);
+    for (int i = 0; i < entries; i++) {
+      random_table_.push_back(std::chrono::nanoseconds(
+          static_cast<int64_t>(1e9 * r(uniform(gen)))));
+    }
+    // Now set up the thread positions
+    for (int i = 0; i < threads; i++) {
+      thread_posns_.push_back(random_table_.begin() + (entries * i) / threads);
+    }
+  }
+  virtual ~InterarrivalTimer(){};
+
+  std::chrono::nanoseconds operator()(int thread_num) {
+    auto ret = *(thread_posns_[thread_num]++);
+    if (thread_posns_[thread_num] == random_table_.end())
+      thread_posns_[thread_num] = random_table_.begin();
+    return ret;
+  }
+
+ private:
+  typedef std::vector<std::chrono::nanoseconds> time_table;
+  std::vector<time_table::const_iterator> thread_posns_;
+  time_table random_table_;
+};
+}
+}
+
+#endif
diff --git a/test/cpp/qps/qps_driver.cc b/test/cpp/qps/qps_driver.cc
index 281e2e8119..acc3098839 100644
--- a/test/cpp/qps/qps_driver.cc
+++ b/test/cpp/qps/qps_driver.cc
@@ -63,11 +63,15 @@ DEFINE_int32(client_channels, 1, "Number of client channels");
 DEFINE_int32(payload_size, 1, "Payload size");
 DEFINE_string(client_type, "SYNCHRONOUS_CLIENT", "Client type");
 DEFINE_int32(async_client_threads, 1, "Async client threads");
+DEFINE_string(load_type, "CLOSED_LOOP", "Load type");
+DEFINE_double(load_param_1, 0.0, "Load parameter 1");
+DEFINE_double(load_param_2, 0.0, "Load parameter 2");
 
 using grpc::testing::ClientConfig;
 using grpc::testing::ServerConfig;
 using grpc::testing::ClientType;
 using grpc::testing::ServerType;
+using grpc::testing::LoadType;
 using grpc::testing::RpcType;
 using grpc::testing::ResourceUsage;
 
@@ -80,11 +84,14 @@ static void QpsDriver() {
 
   ClientType client_type;
   ServerType server_type;
+  LoadType load_type;
   GPR_ASSERT(ClientType_Parse(FLAGS_client_type, &client_type));
   GPR_ASSERT(ServerType_Parse(FLAGS_server_type, &server_type));
+  GPR_ASSERT(LoadType_Parse(FLAGS_load_type, &load_type));
 
   ClientConfig client_config;
   client_config.set_client_type(client_type);
+  client_config.set_load_type(load_type);
   client_config.set_enable_ssl(FLAGS_enable_ssl);
   client_config.set_outstanding_rpcs_per_channel(
       FLAGS_outstanding_rpcs_per_channel);
@@ -93,6 +100,40 @@ static void QpsDriver() {
   client_config.set_async_client_threads(FLAGS_async_client_threads);
   client_config.set_rpc_type(rpc_type);
 
