The Multicast/Reduction Network: A User's Guide to MRNet v2.0

MRNet has been developed to be highly portable; we expect it to run properly on all common Unix-based as well as Microsoft Windows platforms. This being said, we have successfully built and tested MRNet on the following systems:

Our build system attempts to use native system compilers where appropriate, for instance, xlc and xlC in AIX environments.

Here we list the third party tools that MRNet uses and needs for proper installation:

MRNet uses GNU autoconf to discover the platform specific configuration parameters. The script that does this auto-configuration is called configure.

UNIX>  ./configure --help

shows all possible options of the command. Below, we display the MRNet-specific ones:

  --with-libfldir                Directory containing flex library

./configure without any options should give reasonable results, but the user may specify certain options. For example,

UNIX> ./configure CXX=g++ CXXFLAGS=-O3 --with-libfldir=/usr/local/lib

instructs the configure script to use g++ for the C++ compiler with level 3 optimization and /usr/local/lib/libfl.a as the location of the flex library.

To build the MRNet toolkit by type:

UNIX>  make

After a successful build, the following files will be present:

Typing:

UNIX>  make mrnet-tests

builds the mrnet test files. In addition to those files above, you will also generate:

UNIX> mrnet_tests.sh [ -l | -r <hostfile> | -a <hostfile> ] [ -f <sharedobject> ]

MRNet is maintained primarily by the Paradyn Project, University of Wisconsin-Madison. Comments and other feedback whether positive or negative are welcome.

Please report bugs to paradyn@cs.wisc.edu.

The MRNet webpage is https://www.paradyn.org/mrnet/

An MRNet end-point represents a tool or application process or node. In particular, they represent the back-end processes in the system. The front-end can communicate in a unicast or multicast fashion with one or more of these end-points as described below.

MRNet uses communicators to represent groups of network end-points. Like communicators in MPI, MRNet communicators provide a handle that identifies a set of end-points for point-to-point, multicast or broadcast communications. MPI applications typically have a non-hierarchical layout of potentially identical processes. In contrast, MRNet enforces a tree-like layout of all processes, rooted at the tool front-end. Accordingly, MRNet communicators are created and managed by the front-end, and communication is only allowed between a tool's front-end and its back-ends, i.e. back-ends cannot interact with each other directly via MRNet.

A stream is a logical channel that connects the front-end to the end-points of a communicator. All tool-level communication via MRNet must use these streams. Streams carry data packets downstream, from the front-end toward the back-ends, and upstream, from the back-ends toward the front-end. Upward streams are expected to carry data of a specific type allowing data aggregation operations to be associated with a stream. The type is specified using a format string (See Appendix A, MRNET Format Strings) similar to those used in C formatted I/O primitives, e.g. a packet whose data is described by the format string "%d %d %f %s" contains two integers followed by a float then a character string. MRNet expands the standard specification to allow for specifiers that describe arrays of integers and floats.

Data Aggregation is the process of merging multiple input data packets and transforming them into one or more output packets. Though it is not necessary for the aggregation to result in less or even different data, aggregations that reduce or modify data values are most common. MRNet uses data filters to aggregate data packets. Filters specify an operation to perform and the type of the data expected on the bound stream. Filter instances are bound to a stream at stream creation. MRNet uses two types of filters: synchronization filters and transformation filters. Synchronization filters organize data packets from downstream nodes into synchronized waves of data packets, while transformation filters operate on the synchronized data packets yielding one or more output packets. A distinction between synchronization and transformation filters is that synchronization filters are independent of the packet data type, but transformation filters operate on packets of a specific type.

Synchronization filters operate on data flowing upstream in the network, receiving packets one at a time and outputting packets only when the specified synchronization criteria has occurred. Synchronization filters provide a mechanism to deal with the asynchronous arrival of packets from children nodes; the synchronizer collects packets and typically aligns them into waves, passing an entire wave onward at the same time. Therefore, synchronization filters do no data transformation and can operate on packets in a type-independent fashion. MRNet currently supports three synchronization modes:

Transformation filters can be used on both upstream and downstream data flows. Transformation filters input a group of synchronized packets, and combine data from multiple packets by performing an aggregation that yields one or more new data packets. Since transformation filters are expected to perform computational operations on data packets, there is a type requirement for the data packets to be passed to this type of filter: the data format string of the stream's packets and the filter must be the same. Transformation operations must be synchronous, but can carry state from one transformation to the next using static storage structures. MRNet provides several transformation filters that should be of general use:

Chapter 7, Adding New Filters describes facilities a tool developer may use to add new filters to the provided set.

