Running on Edison

ULFM configuration for NERSC Edison

Edison is a Cray XC30, with a peak performance of 2.57 petaflops/sec, 133,824 compute cores, 357 terabytes of memory, and 7.56 petabytes of disk. Hosted at NERSC, it offers access to many researchers and practitioners in the US.

Installing the latest versions of Open MPI (including the development, unstable and stable) on this machine can easily be done by using the provided platform file. The same cannot unfortunately be said about the current ULFM version. This version was based on an older unstable version of Open MPI (1.7) and didnt backported all the new exciting features from the development branch of Open MPI. As a result, getting it to work on Edison is a little challenging, but we all like a little challenge.

Preparing for installation

This step is unfortunately required, as I couldnt find any other way to update the m4 file of our ancient ULFM installation, without dedicating too much time to such a hopeless task (especially as the new ULFM 2.0 is reaching almost stable status). The issue is triggered by a misconfiguration of autoconf 2.69 on Edison, which breaks all the autotool chain. The issue can be easily corrected by downloading the Mercurial version of ULFM on another computer, running, a quick configure with no arguments, followed by “make dist”. This will generate 2 archives (openmpi-1.7.1ULFM.tar.bz2 and openmpi-1.7.1ULFM.tar.gz). Use the one corresponding to your preferred compression algorithm, and move it on Edison.

Compiling from source

With the archive moved on Edison, you can now undergo the last step to configure and compile ULFM there. Lets assume we want to build an optimized version, integrated with Edison resource management software (Slurm), using uGNI, without debugging information. One of the interesting features ULFM inherited from Open MPI is the capability of using platform files. Here is the platform file for Edison, it will eventually make its way into the official ULFM release, meanwhile you can copy and paste directly from here.


With this platform file saved as platform.edisonyou can now proceed to configure ULFM using “./configure —with-platform=platform.edison …”. There are many other possible configuration parameters, and you are strongly encouraged to read the output of configure —helpto gain more insight into the flexible configuration system. Meanwhile here is what I usually add in addition to the platform file “—prefix=… –enable-mpirun-prefix-by-default” (please replace the dots with your prefered installation directory). Keep in mind that the installation directory must be in your $PATH, to have direct access to the wrapper compilers and to the mpiexec.

Preparing for execution

In general we suggest you to use a specific MCA file (the -am parameter) to provide the parameters for your ULFM runs (and not collide with the normal Open MPI runs). As long as you use mpiexec to start your application this will work. If instead you want to rely on srun to so do, you will have to dump the context of your AMCA file directly into your main ${HOME}/.openmpi/mca-params.conf file (and pay attention when you execute normal Open MPI runs).

Running ULFM applications in an interactive job

Congratulations, the hardest steps are now done and you are now supposed to have a fully compiled and ready to run version of ULFM, and its binaries in your $PATH. Allocating an interactive job on Edison is well described in the online documentation. There is however a trick, you need to add an option to prevent Slurm from destroying your job when one of the processes disappear. This option is —no-kill.

Thus, allocating an interactive job with 4 processes for a short duration can be achieved with “`salloc -p debug -N 4 –no-kill“.
Running an application then become easy, as indicated in ULFM setup steps.

Survival readiness

You code has now a sane execution environment that is able to survive faults, and can help your application achieve the same goal.

SC’ 16 paper: Failure Detection and Propagation in HPC System

We have presented our paper on the scalable failure detection, “Failure Detection and Propagation in HPC System” at SC’16.

Building an infrastructure for exascale applications requires, in addition to many other key components, a stable and efficient failure detector. This paper describes the design and evaluation of a robust failure detector, able to maintain and distribute the correct list of alive resources within proven and scalable bounds.
The detection and distribution of the fault information follow different overlay topologies that together guarantee minimal disturbance to the applications. A virtual observation ring minimizes the overhead by allowing each node to be observed by another single node, providing an unobtrusive behavior.
The propagation stage is using a non-uniform variant of a reliable broadcast over a circulant graph overlay network, and guarantees a logarithmic fault propagation. Extensive simulations, together with experiments on the ORNL Titan supercomputer, show that the algorithm performs extremely well and exhibits all the desired properties of an exascale-ready algorithm.

The slides presented by Yves are here and the paper is also available.

Download (PDF, 638KB)

This failure detector is implemented in the ULFM 2.0, an effort that will soon become open to public.

SC’16 Tutorial

The ULFM team is happy to announce that we will be teaching a day-long tutorial on fault tolerance at SC’16 (somewhat similar to last year tutorial). The tutorial will cover multiple theoretical and practical aspects of dealing with faults. It targets a wide scientific community, starting from scientists trying to understand the challenges of different types of failures, and up to advanced users with prior experience with fault-related topics that want to get a more precise understanding of the available tools allowing them to efficiently deal with faults.

