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 The modex stage gate (STG1) has been replaced by a collective operation in the modex itself that performs an allgather on the required modex info.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The other stage gates have been replaced by orte_grpcomm barrier functions.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.
in the OMPI proc structures. For now, use an extension of the modex that is
keyed on strings. Eventually, this should use the attribute put/get that is
part of the RSL interface.
This commit was SVN r15820.
* General TCP cleanup for OPAL / ORTE
* Simplifying the OOB by moving much of the logic into the RML
* Allowing the OOB RML component to do routing of messages
* Adding a component framework for handling routing tables
* Moving the xcast functionality from the OOB base to its own framework
Includes merge from tmp/bwb-oob-rml-merge revisions:
r15506, r15507, r15508, r15510, r15511, r15512, r15513
This commit was SVN r15528.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r15506
r15507
r15508
r15510
r15511
r15512
r15513
Cleanup ALL instances of output involving the printing of orte_process_name_t structures using the ORTE_NAME_ARGS macro so that the number of fields and type of data match. Replace those values with a new macro/function pair ORTE_NAME_PRINT that outputs a string (using the new thread safe data capability) so that any future changes to the printing of those structures can be accomplished with a change to a single point.
Note that I could not possibly find outputs that directly print the orte_process_name_t fields, but only dealt with those that used ORTE_NAME_ARGS. Hence, you may still have a few outputs that bark during compilation. Also, I could only verify those that fall within environments I can compile on, so other environments may yield some minor warnings.
This commit was SVN r15517.
Over to Jeff now for modifying mpi_init accordingly.
Until Jeff makes his changes, nobody should see anything different as the new info just isn't used by anything!
This commit was SVN r13351.
and George on these refinements):
* Rename the static OBJ initializer macro to be
OPAL_OBJ_STATIC_INIT(class)
* Ensure that all static OBJ initializations get a refcount of 1
(doesn't ''really'' matter, since they're static, it should never
get to the point where the OBJ is DESTRUCTed, but more correct
nonetheless)
* Add a "magic number" to the OBJ when compiling with debug support.
The magic number does some rudimentary support to ensure that
you're operating on a valid OBJ (and fails an assertion if you're
not). Check to ensure that the memory contains the magic number
when performing actions of OBJ's. Also remove the magic number
when DESTRUCTing OBJs, so that if, for example, an OBJ is
DESTRUCTed more than once, we'll fail the magic number assert.
This commit was SVN r13338.
The following SVN revision numbers were found above:
r13227 --> open-mpi/ompi@96030de97b
r13228 --> open-mpi/ompi@c2e9075d29
* Add line about heterogeneous support to ompi_info output
* Print warning and abort if heterogeneous detected and
no heterogeneous support available.
Refs trac:587
This commit was SVN r12943.
The following Trac tickets were found above:
Ticket 587 --> https://svn.open-mpi.org/trac/ompi/ticket/587
I found only two places that were looking at the tokens:
1. the odls - we used the tokens to separately process the globals container data from everything else. In this case, I left the subscription that returned the globals data alone, but "stripped" the subscription that returned the launch data for the procs. These subscriptions have nothing to do with the xcast message.
2. the pml_base_modex - the callback function was getting process names from the returned tokens. Actually, this function was doing a very bad thing - it was assuming that the first token returned was *always* the process name. This is currently true, but is one of those assumptions that someone could have easily changed - and suddenly found the system inexplicably failing. I modified the function to (a) get the name sent back to us, (b) "stripped" the value structures of tokens and segment strings, and (c) correctly obtained process names from the returned values. I also reindented the heck out of the code so it was legible (at least, to my old eyes).
This commit was SVN r12813.
* Do not add new procs to the global list during modex callback or
when sharing orte names during accept/connect. For modex, we
cache the modex info for later, in case that proc ever does get
added to the global proc list. For accept/connect orte name
exchange between the roots, we only need the orte name, so no
need to add a proc structure anyway. The procs will be added
to the global process list during the proc exchange later in
the wireup process
* Rename proc_get_namebuf and proc_get_proclist to proc_pack
and proc_unpack and extend them to include all information
needed to build that proc struct on a remote node (which
includes ORTE name, architecture, and hostname). Change
unpack to call pml_add_procs for the entire list of new
procs at once, rather than one at a time.
