优先时效性，全量落后 pq压缩 cb：term召回 近线召回：预测用户下次刷新，将没进入精算的结果加入缓存。用户发起请求之前就会进行预计算 grg:多样性， buffer, context dnn 与模型相关的服务 强id & 离散dnn，模型退化严重 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 DemoProcessor processor; BthreadGraphExecutor executor; GraphBuilder builder; builder.executor(executor); { auto& v = builder.add_vertex(processor); v.named_depend("a").to("A"); v.named_emit("B").to("B"); v.named_emit("C").to("C"); } builder.finish(); auto graph = builder.build(); auto a = graph->find_data("A"); auto b = graph->find_data("B"); auto c = graph->find_data("C"); }