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 The modex stage gate (STG1) has been replaced by a collective operation in the modex itself that performs an allgather on the required modex info.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The other stage gates have been replaced by orte_grpcomm barrier functions.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.
that it is >= 1, so making it a size_t makes it easier to interact
with all the other size_t variables and removes a compiler warning.
This commit was SVN r15935.
used at nce (up to one unique collective module per collective function).
Matches r15795:15921 of the tmp/bwb-coll-select branch
This commit was SVN r15924.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r15795
r15921
mpi_preconnect_oob_simultaneous > np. Need to scale back
simultaneous to equal np in those cases. Reviewed by Brian.
This commit fixes trac:1064.
This commit was SVN r15916.
The following Trac tickets were found above:
Ticket 1064 --> https://svn.open-mpi.org/trac/ompi/ticket/1064
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.
mpi_show_mpi_alloc_mem_leaks
When activated, MPI_FINALIZE displays a list of memory allocations
from MPI_ALLOC_MEM that were not freed by MPI_FREE_MEM (in each MPI
process).
* If set to a positive integer, display only that many leaks.
* If set to a negative integer, display all leaks.
* If set to 0, do not show any leaks.
This commit was SVN r15736.
* 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.
* bml.h had a change that introduced a variable named "_order" to
avoid a conflict with a local variable. The namespace starting
with _ belongs to the os/compiler/kernel/not us. So we can't start
symbols with _. So I replaced it with arg_order, and also updated
the threaded equivalent of the macro that was modified.
* in btl_openib_proc.c, one opal_output accidentally had its string
reverted from "ompi_modex_recv..." to
"mca_pml_base_modex_recv....". This was fixed.
* The change to ompi/runtime/ompi_preconnect.c was entirely
reverted; it was an artifact of debugging.
This commit was SVN r15475.
The following SVN revision numbers were found above:
r15474 --> open-mpi/ompi@8ace07efed
1. Galen's fine-grain control of queue pair resources in the openib
BTL.
1. Pasha's new implementation of asychronous HCA event handling.
Pasha's new implementation doesn't take much explanation, but the new
"multifrag" stuff does.
Note that "svn merge" was not used to bring this new code from the
/tmp/ib_multifrag branch -- something Bad happened in the periodic
trunk pulls on that branch making an actual merge back to the trunk
effectively impossible (i.e., lots and lots of arbitrary conflicts and
artifical changes). :-(
== Fine-grain control of queue pair resources ==
Galen's fine-grain control of queue pair resources to the OpenIB BTL
(thanks to Gleb for fixing broken code and providing additional
functionality, Pasha for finding broken code, and Jeff for doing all
the svn work and regression testing).
Prior to this commit, the OpenIB BTL created two queue pairs: one for
eager size fragments and one for max send size fragments. When the
use of the shared receive queue (SRQ) was specified (via "-mca
btl_openib_use_srq 1"), these QPs would use a shared receive queue for
receive buffers instead of the default per-peer (PP) receive queues
and buffers. One consequence of this design is that receive buffer
utilization (the size of the data received as a percentage of the
receive buffer used for the data) was quite poor for a number of
applications.
The new design allows multiple QPs to be specified at runtime. Each
QP can be setup to use PP or SRQ receive buffers as well as giving
fine-grained control over receive buffer size, number of receive
buffers to post, when to replenish the receive queue (low water mark)
and for SRQ QPs, the number of outstanding sends can also be
specified. The following is an example of the syntax to describe QPs
to the OpenIB BTL using the new MCA parameter btl_openib_receive_queues:
{{{
-mca btl_openib_receive_queues \
"P,128,16,4;S,1024,256,128,32;S,4096,256,128,32;S,65536,256,128,32"
}}}
Each QP description is delimited by ";" (semicolon) with individual
fields of the QP description delimited by "," (comma). The above
example therefore describes 4 QPs.
The first QP is:
P,128,16,4
Meaning: per-peer receive buffer QPs are indicated by a starting field
of "P"; the first QP (shown above) is therefore a per-peer based QP.
The second field indicates the size of the receive buffer in bytes
(128 bytes). The third field indicates the number of receive buffers
to allocate to the QP (16). The fourth field indicates the low
watermark for receive buffers at which time the BTL will repost
receive buffers to the QP (4).
The second QP is:
S,1024,256,128,32
Shared receive queue based QPs are indicated by a starting field of
"S"; the second QP (shown above) is therefore a shared receive queue
based QP. The second, third and fourth fields are the same as in the
per-peer based QP. The fifth field is the number of outstanding sends
that are allowed at a given time on the QP (32). This provides a
"good enough" mechanism of flow control for some regular communication
patterns.
QPs MUST be specified in ascending receive buffer size order. This
requirement may be removed prior to 1.3 release.
This commit was SVN r15474.
Short description: major changes include -
1. singletons now fork/exec a local daemon to manage their operations.
2. the orte daemon code now resides in libopen-rte
3. daemons no longer use the orte triggering system during startup. Instead, they directly call back to their parent pls component to report ready to operate. A base function to count the callbacks has been provided.
I have modified all the pls components except xcpu and poe (don't understand either well enough to do it). Full functionality has been verified for rsh, SLURM, and TM systems. Compile has been verified for xgrid and gridengine.
This commit was SVN r15390.
than just the PML/BTLs these days. Also clean up the code so that it
handles the situation where not all nodes register information for a given
node (rather than just spinning until that node sends information, like
we do today).
Includes r15234 and r15265 from the /tmp/bwb-modex branch.
This commit was SVN r15310.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r15234
r15265
Changes paffinity interface to use a cpu mask for available/preferred cpus
rather than the current coarse grained paffinity that lets the OS choose
which processor.