The tutorial was divided in two parts, one theoretical (covering the different existing approaches and their modeling), and one practical. The slides for the 2 parts are available (theory and practice), as well as the handon examples. Unlike the previous years, we have embraced new technologies to facilitate the public interaction with ULFM: enjoy the ULFM docker.

More information about the tutorial can be found here. Enjoy our promotional video 😉

See you all in Salt Lake City, UT !!!

ULFM-1.1 Release

ULFM has reached the 1.1 milestone, a minor release, crushing few bugs identified by our users and developers.

Focus has been toward improving stability, feature coverage for intercommunicators, and following the updated specification for MPI_ERR_PROC_FAILED_PENDING.

  • Addition of the MPI_ERR_PROC_FAILED_PENDING error code, as per newer specification revision. Properly returned from point-to-point, non-blocking ANY_SOURCE operations.
  • Alias MPI_ERR_PROC_FAILED, MPI_ERR_PROC_FAILED_PENDING and MPI_ERR_REVOKED to the corresponding standard blessed – extension- names MPIX_ERR_xxx.
  • Support for Intercommunicators:
    • Support for the blocking version of the agreement, MPI_COMM_AGREE on Intercommunicators.
    • MPI_COMM_REVOKE tested on intercommunicators.
  • Disabled completely (.ompi_ignore) many untested components
  • Changed the default ORTE failure notification propagation aggregation delay from 1s to 25ms.
  • Added an OMPI internal failure propagator; failure propagation between SM domains is now immediate.
  • Bugfixes:
    • SendRecv would not always report MPI_ERR_PROC_FAILED correctly.
    • SendRecv could incorrectly update the status with errors pertaining to the Send portion of the Sendrecv.
  • Revoked send operations are now always completed or remote cancelled and may not deadlock anymore.
  • Cancelled send operations to a dead peer will not trigger an assert when the BTL reports that same failure.
  • Repeat calls to operations returning MPI_ERR_PROC_FAILED will eventually return MPI_ERR_REVOKED when another process revokes the communicator.

Get the source and happy hacking,
The ULFM team

SC’15 tutorial

The ULFM team is happy to announce that we will be teaching a day-long tutorial on fault tolerance at SC’15 (somewhat similar to last year tutorial). The tutorial will cover multiple theoretical and practical aspects of dealing with faults. It targets a wide scientific community, starting from scientists trying to understand the challenges of different types of failures, and up to advanced users with prior experience with fault-related topics that want to get a more precise understanding of the available tools allowing them to efficiently deal with faults.

Get the slides part1, part2, and the examples

More information about the tutorial can be found here. Enjoy our promotional video 😉

See you all in Austin, TX !!!


Reliability is one of the major concerns when envisioning the future Exascale platforms. The IESP projects an increase in node performance and node concurrency by one or two orders of magnitude, which translates, even under the most optimistic perspectives, in a mechanical decrease of the mean time to interruption (MTTI) of at least one order of magnitude. Because of this tendency, platform providers, software implementors, and high-performance application users who target capability runs on such machines cannot regard the occurrence of interruption due to a failure as a rare dramatic event, but must consider faults inevitable and take them into account by integrating some form of fault-tolerance.

One easy way to get ready is to join us at SC’14 in New Orleans for a tutorial on fault tolerance, a middle-ground between theoretical understanding and practical knowledge. This tutorial will present a comprehensive survey of the techniques proposed to deal with failures in high performance systems. The main goal is to provide the attendees with a clear picture of this important topic: what are the techniques, how do they work, and how can they be evaluated? The tutorial is organized in four parts: (i) An overview of failure types (software/hardware, transient/fail-stop), and typical probability distributions (Exponential, Weibull, Log-Normal); (ii) General-purpose techniques, which include several checkpoint and rollback recovery protocols, replication, prediction and silent error detection; (iii) Application-specific techniques, such as ABFT for grid-based algorithm or fixed-point convergence for iterative applications; and (iv) Practical deployment of fault tolerant techniques with User Level Fault Mitigation (a fault tolerant MPI extension recently proposed to the MPI forum). Relevant examples based on widespread computational solver routines will be protected with a mix of checkpoint-restart and ABFT techniques in a hands-on session.

In preparation for the hand-on session one needs to get ready by installing ULFM and setting up the paths to access it. You can either follow the post or the steps below.
1. Download the version prepared for this tutorial.
2. Untar it in some convenient location.

3. Go inside the newly untarred directory and launch

4. Configure Open MPI. You should change the –prefix in the following command, and make sure that –enable-mpi-ext=ftmpi –with-ft=mpi is specified.

5. Compile, link and install

6. Add the directory provided as a –prefix in Step 4 to your PATH and LD_LIBRARY_PATH. As an example for bash one can add the following two lines to ${HOME}/.bashrc

7. You’re almost ready for the hand-on session.
8. The archive with the examples and skeletons is available here.

Let’s rock the faults!