* Remove ompi_proc_find_and_add from the public proc
interface and make it a private function. This function
would add a half-created proc to the global proc list, so
making it harder to call is a good thing.
This means that there's only two ways to add new procs into the global proc list at this time: During MPI_INIT via the call to ompi_proc_init, where my job is added to the list and via ompi_proc_unpack using a buffer from a packed proc list sent to us by someone else. Currently, this is enough to implement MPI semantics. We can extend the interface more if we like, but that may require HNP communication to get the remote proc information and I wanted to avoid that if at all possible.
Refs trac:564
This commit was SVN r12798.
The following Trac tickets were found above:
Ticket 564 --> https://svn.open-mpi.org/trac/ompi/ticket/564
Accordingly, there are new APIs to the name service to support the ability to get a job's parent, root, immediate children, and all its descendants. In addition, the terminate_job, terminate_orted, and signal_job APIs for the PLS have been modified to accept attributes that define the extent of their actions. For example, doing a "terminate_job" with an attribute of ORTE_NS_INCLUDE_DESCENDANTS will terminate the given jobid AND all jobs that descended from it.
I have tested this capability on a MacBook under rsh, Odin under SLURM, and LANL's Flash (bproc). It worked successfully on non-MPI jobs (both simple and including a spawn), and MPI jobs (again, both simple and with a spawn).
This commit was SVN r12597.
Clean up the remainder of the size_t references in the runtime itself. Convert to orte_std_cntr_t wherever it makes sense (only avoid those places where the actual memory size is referenced).
Remove the obsolete oob barrier function (we actually obsoleted it a long time ago - just never bothered to clean it up).
I have done my best to go through all the components and catch everything, even if I couldn't test compile them since I wasn't on that type of system. Still, I cannot guarantee that problems won't show up when you test this on specific systems. Usually, these will just show as "warning: comparison between signed and unsigned" notes which are easily fixed (just change a size_t to orte_std_cntr_t).
In some places, people didn't use size_t, but instead used some other variant (e.g., I found several places with uint32_t). I tried to catch all of them, but...
Once we get all the instances caught and fixed, this should once and for all resolve many of the heterogeneity problems.
This commit was SVN r11204.
interconnects that provide matching logic in the library.
Currently includes support for MX and some support for
Portals
* Fix overuse of proc_pml pointer on the ompi_proc structuer,
splitting into proc_pml for pml data and proc_bml for
the BML endpoint data
* bug fixes in bsend init code, which wasn't being used by
the OB1 or DR PMLs...
This commit was SVN r10642.
(which is currently the default, although we may argue over this later
:-) ), a new field in the ompi_proc_t named proc_hostname will have
the string hostname of that peer. If 0, this field will be NULL.
This allows for printing nicer error messages in environments where
peer hostnames are not otherwise easily obtainable, such as the mvapi
BTL (requested by Sandia, who has both a *huge* number of nodes and
6GB of RAM per node, so they don't care about the extra memory usage
;-) ).
This commit was SVN r9902.
procs architecture ints to the architecture of the last proc we got
in the list from the gpr, which is probably not what we intended to
do...
This commit was SVN r9144.
- move files out of toplevel include/ and etc/, moving it into the
sub-projects
- rather than including config headers with <project>/include,
have them as <project>
- require all headers to be included with a project prefix, with
the exception of the config headers ({opal,orte,ompi}_config.h
mpi.h, and mpif.h)
This commit was SVN r8985.
originally suggested by Ralf Wildenhues, to try to speed autogen, configure,
and make (and possibly even make install). Use automake's include directive
to drastically reduce the number of Makefile files (although the number of
Makefile.am files is the same - most are just included in a top-level
Makefile.am). Also use an Automake SUBDIRs feature to eliminate the
dynamic-mca tree, which was no longer really needed. This makes adding
a framework easier (since you don't have to remember the dynamic-mca
tree) and makes building faster (as make doesn't have to recurse through
the dynamic-mca tree)
This commit was SVN r7777.
AM_INIT_AUTOMAKE, instead of the deprecated version.
