ABAP使用异步远程RFC实现并行处理
1、使用场景
当开发复杂报表，需要处理大量数据，不管怎么优化计算和查询语句，程序的运行效率还是达不到用户要求，怎么办？
为了解决这个问题，就需要程序实现并行处理。
本文档就是通过异步调用远程RFC的办法，实现对大量数据的计算，以并行的方式，更快的计算出最终结果。
2、代码实现
在实现并行处理时，首先要看系统当前能并行的资源数
"--------------------@斌将军--------------------
"获取服务器标识
CALL 'C_SAPGPARAM'
ID 'NAME' FIELD 'rdisp/myname'
ID 'VALUE' FIELD gv_applserver.
"获取登录/服务器组名称
SELECT SINGLE
classname
FROM rzllitab
INTO gv_classname "Server Group Name
WHERE applserver = gv_applserver
AND grouptype = 'S' . "S:服务器组，空:登陆组
CALL FUNCTION 'SPBT_INITIALIZE'
EXPORTING
group_name = gv_classname
IMPORTING
max_pbt_wps = gv_total "可用资源总数
free_pbt_wps = gv_available "空闲资源数
EXCEPTIONS
invalid_group_name = 1
internal_error = 2
pbt_env_already_initialized = 3
currently_no_resources_avail = 4
no_pbt_resources_found = 5
cant_init_different_pbt_groups = 6
OTHERS = 7.
"--------------------@斌将军--------------------将逻辑处理封装为远程RFC，比如现在要计算1000行数据，每行数据乘以一百万次循环的累加数，然后展示出结果。远程RFC的计算逻辑如下：
"--------------------@斌将军--------------------
FUNCTION ytest001_001.
*"----------------------------------------------------------------------
*"*"本地接口：
*" IMPORTING
*" VALUE(I_INDEX) TYPE I
*" VALUE(I_COUNT) TYPE I
*" EXPORTING
*" VALUE(E_INDEX) TYPE I
*" VALUE(E_COUNT) TYPE I
*"----------------------------------------------------------------------
DO 1000000 TIMES.
i_count = i_count + 1.
ENDDO.
i_count = i_count * i_index.
e_index = i_index.
e_count = i_count.
ENDFUNCTION.
"--------------------@斌将军--------------------在主程序中调用远程RFC，并且添加远程调用的系统报错异常communication_failure和system_failure
"--------------------@斌将军--------------------
CALL FUNCTION 'YTEST001_001' STARTING NEW TASK gv_taskname
DESTINATION IN GROUP gv_classname
PERFORMING frm_ytest ON END OF TASK "调用每条线程的处理方法，接收处理结果
EXPORTING
i_index = ps_alv-index
i_count = ps_alv-count
EXCEPTIONS
communication_failure = 1 MESSAGE lv_message
system_failure = 2 MESSAGE lv_message
resource_failure = 3.
"--------------------@斌将军--------------------并在主程序中接收远程RFC的返回消息
"--------------------@斌将军--------------------
"返回消息处理
FORM frm_ytest USING taskname.
DATA:ls_alv TYPE ty_alv.
"接收处理数据的返回消息
RECEIVE RESULTS FROM FUNCTION 'YTEST001_001'
IMPORTING
e_index = ls_alv-index
e_count = ls_alv-count
EXCEPTIONS
communication_failure = 1 MESSAGE lv_message
system_failure = 2 MESSAGE lv_message.
MODIFY gt_alv FROM ls_alv TRANSPORTING count WHERE index = ls_alv-index.
"已完成进程 + 1
gv_end_jobs = gv_end_jobs + 1.
ENDFORM.
"--------------------@斌将军--------------------完整参考代码：
"--------------------@斌将军--------------------
TYPES:BEGIN OF ty_alv,
index TYPE i,
count TYPE i,
END OF ty_alv.
DATA:gt_alv TYPE TABLE OF ty_alv,
gs_alv TYPE ty_alv.
