Add Power-of-Two-Choices Peak-EWMA load balancer (p2c)

docs/cn/client.md

@@ -260,6 +260,10 @@ int main() {
 
 locality-aware，优先选择延时低的下游，直到其延时高于其他机器，无需其他设置。实现原理请查看[Locality-aware load balancing](lalb.md)。
 
+### p2c
+
+即power-of-two-choices加peak-EWMA延时评分。每次选择随机采样两台服务器，把请求发给`延时 * (inflight + 1) / 权重`得分较低的那台。其中延时是对尖峰敏感的滑动平均：延时上升立即生效，恢复则按`tau_ms`（默认10秒）衰减。变慢或出错的服务器在一次观察内即被避开，且选择开销与集群规模无关，为O(1)。权重取自实例tag（同wrr，默认为1）。可选参数：`p2c:choices=4`（每次比较4台采样服务器，适合多台机器同时劣化的场景）、`p2c:tau_ms=5000`。
+
 ### c_murmurhash or c_md5
 
 一致性哈希，与简单hash的不同之处在于增加或删除机器时不会使分桶结果剧烈变化，特别适合cache类服务。

docs/cn/rpc_press.md

@@ -33,7 +33,7 @@ json也可以写在文件中，假如./input.json包含了上述两个请求，-
 可选参数:
 
 - -inc: 包含被import的proto文件的路径。rpc_press默认支持import目录下的其他proto文件，但如果proto文件在其他目录，就要通过这个参数指定，多个路径用分号(;)分隔。
-- -lb_policy: 指定负载均衡算法，默认为空，可选项为: rr random la c_murmurhash c_md5，具体见[负载均衡](client.md#负载均衡)。
+- -lb_policy: 指定负载均衡算法，默认为空，可选项为: rr random la p2c c_murmurhash c_md5，具体见[负载均衡](client.md#负载均衡)。
 - -timeout_ms: 设定超时,单位是毫秒(milliseconds),默认是1000(1秒)
 - -max_retry: 最大的重试次数,默认是3, 一般无需修改. brpc的重试行为具体请见[这里](client.md#重试).
 - -protocol: 连接server使用的协议，可选项见[协议](client.md#协议), 默认是baidu_std

docs/en/client.md

@@ -258,6 +258,10 @@ Requirements of instance tag is the same as wrr.
 
 which is locality-aware. Perfer servers with lower latencies, until the latency is higher than others, no other settings. Check out [Locality-aware load balancing](lalb.md) for more details.
 
+### p2c
+
+which is power-of-two-choices with peak-EWMA latency scoring. Each selection samples two random servers and routes to the one with the lower `latency * (inflight + 1) / weight` score, where the latency is a peak-sensitive moving average: an upward spike takes effect immediately while recovery decays over `tau_ms`(default 10s). A slow or failing server is shed within one observation and selection cost is O(1) regardless of cluster size. Weight comes from the instance tag as in wrr(default 1). Optional parameters: `p2c:choices=4`(compare 4 sampled servers instead of 2, useful when many servers degrade at once), `p2c:tau_ms=5000`.
+
 ### c_murmurhash or c_md5
 
 which is consistent hashing. Adding or removing servers does not make destinations of requests change as dramatically as in simple hashing. It's especially suitable for caching services.

src/brpc/global.cpp

@@ -46,6 +46,7 @@
 #include "brpc/policy/randomized_load_balancer.h"
 #include "brpc/policy/weighted_randomized_load_balancer.h"
 #include "brpc/policy/locality_aware_load_balancer.h"
+#include "brpc/policy/p2c_ewma_load_balancer.h"
 #include "brpc/policy/consistent_hashing_load_balancer.h"
 #include "brpc/policy/hasher.h"
 #include "brpc/policy/dynpart_load_balancer.h"
@@ -155,6 +156,7 @@ struct GlobalExtensions {
     RandomizedLoadBalancer randomized_lb;
     WeightedRandomizedLoadBalancer wr_lb;
     LocalityAwareLoadBalancer la_lb;
+    P2CEwmaLoadBalancer p2c_ewma_lb;
     ConsistentHashingLoadBalancer ch_mh_lb;
     ConsistentHashingLoadBalancer ch_md5_lb;
     ConsistentHashingLoadBalancer ch_ketama_lb;
@@ -391,6 +393,7 @@ static void GlobalInitializeOrDieImpl() {
     LoadBalancerExtension()->RegisterOrDie("random", &g_ext->randomized_lb);
     LoadBalancerExtension()->RegisterOrDie("wr", &g_ext->wr_lb);
     LoadBalancerExtension()->RegisterOrDie("la", &g_ext->la_lb);
+    LoadBalancerExtension()->RegisterOrDie("p2c", &g_ext->p2c_ewma_lb);
     LoadBalancerExtension()->RegisterOrDie("c_murmurhash", &g_ext->ch_mh_lb);
     LoadBalancerExtension()->RegisterOrDie("c_md5", &g_ext->ch_md5_lb);
     LoadBalancerExtension()->RegisterOrDie("c_ketama", &g_ext->ch_ketama_lb);