1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
|
// Copyright 2017 Yahoo Holdings. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
#include <vespa/vespalib/testkit/test_kit.h>
#include <vespa/vespalib/util/stringfmt.h>
#include <vespa/fnet/frt/frt.h>
#include <vespa/vespalib/util/benchmark_timer.h>
#include <thread>
using vespalib::BenchmarkTimer;
struct Rpc : FRT_Invokable {
FastOS_ThreadPool thread_pool;
FNET_Transport transport;
FRT_Supervisor orb;
Rpc(size_t num_threads)
: thread_pool(128 * 1024), transport(num_threads), orb(&transport, &thread_pool) {}
void start() {
ASSERT_TRUE(transport.Start(&thread_pool));
}
uint32_t listen() {
ASSERT_TRUE(orb.Listen(0));
return orb.GetListenPort();
}
FRT_Target *connect(uint32_t port) {
return orb.GetTarget(port);
}
~Rpc() {
transport.ShutDown(true);
thread_pool.Close();
}
};
struct Server : Rpc {
uint32_t port;
Server(size_t num_threads) : Rpc(num_threads), port(listen()) {
init_rpc();
start();
}
void init_rpc() {
FRT_ReflectionBuilder rb(&orb);
rb.DefineMethod("inc", "l", "l", true, FRT_METHOD(Server::rpc_inc), this);
rb.MethodDesc("increment a 64-bit integer");
rb.ParamDesc("in", "an integer (64 bit)");
rb.ReturnDesc("out", "in + 1 (64 bit)");
}
void rpc_inc(FRT_RPCRequest *req) {
FRT_Values ¶ms = *req->GetParams();
FRT_Values &ret = *req->GetReturn();
ret.AddInt64(params[0]._intval64 + 1);
}
};
struct Client : Rpc {
uint32_t port;
Client(size_t num_threads, const Server &server) : Rpc(num_threads), port(server.port) {
start();
}
FRT_Target *connect() { return Rpc::connect(port); }
};
struct Result {
std::vector<double> req_per_sec;
Result(size_t num_threads) : req_per_sec(num_threads, 0.0) {}
double throughput() const {
double sum = 0.0;
for (double sample: req_per_sec) {
sum += sample;
}
return sum;
}
double latency_ms() const {
double avg_req_per_sec = throughput() / req_per_sec.size();
double avg_sec_per_req = 1.0 / avg_req_per_sec;
return avg_sec_per_req * 1000.0;
}
void print() const {
fprintf(stderr, "total throughput: %f req/s\n", throughput());
fprintf(stderr, "average latency : %f ms\n", latency_ms());
}
};
void perform_test(size_t thread_id, Client &client, Result &result) {
uint64_t seq = 0;
FRT_Target *target = client.connect();
FRT_RPCRequest *req = client.orb.AllocRPCRequest();
auto invoke = [&seq, target, &client, &req](){
req = client.orb.AllocRPCRequest(req);
req->SetMethodName("inc");
req->GetParams()->AddInt64(seq);
target->InvokeSync(req, 60.0);
ASSERT_TRUE(req->CheckReturnTypes("l"));
uint64_t ret = req->GetReturn()->GetValue(0)._intval64;
EXPECT_EQUAL(ret, seq + 1);
seq = ret;
};
size_t loop_cnt = 128;
BenchmarkTimer::benchmark(invoke, invoke, 1.0);
BenchmarkTimer timer(3.0);
while (timer.has_budget()) {
timer.before();
for (size_t i = 0; i < loop_cnt; ++i) {
invoke();
}
timer.after();
}
double t = timer.min_time();
BenchmarkTimer::benchmark(invoke, invoke, 1.0);
EXPECT_GREATER_EQUAL(seq, loop_cnt);
result.req_per_sec[thread_id] = double(loop_cnt) / t;
req->SubRef();
target->SubRef();
TEST_BARRIER();
if (thread_id == 0) {
result.print();
}
}
TEST_MT_FFF("parallel rpc with 1/1 transport threads and 128 user threads",
128, Server(1), Client(1, f1), Result(num_threads)) { perform_test(thread_id, f2, f3); }
TEST_MT_FFF("parallel rpc with 1/8 transport threads and 128 user threads",
128, Server(8), Client(1, f1), Result(num_threads)) { perform_test(thread_id, f2, f3); }
TEST_MT_FFF("parallel rpc with 8/1 transport threads and 128 user threads",
128, Server(1), Client(8, f1), Result(num_threads)) { perform_test(thread_id, f2, f3); }
TEST_MT_FFF("parallel rpc with 8/8 transport threads and 128 user threads",
128, Server(8), Client(8, f1), Result(num_threads)) { perform_test(thread_id, f2, f3); }
TEST_MAIN() { TEST_RUN_ALL(); }
|