DATA:gv_open_jobs TYPE i, "开启的进程
gv_jobs TYPE i, "可用的进程
gv_end_jobs TYPE i. "结束的进程
DATA:gv_applserver TYPE rzlli_asvr, "服务器标识 实例名称（用于登录/服务器组分配）
gv_classname TYPE rzlli_apcl, "登录/服务器组名称
gv_taskname TYPE char10, "进程名称
gv_init_flag TYPE char1,
gv_total TYPE i, "可用资源总数
gv_available TYPE i. "空闲资源数
DATA:lv_count TYPE i,
lv_flat TYPE p DECIMALS 2,
lv_message TYPE char200.
"准备1000行测试数据
DO 1000 TIMES.
gs_alv-index = sy-index.
gs_alv-count = 1.
APPEND gs_alv TO gt_alv.
ENDDO.
"获取服务器标识
PERFORM frm_get_server.
"获取服务器组对应的可用资源数
PERFORM frm_get_jobs_available.
"处理每行数据，实际应用时，看如何将数据“分批”
CLEAR:lv_count.
LOOP AT gt_alv INTO gs_alv.
lv_count = lv_count + 1."进程任务计数器
IF gv_open_jobs - gv_end_jobs = gv_jobs."已开进程 - 已结束进程 = 可用进程
WAIT UNTIL gv_open_jobs - gv_end_jobs < gv_jobs."等待 已开进程 - 已结束进程 < 可用进程
ENDIF.
"拼接进程名称
gv_taskname = 'Task' && lv_count.
CONDENSE gv_taskname.
"逻辑处理
PERFORM frm_deal_task USING gs_alv.
CLEAR:gs_alv.
ENDLOOP.
"等待所有的进程执行完毕
WAIT UNTIL gv_end_jobs >= gv_open_jobs.
"展示结果
CALL METHOD cl_demo_output=>display
EXPORTING
data = gt_alv.
*&---------------------------------------------------------------------*
*& Form FRM_GET_SERVER
*&---------------------------------------------------------------------*
* text 获取服务器标识
*----------------------------------------------------------------------*
FORM frm_get_server.
"获取服务器标识
CALL 'C_SAPGPARAM'
ID 'NAME' FIELD 'rdisp/myname'
ID 'VALUE' FIELD gv_applserver.
"获取登录/服务器组名称
SELECT SINGLE
classname
FROM rzllitab
INTO gv_classname "Server Group Name
WHERE applserver = gv_applserver
AND grouptype = 'S' . "S:服务器组，空:登陆组
ENDFORM.
*&---------------------------------------------------------------------*
*& Form FRM_GET_JOBS_AVAILABLE
*&---------------------------------------------------------------------*
* text 获取服务器组对应的可用资源数
*----------------------------------------------------------------------*
FORM frm_get_jobs_available.
gv_jobs = 0."可用资源数
IF gv_init_flag IS INITIAL."第一次获取资源
"资源查询 - 获取最多jobs 个数
CALL FUNCTION 'SPBT_INITIALIZE'
EXPORTING
group_name = gv_classname
IMPORTING
max_pbt_wps = gv_total "可用资源总数
free_pbt_wps = gv_available "空闲资源数
EXCEPTIONS
invalid_group_name = 1
internal_error = 2
pbt_env_already_initialized = 3
currently_no_resources_avail = 4
no_pbt_resources_found = 5
cant_init_different_pbt_groups = 6
OTHERS = 7.
CASE sy-subrc.
WHEN 0.
lv_flat = gv_available * 9 / 10.
gv_jobs = floor( lv_flat ). "拿其中一部分的空闲资源数来执行
IF gv_jobs = 0 AND gv_available = 1.
gv_jobs = 1.
ENDIF.
gv_init_flag = 'X'.
"按照jobs来进行数据拆分
WHEN 1.
* MESSAGE E836."未定义 PBT 服务器组
WHEN 2.
WHEN 3.