+  // set up the load parameters
+  switch (load_type) {
+    case grpc::testing::CLOSED_LOOP:
+      break;
+    case grpc::testing::POISSON: {
+      auto poisson = client_config.mutable_load_params()->mutable_poisson();
+      GPR_ASSERT(FLAGS_load_param_1 != 0.0);
+      poisson->set_offered_load(FLAGS_load_param_1);
+      break;
+    }
+    case grpc::testing::UNIFORM: {
+      auto uniform = client_config.mutable_load_params()->mutable_uniform();
+      GPR_ASSERT(FLAGS_load_param_1 != 0.0);
+      GPR_ASSERT(FLAGS_load_param_2 != 0.0);
+      uniform->set_interarrival_lo(FLAGS_load_param_1 / 1e6);
+      uniform->set_interarrival_hi(FLAGS_load_param_2 / 1e6);
+      break;
+    }
+    case grpc::testing::DETERMINISTIC: {
+      auto determ = client_config.mutable_load_params()->mutable_determ();
+      GPR_ASSERT(FLAGS_load_param_1 != 0.0);
+      determ->set_offered_load(FLAGS_load_param_1);
+      break;
+    }
+    case grpc::testing::PARETO: {
+      auto pareto = client_config.mutable_load_params()->mutable_pareto();
+      GPR_ASSERT(FLAGS_load_param_1 != 0.0);
+      GPR_ASSERT(FLAGS_load_param_2 != 0.0);
+      pareto->set_interarrival_base(FLAGS_load_param_1 / 1e6);
+      pareto->set_alpha(FLAGS_load_param_2);
+      break;
+    }
+  }
+
   ServerConfig server_config;
   server_config.set_server_type(server_type);
   server_config.set_threads(FLAGS_server_threads);
diff --git a/test/cpp/qps/qps_interarrival_test.cc b/test/cpp/qps/qps_interarrival_test.cc
new file mode 100644
index 0000000000..cecd1be03f
--- /dev/null
+++ b/test/cpp/qps/qps_interarrival_test.cc
@@ -0,0 +1,76 @@
+/*
+ *
+ * Copyright 2015, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include "test/cpp/qps/interarrival.h"
+#include <chrono>
+#include <iostream>
+
+// Use the C histogram rather than C++ to avoid depending on proto
+#include <grpc/support/histogram.h>
+#include <grpc++/config.h>
+
+using grpc::testing::RandomDist;
+using grpc::testing::InterarrivalTimer;
+
+void RunTest(RandomDist&& r, int threads, std::string title) {
+  InterarrivalTimer timer;
+  timer.init(r, threads);
+  gpr_histogram *h(gpr_histogram_create(0.01, 60e9));
+
+  for (int i = 0; i < 10000000; i++) {
+    for (int j = 0; j < threads; j++) {
+      gpr_histogram_add(h, timer(j).count());
+    }
+  }
+
+  std::cout << title << " Distribution" << std::endl;
+  std::cout << "Value, Percentile" << std::endl;
+  for (double pct = 0.0; pct < 100.0; pct += 1.0) {
+    std::cout << gpr_histogram_percentile(h, pct) << "," << pct << std::endl;
+  }
+
+  gpr_histogram_destroy(h);
+}
+
+using grpc::testing::ExpDist;
+using grpc::testing::DetDist;
+using grpc::testing::UniformDist;
+using grpc::testing::ParetoDist;
+
+int main(int argc, char **argv) {
+  RunTest(ExpDist(10.0), 5, std::string("Exponential(10)"));
+  RunTest(DetDist(5.0), 5, std::string("Det(5)"));
+  RunTest(UniformDist(0.0, 10.0), 5, std::string("Uniform(1,10)"));
+  RunTest(ParetoDist(1.0, 1.0), 5, std::string("Pareto(1,1)"));
+  return 0;
+}
diff --git a/test/cpp/qps/qps_test_openloop.cc b/test/cpp/qps/qps_test_openloop.cc
new file mode 100644
index 0000000000..0f6d8e8530
--- /dev/null
+++ b/test/cpp/qps/qps_test_openloop.cc
@@ -0,0 +1,87 @@
+/*
+ *
+ * Copyright 2015, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <set>
+
+#include <grpc/support/log.h>
+
+#include <signal.h>
+
+#include "test/cpp/qps/driver.h"
+#include "test/cpp/qps/report.h"
+#include "test/cpp/util/benchmark_config.h"
+
+namespace grpc {
+namespace testing {
+
+static const int WARMUP = 5;
+static const int BENCHMARK = 10;
+
+static void RunQPS() {
+  gpr_log(GPR_INFO, "Running QPS test, open-loop");
+
+  ClientConfig client_config;
+  client_config.set_client_type(ASYNC_CLIENT);
+  client_config.set_enable_ssl(false);
+  client_config.set_outstanding_rpcs_per_channel(1000);
+  client_config.set_client_channels(8);
+  client_config.set_payload_size(1);
+  client_config.set_async_client_threads(8);
+  client_config.set_rpc_type(UNARY);
+  client_config.set_load_type(POISSON);
+  client_config.mutable_load_params()->
+    mutable_poisson()->set_offered_load(10000.0);
+  
+  ServerConfig server_config;
+  server_config.set_server_type(ASYNC_SERVER);
+  server_config.set_enable_ssl(false);
+  server_config.set_threads(4);
+
+  const auto result =
+      RunScenario(client_config, 1, server_config, 1, WARMUP, BENCHMARK, -2);
+
+  GetReporter()->ReportQPSPerCore(*result, server_config);
+  GetReporter()->ReportLatency(*result);
+}
+
+}  // namespace testing
+}  // namespace grpc
+
+int main(int argc, char** argv) {
+  grpc::testing::InitBenchmark(&argc, &argv, true);
+
+  signal(SIGPIPE, SIG_IGN);
+  grpc::testing::RunQPS();
+
+  return 0;
+}
diff --git a/test/cpp/qps/qpstest.proto b/test/cpp/qps/qpstest.proto
index 122a7df1ac..d977c9b48b 100644
--- a/test/cpp/qps/qpstest.proto
+++ b/test/cpp/qps/qpstest.proto
@@ -36,7 +36,7 @@ package grpc.testing;
 