Macros for setting and clearing masks are provided.
Solaris and windows changes have not been made. Solaris subdirectory has some
suggested changes - however the relevant man pages for the Solaris 10 APIs
have some ambiguity regarding order in which one create and sets a processor
set. As we did not have access to a solaris 10 machine we could not test to
see the correct way to do the work under solaris.
This commit was SVN r14887.
This commit moves the initalization/finalization of opal_event and opal_progress
to opal_init/finalize. These were previously init/final in ORTE which is an
abstraction violation. After talking about it we concluded that there are no
ordering issues that require these to be init/final in ORTE instead of OPAL.
I ran the IBM test suite against this commit and it didn't turn up any new
failures so I think it is good to go.
Let us know if this causes problems.
This commit was SVN r14773.
wireup. For small clusters or clusters with decent ARP lookup and
connect times, this will have marginal impact. For systems with either
bad ARP lookup times or long connect times, increasing this number
to something much closer to SOMAXCONN (128 on most modern machines) will
result in a faster OOB wireup. Don't set higher than SOMAXCONN or you
can end up with lots of connect() retries and we'll end up slower.
This commit was SVN r14742.
The primary change that underlies all this is in the OOB. Specifically, the problem in the code until now has been that the OOB attempts to resolve an address when we call the "send" to an unknown recipient. The OOB would then wait forever if that recipient never actually started (and hence, never reported back its OOB contact info). In the case of an orted that failed to start, we would correctly detect that the orted hadn't started, but then we would attempt to order all orteds (including the one that failed to start) to die. This would cause the OOB to "hang" the system.
Unfortunately, revising how the OOB resolves addresses introduced a number of additional problems. Specifically, and most troublesome, was the fact that comm_spawn involved the immediate transmission of the rendezvous point from parent-to-child after the child was spawned. The current code used the OOB address resolution as a "barrier" - basically, the parent would attempt to send the info to the child, and then "hold" there until the child's contact info had arrived (meaning the child had started) and the send could be completed.
Note that this also caused comm_spawn to "hang" the entire system if the child never started... The app-failed-to-start helped improve that behavior - this code provides additional relief.
With this change, the OOB will return an ADDRESSEE_UNKNOWN error if you attempt to send to a recipient whose contact info isn't already in the OOB's hash tables. To resolve comm_spawn issues, we also now force the cross-sharing of connection info between parent and child jobs during spawn.
Finally, to aid in setting triggers to the right values, we introduce the "arith" API for the GPR. This function allows you to atomically change the value in a registry location (either divide, multiply, add, or subtract) by the provided operand. It is equivalent to first fetching the value using a "get", then modifying it, and then putting the result back into the registry via a "put".
This commit was SVN r14711.
* Don't need the 2 process case -- we'll send an extra message, but
at very little cost and less code is better.
* Use COMPLETE sends instead of STANDARD sends so that the connection
is fully established before we move on to the next connection. The
previous code was still causing minor connection flooding for huge
numbers of processes.
* mpi_preconnect_all now connects both OOB and MPI layers. There's
also mpi_preconnect_mpi and mpi_preconnect_oob should you want to
be more specific.
* Since we're only using the MCA parameters once at the beginning
of time, no need for global constants. Just do the quick param
lookup right before the parameter is needed. Save some of that
global variable space for the next guy.
Fixes trac:963
This commit was SVN r14553.
The following Trac tickets were found above:
Ticket 963 --> https://svn.open-mpi.org/trac/ompi/ticket/963
There is a binomial algorithm in the code (i.e., the HNP would send to a subset of the orteds, which then relay it on according to the typical log-2 algo), but that has a bug in it so the code won't let you select it even if you tried (and the mca param doesn't show, so you'd *really* have to try).
This also involved a slight change to the oob.xcast API, so propagated that as required.
Note: this has *only* been tested on rsh, SLURM, and Bproc environments (now that it has been transferred to the OMPI trunk, I'll need to re-test it [only done rsh so far]). It should work fine on any environment that uses the ORTE daemons - anywhere else, you are on your own... :-)
Also, correct a mistake where the orte_debug_flag was declared an int, but the mca param was set as a bool. Move the storage for that flag to the orte/runtime/params.c and orte/runtime/params.h files appropriately.
This commit was SVN r14475.
Fix for memory corruption in the restarted process stack. This stemed from
the brute force method we were previously using. This commit fixes this by
using a lighter weight solution focused in the r2 BML instead of above the PML.
This is a more efficient and flexible solution, and it solves the original
problem.
In the process I pulled out the ft_event function in the tcp BTL and r2 BML
into a set of *_ft.[c|h] files just to keep any updates to these code paths
as isolated as possible to make merging easier on everyone.
This commit was SVN r14371.
The following SVN revision numbers were found above:
r2 --> open-mpi/ompi@58fdc18855
The following Trac tickets were found above:
Ticket 977 --> https://svn.open-mpi.org/trac/ompi/ticket/977
This merge adds Checkpoint/Restart support to Open MPI. The initial
frameworks and components support a LAM/MPI-like implementation.
This commit follows the risk assessment presented to the Open MPI core
development group on Feb. 22, 2007.
This commit closes trac:158
More details to follow.
This commit was SVN r14051.
The following SVN revisions from the original message are invalid or
inconsistent and therefore were not cross-referenced:
r13912
The following Trac tickets were found above:
Ticket 158 --> https://svn.open-mpi.org/trac/ompi/ticket/158
support SEND_IN_PLACE causes badness because the BTL tries to use the
not-exactly-complete convertor. Don't need it in this situation anyway.
This commit was SVN r13700.
