54b2cf747e
The commit has been tested for C/R and Cray operations, and on Odin (SLURM, rsh) and RoadRunner (TM). I tried to update all environments, but obviously could not test them. I know that Windows needs some work, and have highlighted what is know to be needed in the odls process component. This represents a lot of work by Brian, Tim P, Josh, and myself, with much advice from Jeff and others. For posterity, I have appended a copy of the email describing the work that was done: As we have repeatedly noted, the modex operation in MPI_Init is the single greatest consumer of time during startup. To-date, we have executed that operation as an ORTE stage gate that held the process until a startup message containing all required modex (and OOB contact info - see #3 below) info could be sent to it. Each process would send its data to the HNP's registry, which assembled and sent the message when all processes had reported in. In addition, ORTE had taken responsibility for monitoring process status as it progressed through a series of "stage gates". The process reported its status at each gate, and ORTE would then send a "release" message once all procs had reported in. The incoming changes revamp these procedures in three ways: 1. eliminating the ORTE stage gate system and cleanly delineating responsibility between the OMPI and ORTE layers for MPI init/finalize. The modex stage gate (STG1) has been replaced by a collective operation in the modex itself that performs an allgather on the required modex info. The allgather is implemented using the orte_grpcomm framework since the BTL's are not active at that point. At the moment, the grpcomm framework only has a "basic" component analogous to OMPI's "basic" coll framework - I would recommend that the MPI team create additional, more advanced components to improve performance of this step. The other stage gates have been replaced by orte_grpcomm barrier functions. We tried to use MPI barriers instead (since the BTL's are active at that point), but - as we discussed on the telecon - these are not currently true barriers so the job would hang when we fell through while messages were still in process. Note that the grpcomm barrier doesn't actually resolve that problem, but Brian has pointed out that we are unlikely to ever see it violated. Again, you might want to spend a little time on an advanced barrier algorithm as the one in "basic" is very simplistic. Summarizing this change: ORTE no longer tracks process state nor has direct responsibility for synchronizing jobs. This is now done via collective operations within the MPI layer, albeit using ORTE collective communication services. I -strongly- urge the MPI team to implement advanced collective algorithms to improve the performance of this critical procedure. 2. reducing the volume of data exchanged during modex. Data in the modex consisted of the process name, the name of the node where that process is located (expressed as a string), plus a string representation of all contact info. The nodename was required in order for the modex to determine if the process was local or not - in addition, some people like to have it to print pretty error messages when a connection failed. The size of this data has been reduced in three ways: (a) reducing the size of the process name itself. The process name consisted of two 32-bit fields for the jobid and vpid. This is far larger than any current system, or system likely to exist in the near future, can support. Accordingly, the default size of these fields has been reduced to 16-bits, which means you can have 32k procs in each of 32k jobs. Since the daemons must have a vpid, and we require one daemon/node, this also restricts the default configuration to 32k nodes. To support any future "mega-clusters", a configuration option --enable-jumbo-apps has been added. This option increases the jobid and vpid field sizes to 32-bits. Someday, if necessary, someone can add yet another option to increase them to 64-bits, I suppose. (b) replacing the string nodename with an integer nodeid. Since we have one daemon/node, the nodeid corresponds to the local daemon's vpid. This replaces an often lengthy string with only 2 (or at most 4) bytes, a substantial reduction. (c) when the mca param requesting that nodenames be sent to support pretty error messages, a second mca param is now used to request FQDN - otherwise, the domain name is stripped (by default) from the message to save space. If someone wants to combine those into a single param somehow (perhaps with an argument?), they are welcome to do so - I didn't want to alter what people are already using. While these may seem like small savings, they actually amount to a significant impact when aggregated across the entire modex operation. Since every proc must receive the modex data regardless of the collective used to send it, just reducing the size of the process name removes nearly 400MBytes of communication from a 32k proc job (admittedly, much of this comm may occur in parallel). So it does add up pretty quickly. 3. routing RML messages to reduce connections. The default messaging system remains point-to-point - i.e., each proc opens a socket to every proc it communicates with and sends its messages directly. A new option uses the orteds as routers - i.e., each proc only opens a single socket to its local orted. All messages are sent from the proc to the orted, which forwards the message to the orted on the node where the intended recipient proc is located - that orted then forwards the message to its local proc (the recipient). This greatly reduces the connection storm we have encountered during startup. It also has the benefit of removing the sharing of every proc's OOB contact with every other proc. The orted routing tables are populated during launch since every orted gets a map of where every proc is being placed. Each proc, therefore, only needs to know the contact info for its local daemon, which is passed in via the environment when the proc is fork/exec'd by the daemon. This alone removes ~50 bytes/process of communication that was in the current STG1 startup message - so for our 32k proc job, this saves us roughly 32k*50 = 1.6MBytes sent to 32k procs = 51GBytes of messaging. Note that you can use the new routing method by specifying -mca routed tree - if you so desire. This mode will become the default at some point in the future. There are a few minor additional changes in the commit that I'll just note in passing: * propagation of command line mca params to the orteds - fixes ticket #1073. See note there for details. * requiring of "finalize" prior to "exit" for MPI procs - fixes ticket #1144. See note there for details. * cleanup of some stale header files This commit was SVN r16364.