 enum PayloadType {
   // Compressable text format.
-  COMPRESSABLE= 1;
+  COMPRESSABLE = 1;
 
   // Uncompressable binary format.
   UNCOMPRESSABLE = 2;
@@ -92,21 +92,59 @@ enum RpcType {
   STREAMING = 2;
 }
 
+enum LoadType {
+  CLOSED_LOOP = 1;
+  POISSON = 2;
+  UNIFORM = 3;
+  DETERMINISTIC = 4;
+  PARETO = 5;
+}
+
+message PoissonParams {
+  optional double offered_load = 1;
+}
+
+message UniformParams {
+  optional double interarrival_lo = 1;
+  optional double interarrival_hi = 2;
+}
+
+message DeterministicParams {
+  optional double offered_load = 1;
+}
+
+message ParetoParams {
+  optional double interarrival_base = 1;
+  optional double alpha = 2;
+}
+
+message LoadParams {
+  oneof load {
+    PoissonParams poisson = 1;
+    UniformParams uniform = 2;
+    DeterministicParams determ = 3;
+    ParetoParams pareto = 4;
+  };
+}
+
 message ClientConfig {
   repeated string server_targets = 1;
   required ClientType client_type = 2;
-  optional bool enable_ssl = 3 [default=false];
+  optional bool enable_ssl = 3 [default = false];
   required int32 outstanding_rpcs_per_channel = 4;
   required int32 client_channels = 5;
   required int32 payload_size = 6;
   // only for async client:
   optional int32 async_client_threads = 7;
-  optional RpcType rpc_type = 8 [default=UNARY];
+  optional RpcType rpc_type = 8 [default = UNARY];
   optional string host = 9;
+  optional LoadType load_type = 10 [default = CLOSED_LOOP];
+  optional LoadParams load_params = 11;
 }
 
 // Request current stats
-message Mark {}
+message Mark {
+}
 
 message ClientArgs {
   oneof argtype {
@@ -128,8 +166,8 @@ message ClientStatus {
 
 message ServerConfig {
   required ServerType server_type = 1;
-  optional int32 threads = 2 [default=1];
-  optional bool enable_ssl = 3 [default=false];
+  optional int32 threads = 2 [default = 1];
+  optional bool enable_ssl = 3 [default = false];
   optional string host = 4;
 }
 
@@ -148,11 +186,11 @@ message ServerStatus {
 message SimpleRequest {
   // Desired payload type in the response from the server.
   // If response_type is RANDOM, server randomly chooses one from other formats.
-  optional PayloadType response_type = 1 [default=COMPRESSABLE];
+  optional PayloadType response_type = 1 [default = COMPRESSABLE];
 
   // Desired payload size in the response from the server.
   // If response_type is COMPRESSABLE, this denotes the size before compression.
-  optional int32 response_size = 2 [default=0];
+  optional int32 response_size = 2 [default = 0];
 
   // Optional input payload sent along with the request.
   optional Payload payload = 3;