* Work around dumbness in modern AC_INIT that requires the version
number to be set at autoconf time (instead of at configure time, as
it was before). Set the version number, minus the subversion r number,
at autoconf time. Override the internal variables to include the r
number (if needed) at configure time. Basically, the right thing
should always happen. The only place it might not is the version
reported as part of configure --help will not have an r number.
* Since AM_INIT_AUTOMAKE taks a list of options, no need to specify
them in all the Makefile.am files.
* Addes support for subdir-objects, meaning that object files are put
in the directory containing source files, even if the Makefile.am is
in another directory. This should start making it feasible to
reduce the number of Makefile.am files we have in the tree, which
will greatly reduce the time to run autogen and configure.
This commit was SVN r7211.
1. Added OMPI_PROC_ARCH as a defined registry key and added the code so that the architecture info gets properly transmitted across all processes using the startup message.
2. Added an OMPI_MODEX_KEY definition and removed the hard-coded "modex" key from pml_modex_exchange
This commit was SVN r7129.
Here's the huge registry check-in you've all been waiting for with baited breath. The revised version sends a single message to all processes at the various stage gates, thus making the startup much more scalable. I could provide you with all the tawdry details, but won't for now - you are welcome to ask, though, and I'll merrily bore your ears to tears.
In addition, the commit contains the following:
1. set the ignore properties on ompi/debuggers and orte/mca/pls/poe
2. Added simplified subscribe and put functions to the registry's API. I have also converted all of the ompi functions that registered subscriptions to the new API, and caught their associated put's as well.
In a follow-on commit, I'll be adding support for George's hetero arch registry subscription (wanted to get this one in first).
This commit was SVN r7118.
tree.
- fix up #include's throughout the tree (yay contrib/search_replace.pl!)
- remove a few extraneous #include's
- remove orte_sys_info*() from opal_init()/opal_finalize() (it's
already in orte_init_stage1() and orte_system_finalize())
- remove dependencies in opal on orte_system_info -- util/os_path.c
and util/os_create_dirpath.c (they only used path_sep, anyway --
easily changed to #defines)
This commit was SVN r7059.
Change all the places where they are used to fit the new name.
Remove the code to check the remote arch from the PML. We will have a GPR mechanism
in ompi_mpi_initialize to do that.
This commit was SVN r6750.
1. Modify the registry to eliminate redundant data copying for startup messages.
2. Revise the subscription/trigger system to avoid redundant storage of triggers and subscriptions. This dramatically reduces the search time when a registry action occurs - to illustrate the point, there are now only a handful of triggers on the system for each job. Before, there were a handful of triggers for each PROCESS in the job, all of which had to be checked every time something happened on the registry. This is much, much faster now.
3. Update all subscriptions to the new format. There are now "named" subscriptions - this allows you to "name" a subscription that all the processes will be using. The first one to hit the registry actually defines the subscription. From then on, any subsequent "subscribes" to the same name just cause that process to "attach" to the existing subscription. This keeps the number of subscriptions being tracked by the registry to a minimum, while ensuring that each process still gets notified.
4. Do the same for triggers.
Also fixed a duplicate subscription problem that was causing people to receive data equal to the number of processes times the data they should have received from a trigger/subscription. Sorry about that... :-( ...but it's all better now!
Uncovered a situation where the modex data seems to be getting entered on the registry a second time - the latter time coming after the compound command has been "fired", thereby causing all the subscriptions to fire. Asked Tim and Jeff to look into this.
Second phase of the changes will involve modifying the xcast system so that the same message gets sent to all processes. This will further reduce the message traffic, and - once we have a true "broadcast" version of xcast - really speed things up and improve scalability.
This commit was SVN r6542.
- Change ompi_proc_world() to only return the procs in this job (as
opposed to all of them)
- Add a subscription that fires during MPI_INIT (stg1) for figuring
out which procs are on my local node. Need to figure out what to do
in the esoteric cases -- but the obvious one (Red Storm), where
subscriptions are never fired, is ok, because by definition, no
other procs will be on my node, so their default value (not on my
node) is ok.
--> Need to have RHC check this code; it seems to work, but I think
I'm getting too much data back from the subscription.
- End result is that any proc that is on my node will have its
OMPI_PROC_FLAG_LOCAL bit set on its proc->proc_flags field.
- Added/corrected a few comments in proc.h.
This commit was SVN r6507.