274 строки
10 KiB
C
274 строки
10 KiB
C
/*
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
* University Research and Technology
|
|
* Corporation. All rights reserved.
|
|
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
|
* of Tennessee Research Foundation. All rights
|
|
* reserved.
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
* University of Stuttgart. All rights reserved.
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
* All rights reserved.
|
|
* $COPYRIGHT$
|
|
*
|
|
* Additional copyrights may follow
|
|
*
|
|
* $HEADER$
|
|
*/
|
|
/** @file:
|
|
*/
|
|
|
|
#ifndef MCA_NS_PRIVATE_H
|
|
#define MCA_NS_PRIVATE_H
|
|
|
|
/*
|
|
* includes
|
|
*/
|
|
#include "orte_config.h"
|
|
#include "orte/orte_constants.h"
|
|
|
|
#include "opal/class/opal_list.h"
|
|
#include "opal/mca/mca.h"
|
|
|
|
#include "orte/dss/dss_types.h"
|
|
#include "orte/mca/ns/ns_types.h"
|
|
#include "orte/mca/rml/rml_types.h"
|
|
|
|
/*
|
|
* Global functions for MCA overall collective open and close
|
|
*/
|
|
#if defined(c_plusplus) || defined(__cplusplus)
|
|
extern "C" {
|
|
#endif
|
|
|
|
/* default limits */
|
|
#define ORTE_NS_ARRAY_MAX_SIZE INT_MAX
|
|
#define ORTE_NS_ARRAY_BLOCK_SIZE 512
|
|
/*
|
|
* Internal definitions
|
|
*/
|
|
typedef uint8_t orte_ns_cmd_bitmask_t;
|
|
typedef uint8_t orte_ns_cmd_flag_t;
|
|
|
|
/*
|
|
* packing type definitions
|
|
*/
|
|
/* CAUTION - any changes here must also change corresponding
|
|
* typedefs above and in ns_types.h
|
|
*/
|
|
#define ORTE_NS_CMD ORTE_INT8
|
|
|
|
|
|
#if ORTE_ENABLE_JUMBO_APPS
|
|
#define ORTE_JOBID_T ORTE_INT32
|
|
#define ORTE_VPID_T ORTE_INT32
|
|
#define ORTE_NODEID_T ORTE_INT32
|
|
|
|
#else
|
|
#define ORTE_JOBID_T ORTE_INT16
|
|
#define ORTE_VPID_T ORTE_INT16
|
|
#define ORTE_NODEID_T ORTE_INT16
|
|
#endif
|
|
|
|
|
|
/*
|
|
* define flag values for remote commands - only used internally
|
|
*/
|
|
#define ORTE_NS_CREATE_NODEID_CMD (int8_t) 3
|
|
#define ORTE_NS_GET_NODE_INFO_CMD (int8_t) 4
|
|
#define ORTE_NS_CREATE_JOBID_CMD (int8_t) 5
|
|
#define ORTE_NS_GET_JOB_DESC_CMD (int8_t) 6
|
|
#define ORTE_NS_GET_JOB_CHILD_CMD (int8_t) 7
|
|
#define ORTE_NS_GET_ROOT_JOB_CMD (int8_t) 8
|
|
#define ORTE_NS_GET_PARENT_JOB_CMD (int8_t) 9
|
|
#define ORTE_NS_RESERVE_RANGE_CMD (int8_t) 10
|
|
#define ORTE_NS_ASSIGN_OOB_TAG_CMD (int8_t) 11
|
|
#define ORTE_NS_GET_PEERS_CMD (int8_t) 12
|
|
#define ORTE_NS_DEFINE_DATA_TYPE_CMD (int8_t) 13
|
|
|
|
#define ORTE_NS_DUMP_JOBIDS_CMD (int8_t) 16
|
|
#define ORTE_NS_DUMP_TAGS_CMD (int8_t) 17
|
|
#define ORTE_NS_DUMP_DATATYPES_CMD (int8_t) 18
|
|
#define ORTE_NS_GET_VPID_RANGE_CMD (int8_t) 19
|
|
#define ORTE_NS_GET_JOB_FAMILY_CMD (int8_t) 20
|
|
|
|
|
|
/*
|
|
* Base functions that are common to all implementations - can be overridden
|
|
*/
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_create_process_name(orte_process_name_t **name,
|
|
orte_jobid_t job,
|
|
orte_vpid_t vpid);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_string_to_process_name(orte_process_name_t **name,
|
|
const char* name_string);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_proc_name_string(char **name_string,
|
|