The slides used during this tutorial are available here, here and here.

Uniform Intercomm Creation

A question about uniformly creating an inter-communicator using MPI_Intercomm_create has been posted on the ULFM mailing list. Initially, I though it is an easy corner-case, that can be solved with few barriers and/or agreements. It turns out this issue is more complicated that initially expected, with few twists on the way. Let me detail our adventure toward writing a uniform intercomm creation function.

Before moving further, let’s clarify what MPI_Intercomm_create is about. The MPI standard is not very explicit about the scope of this function, but we can gather enough info to start talking about (page 262 line 6):

This call [MPI_Intercomm_create] creates an inter-communicator. It is collective over the union of the local and remote groups. Processes should provide identical local_comm and local_leader arguments within each group. Wildcards are not permitted for remote_leader, local_leader, and tag.

In other words, if you provide two intra-communicators, a leader on each one and a bridge communicator where the leaders can talk together, you will be able to bind the two groups of processes corresponding to each of the intra-communicators into a inter-communicator. Neat! Graphically speaking this should look like

So far so good, but what “uniformly” means? Based on some web resources uniformly is about consistency, in this particular instance about the fact that all participants will return the same answer (aka. the some inter-communicator or the same error code) despite all odds. Here we are looking specifically from the point of view of process failures, more precisely a single failure. Obviously, the trivial approach where one agrees on the resulting intercomm doesn’t work, as due to process failures some processes might not have an intercomm. Clearly, we need a different, better approach.

While talking about this, Aurelien proposed the following solution, entirely based on MPI_Barrier.

Looks quite simple and straightforward. Does it do reach the expected consensus at the end? Unfortunately not, it doesn’t work in all cases. Imagine that all peers in one side succeeded to create the intercomm while everyone in the remote communicator failed. Then on one side the interexists is TRUE while on the other side it is FALSE. Thus the local barrier (line 3) succeed on the side with interexists equal to TRUE and the intercomm is not revoked. As the intercomm exists, these processes will then go in the second MPI_Barrier (line 10), on the “half-created” intercomm, and will deadlock as there will never be a remote process to participate to the barrier.

What we are missing here is an agreement between the two sides (A and B) that things went well for everyone. So let’s try to use agreement to reach this goal. In the remaining of this post I will use the function AGREE, which is a point-to-point agreement between two processes (the two leaders). We need this particular function, as we cannot call MPI_Comm_agree on the bridge communicator, simply because the two leaders are not the only participants in the bridge communicator. Fortunately, implementing an agreement between two processes is almost a trivial task, so I will consider the AGREE function as existing.

If there is no failure the code works, but it’s an overkill. Now, if we have one failure during the execution there are several cases.
1. The failure happens before the MPI_Intercomm_create. Some processes will have an intercomm, and some will not. Thus the local agreement will make everyone on the same group (A and B) agree about the local existence of the intercomm. The AGREE between the leaders will then propagate this knowledge to the remote group leader. At this point the two leaders have the correct answer, and then the MPI_Comm_agree at line 8 will finally distribute this answer on the two groups. This case works!
2. The failure happen after the MPI_Intercomm_create. One can think about this case as a false positive (because the intercomm exists on all alive processes), but the real question is if we are able to detect it and reach a consensus. The MPI_Comm_agree at line 3 will make the flag TRUE on both groups. Note that on one side the MPI_Comm_agree will return MPI_ERR_PROC_FAILED as required by the MPI standard, but we do not take in account the return code here.

Here is the reasoning leading to this form of the algorithm.
1. The MPI_Comm_agree on line 3 is a little tricky. We use the returned value of the flag (bitwise AND operation between all living processes), but we willingly choose to ignore the return value. What really matters is that all alive participants have the resulting intercomm, and this will be reflected in the value of flag after the call. All other cases are trapped either by partially existing intercomm, or by the REVOKE call. Moreover, it is extremely important that the resulting value of flag is computed through an agreement and not through a traditional collective communication (which lack the consistency factor).
2. For a similar reason the MPI_Comm_agree on line 9 cannot be replaced by a MPI_Broadcast, even as its meaning is to broadcast the information collected at the local root into the local group of processes. Without the use of MPI_Comm_agree here one cannot distinguish between every process still alive has an intercomm with known faults or only some participants have an intercomm (with faults). Thus, as the intercomm can be revoked only if all alive processes know about it, we have to enforce this consistent view by the use of the agreement.

While this inter-communicator creation looks like an extremely complicated and expensive matter, one should keep in mind that such communicator creation are not supposed to be in the critical path, so they might not be as prohibitive as one might think. If Everything Was Easy Nothing Would Be Worth It!