A complete description of the MRNet API is in Chapter 5, The MRNet C++ API Reference. This section offers a brief overview only. Using libmrnet.a, a tool can leverage a system of internal processes, instances of the mrnet_commnode program, as a communication substrate. After instantiation of the MRNet network (discussed in the section called “MRNet Instantiation”, the front-end and back-end processes are connected by the internal processes. The connection topology and host assignment of these processes is determined by a configuration file, thus the geometry of MRNet's process tree can be customized to suit the physical topology of the underlying hardware resources. While MRNet can generate a variety of standard topologies, users can easily specify their own topologies; see Chapter 6, MRNET Process-tree Topologies for further discussion.

The MRNet API contains network, end-point, communicator, and stream objects that a tool's front-end and back-end use for communication. The network object is used to instantiate the MRNet network and access end-point objects that represent available tool back-ends. The communicator object is a container for groups of end-points, and streams are used to send data to the end-points in a communicator.

   front_end_main(...) {
1.     Network * net;
2.     Communicator * comm;
3.     Stream * stream;
4.     PacketPtr packet;
5.     int tag = FirstApplicationTag;
6.     float result;

7.     net = new Network(topol_config_file, backend_exe, backend_argv);

8.     comm = net->get_BroadcastCommunicator( );

9.     stream = net->new_Stream(comm, TFILTER_MAX);

10.    stream->send(tag, "%d", FLOAT_MAX_INIT);

11.    stream->recv(&tag, packet)

12.    packet->unpack("%f", &result);
   }

A simplified version of code from an example tool front-end is shown in Example 4.1, “MRNet Front-end Sample Code”. In the front-end code, after some variable definitions in lines 1-6, an instance of the MRNet network is created on line 7 using the topology specification in topol_config_file. In line 8, the newly created network object is queried for an auto-generated broadcast communicator that contains all available end-points. In line 9, this communicator is used to establish a stream for which the MRNet internal processes will use a built-in filter that finds the maximum value of the data sent upstream. The front-end then sends one or more initialization messages to the backends; in our example code on line 10, we broadcast an integer initializer on the new stream. The tag parameter is an application-specific value denoting the nature of the message being transmitted. After the send operation, the front-end performs a blocking stream receive at line 11. This call returns a tag and a packet. Finally, line 12 calls unpack to deserialize the floating point value contained in packet.

   back_end_main(int argc, char** argv) {
1.     Stream * stream;
2.     PacketPtr packet;
3.     int val, tag;
4.     float random_float = (float) random( );

5.     Network * net = new Network(argc,argv);

6.     net->recv(&tag, packet, &stream);

7.     packet->unpack("%d", &val );

8.     if( val == FLOAT_MAX_INIT )
9.         stream->send(tag, "%f", random_float);
   }

Example 4.2, “MRNet Back-end Sample Code” shows the code for the back-end that reciprocates the actions of the front-end. Each tool back-end first connects to the appropriate internal process in line 5, using the back-end version of the network constructor that receives its arguments via the program argument vector (argc/argv). While the front-end makes a stream-specific receive call, the back-ends use a stream-anonymous network receive that returns the tag sent by the front-end, the packet containing the actual data sent, and a stream object representing the stream that the front-end has established. Finally, each back-end sends a scalar floating point value upstream toward the front-end.

A complete example of MRNet code can be found below in Appendix B, A Complete MRNet Example: Integer Addition.

While conceptually simple, creating and connecting the internal processes is complicated by interactions with the various job scheduling systems. In the simplest environments, we can launch jobs manually using facilities like rsh or ssh. In more complex environments, it is necessary to submit all requests to a job management system. In this case, we are constrained by the operations provided by the job manager (and these vary from system to system). We currently support two modes of instantiating MRNet-based tools.

In the first mode of process instantiation, MRNet creates the internal and back-end processes, using the specified MRNet topology configuration to determine the hosts on which the components should be located. First, the front-end consults the configuration and uses rsh or ssh to create internal processes for the first level of the communication tree on the appropriate hosts. Upon instantiation, the newly created processes establish a network connection to the process that created it. The first activity on this connection is a message from parent to child containing the portion of the configuration relevant to that child. The child then uses this information to begin instantiation of the sub-tree rooted at that child. When a sub-tree has been established, the root of that sub-tree sends a report to its parent containing the end-points accessible via that sub-tree. Each internal node establishes its children processes and their respective connections sequentially. However, since the various processes are expected to run on different compute nodes, sub-trees in different branches of the network are created concurrently, maximizing the efficiency of network instantiation.