const orte_process_name_t* name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_vpid_string(char **vpid_string, const orte_process_name_t* name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_vpid_to_string(char **vpid_string, const orte_vpid_t vpid);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_string_to_vpid(orte_vpid_t *vpid, const char* vpidstring);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_jobid_string(char **jobid_string, const orte_process_name_t* name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_jobid_to_string(char **jobid_string, const orte_jobid_t jobid);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_string_to_jobid(orte_jobid_t *jobid, const char* jobidstring);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_vpid(orte_vpid_t *vpid, const orte_process_name_t* name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_jobid(orte_jobid_t *jobid, const orte_process_name_t* name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_string_to_nodeid(orte_nodeid_t *nodeid, const char *string);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_convert_nodeid_to_string(char **nodeid_string, const orte_nodeid_t nodeid);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_compare_fields(orte_ns_cmp_bitmask_t fields,
|
|
const orte_process_name_t* name1,
|
|
const orte_process_name_t* name2);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_print_dump(orte_buffer_t *buffer);
|
|
|
|
|
|
/* not available functions */
|
|
ORTE_DECLSPEC int orte_ns_base_module_init_not_available(void);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_create_nodeids_not_available(orte_nodeid_t **nodeids, orte_std_cntr_t *nnodes,
|
|
char **nodename);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_node_info_not_available(char ***nodename,
|
|
orte_std_cntr_t num_nodes, orte_nodeid_t *nodeids);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_create_jobid_not_available(orte_jobid_t *jobid, opal_list_t *attrs);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_job_descendants_not_available(orte_jobid_t** descendants,
|
|
orte_std_cntr_t *num_desc,
|
|
orte_jobid_t job);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_job_children_not_available(orte_jobid_t** children,
|
|
orte_std_cntr_t *num_childs,
|
|
orte_jobid_t job);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_root_job_not_available(orte_jobid_t *root_job, orte_jobid_t job);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_parent_job_not_available(orte_jobid_t *parent, orte_jobid_t job);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_job_family_not_available(orte_jobid_t **family, orte_std_cntr_t *num_members, orte_jobid_t job);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_reserve_range_not_available(orte_jobid_t job,
|
|
orte_vpid_t range,
|
|
orte_vpid_t *startvpid);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_vpid_range_not_available(orte_jobid_t job,
|
|
orte_vpid_t *range);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_assign_rml_tag_not_available(orte_rml_tag_t *tag, char *name);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_define_data_type_not_available(
|
|
const char *name,
|
|