* MESSAGE E833."已为组初始化了 PBT 环境
WHEN 4.
* MESSAGE E837."所有服务器正忙：没有可用的资源
WHEN 5.
WHEN 6.
ENDCASE.
"刷新资源数量
ELSE.
"第二次获取资源调用函数
CALL FUNCTION 'SPBT_GET_CURR_RESOURCE_INFO'
IMPORTING
max_pbt_wps = gv_total
free_pbt_wps = gv_available
EXCEPTIONS
internal_error = 1
pbt_env_not_initialized_yet = 2
OTHERS = 3.
CASE sy-subrc .
WHEN 0.
lv_flat = gv_available * 9 / 10.
gv_jobs = floor( lv_flat ). "拿其中一部分的空闲资源数来执行
IF gv_jobs = 0 AND gv_available = 1.
gv_jobs = 1.
ENDIF.
WHEN 1.
WHEN 2.
CLEAR gv_init_flag.
PERFORM frm_get_jobs_available. "获取失败则递归调用
WHEN 3.
ENDCASE.
ENDIF.
ENDFORM. " FRM_GET_JOBS_AVAILABLE
*&---------------------------------------------------------------------*
*& Form FRM_DEAL_TASK
*&---------------------------------------------------------------------*
* text 逻辑处理
*----------------------------------------------------------------------*
FORM frm_deal_task USING ps_alv TYPE ty_alv.
"调用需要并行执行的函数，此函数必须为远程函数
CALL FUNCTION 'YTEST001_001' STARTING NEW TASK gv_taskname
DESTINATION IN GROUP gv_classname
PERFORMING frm_ytest ON END OF TASK "调用每条进程的处理方法，接收处理结果
EXPORTING
i_index = ps_alv-index
i_count = ps_alv-count
EXCEPTIONS
communication_failure = 1 MESSAGE lv_message
system_failure = 2 MESSAGE lv_message
resource_failure = 3.
IF sy-subrc = 0.
gv_open_jobs = gv_open_jobs + 1. "开启进程成功，则已开启变量 + 1
ELSEIF sy-subrc = 3.
WAIT UP TO '0.1' SECONDS.
WAIT UNTIL gv_open_jobs - gv_end_jobs < gv_jobs."等待 已开进程 - 已结束进程 < 可用进程
PERFORM frm_deal_task USING ps_alv."递归调用，重复开启进程
ELSE.
* WRITE: sy-index,sy-subrc,lv_message.
ENDIF.
ENDFORM.
*&---------------------------------------------------------------------*
*& Form FRM_YTEST
*&---------------------------------------------------------------------*
* text 返回消息处理
*----------------------------------------------------------------------*
FORM frm_ytest USING taskname.
DATA:ls_alv TYPE ty_alv.
"接收处理数据的返回消息
RECEIVE RESULTS FROM FUNCTION 'YTEST001_001'
IMPORTING
e_index = ls_alv-index
e_count = ls_alv-count
EXCEPTIONS
communication_failure = 1 MESSAGE lv_message
system_failure = 2 MESSAGE lv_message.
MODIFY gt_alv FROM ls_alv TRANSPORTING count WHERE index = ls_alv-index.
"已完成进程 + 1
gv_end_jobs = gv_end_jobs + 1.
ENDFORM.
"--------------------@斌将军--------------------3、效率对比
常规处理方式的代码：
"--------------------@斌将军--------------------
LOOP AT gt_alv INTO gs_alv.
CALL FUNCTION 'YTEST001_001'
EXPORTING
i_index = gs_alv-index
i_count = gs_alv-count
IMPORTING
e_index = gs_alv-index
e_count = gs_alv-count.
MODIFY gt_alv FROM gs_alv TRANSPORTING count WHERE index = gs_alv-index.
ENDLOOP.
"--------------------@斌将军--------------------耗时对比
并行处理耗时1.7秒，常规处理耗时172秒，效果立竿见影
定期更文，欢迎关注