orte_data_type_t *type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_create_my_name_not_available(void);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_get_peers_not_available(orte_process_name_t **procs,
|
|
orte_std_cntr_t *num_procs, opal_list_t *attributes);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_dump_jobs_not_available(void);
|
|
ORTE_DECLSPEC int orte_ns_base_dump_tags_not_available(void);
|
|
ORTE_DECLSPEC int orte_ns_base_dump_datatypes_not_available(void);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_ft_event_not_available(int state);
|
|
|
|
|
|
/* Base functions used everywhere */
|
|
ORTE_DECLSPEC int orte_ns_base_pack_name(orte_buffer_t *buffer, const void *src,
|
|
orte_std_cntr_t num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_pack_nodeid(orte_buffer_t *buffer, const void *src,
|
|
orte_std_cntr_t num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_pack_jobid(orte_buffer_t *buffer, const void *src,
|
|
orte_std_cntr_t num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_pack_vpid(orte_buffer_t *buffer, const void *src,
|
|
orte_std_cntr_t num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_unpack_name(orte_buffer_t *buffer, void *dest,
|
|
orte_std_cntr_t *num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_unpack_nodeid(orte_buffer_t *buffer, void *dest,
|
|
orte_std_cntr_t *num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_unpack_jobid(orte_buffer_t *buffer, void *dest,
|
|
orte_std_cntr_t *num_vals, orte_data_type_t type);
|
|
|
|
ORTE_DECLSPEC int orte_ns_base_unpack_vpid(orte_buffer_t *buffer, void *dest,
|
|
orte_std_cntr_t *num_vals, orte_data_type_t type);
|
|
|
|
/*
|
|
* copy functions
|
|
*/
|
|
|
|
int orte_ns_base_copy_name(orte_process_name_t **dest, orte_process_name_t *src, orte_data_type_t type);
|
|
|
|
int orte_ns_base_copy_vpid(orte_vpid_t **dest, orte_vpid_t *src, orte_data_type_t type);
|
|
|
|
int orte_ns_base_copy_nodeid(orte_nodeid_t **dest, orte_nodeid_t *src, orte_data_type_t type);
|
|
|
|
int orte_ns_base_copy_jobid(orte_jobid_t **dest, orte_jobid_t *src, orte_data_type_t type);
|
|
|
|
/*
|
|
* compare functions
|
|
*/
|
|
|
|
int orte_ns_base_compare_name(orte_process_name_t *value1,
|
|
orte_process_name_t *value2,
|
|
orte_data_type_t type);
|
|
|
|
|
|
int orte_ns_base_compare_vpid(orte_vpid_t *value1,
|
|
orte_vpid_t *value2,
|
|
orte_data_type_t type);
|
|
|
|
int orte_ns_base_compare_jobid(orte_jobid_t *value1,
|
|
orte_jobid_t *value2,
|
|
orte_data_type_t type);
|
|
|
|
int orte_ns_base_compare_nodeid(orte_nodeid_t *value1,
|
|
orte_nodeid_t *value2,
|
|
orte_data_type_t type);
|
|
|
|
/*
|
|
* size functions
|
|
*/
|
|
|
|
int orte_ns_base_std_size(size_t *size, void *src, orte_data_type_t type);
|
|
|
|
/*
|
|
* release functions
|
|
*/
|
|
|
|
void orte_ns_base_std_release(orte_data_value_t *value);
|
|
|
|
/*
|
|
* print functions
|
|
*/
|
|
|
|
int orte_ns_base_std_print(char **output, char *prefix, void *src, orte_data_type_t type);
|
|
|
|
int orte_ns_base_print_name(char **output, char *prefix, orte_process_name_t *name, orte_data_type_t type);
|
|
|
|
|
|
/*
|
|
* external API functions will be documented in the mca/ns/ns.h file
|
|
*/
|
|
|
|
#if defined(c_plusplus) || defined(__cplusplus)
|
|
}
|
|
#endif
|
|
#endif
|