In the second mode of process instantiation, MRNet relies on a process management system to create some or all of the MRNet processes. This mode accommodates tools that require their back-ends to create, monitor, and control other processes. For example, IBM's POE uses environment variables to pass information, such as the process' rank within the application's global MPI communicator, to the MPI run-time library in each application process. In cases like this, MRNet cannot provide back-end processes with the environment necessary to start MPI application processes. As a result, MRNet creates its internal processes recursively as in the first instantiation mode, but does not instantiate any back-end processes. MRNet then starts the tool back-ends using the process management system to ensure they have the environment needed to create application processes successfully. When starting the back-ends, the front-end must provide them with the information needed to connect to the MRNet internal processes, such as the leaf processes' host names and connection port numbers. This information can be provided via the environment, using shared filesystems or other information services as available on the target system. To collect the necessary information, the front-end can use the MRNet API methods for discovering the network topology details.

Instances of NetworkTopology are network specific, so they are created when a Network is instantiated. MRNet API users should not need to create their own NetworkTopology instances.

Instances of Communicator are network specific, so their creation methods are functions of an instantiated Network object.

Instances of Stream are network specific, so their creation methods are functions of an instantiated Network object.

A Packet encapsulates a chunk of formatted data sent on a Stream. Packets are created using a format string (e.g., "%s %d" describes a null-terminated string followed by a 32-bit integer, and the Packet is said to contain 2 data elements). MRNet front-end and back-end processes do not create instances of Packet; instead they are automatically produced from the formatted data passed to Stream::send. Appendix A, MRNET Format Strings contains the full listing of data types that can be sent in a Packet.

When receiving a Packet via Stream::recv or Network::recv, the Packet instance is stored within a PacketPtr object. PacketPtr is a class based on the Boost library shared_ptr class, and helps with memory management of Packets. A PacketPtr can be assumed to be equivalent to "Packet *", and all operations on Packets require use of PacketPtr.

The first parameter to the Network::Network() front-end constructor is the name of an MRNet topology file. This file defines the topological layout of the front-end, internal nodes, and back-end MRNet processes. In the syntax of the topology file, the hostname:id tuple represents a process with a MRNet instance id running on hostname. It is important to note that the id is of symbolic value only and does not reflect a port or process number associated with the system. A line in the topology file is always of the form:

hostname1:0 => hostname1:1 hostname1:2 ;

meaning a process on hostname1 with MRNet id 0 has two children, with MRNet ids 1 and 2, running on the same host. MRNet will parse the topology file without error if the file properly defines a tree in the mathematical sense (i.e. a tree must have a single root, no cycles, full connection, and no node can be its own descendant).

   hostname1:0 => 
                  hostname1:1 
                  hostname1:2 
                  ;

When the MRNet test programs are built, a topology generator program, $MRNET_ROOT/bin/$MRNET_ARCH/mrnet_topgen, will also be created. The usage of this program is:

mrnet_topgen <OPTION> [INFILE] [OUTFILE]

Create a MRNet topology specification from machine list in INFILE
or standard input, and writes output to OUTFILE or standard output.

        -b topology, --balanced=topology
                Create a balanced tree using "topology" specification. The specification
                is in the format F^D, where F is the fan-out (or out-degree) and D is the
                tree depth. The number of tree leaves (or back-ends) will be F^D.

                Example: "16^3" means a tree of depth 3 with fan-out 16, with 4096 leaves.

        -o topology, --other=topology
                Create a generic tree using "topology" specification. The specification
                for this option is (the agreeably complicated) N:N,N,N:... where N specifies
                the number of children a node has, ',' distinguishes nodes on the same level,
                and ':' separates the tree into levels.

                Example 1: "2:2,2" specifies a tree where the root has 2 children
                           and each child on the 2nd level has 2 children.
                Example 2: "2:8,4" specifies a tree where the root has 2 children.
                           At the 2nd level, the 1st child has 8 children, and the
                           2nd child has 4 children

The specified input machine list must contain enough hosts to support the entire process tree. mrnet_topgen assumes one process will be placed on each host. To place multiple processes on the same host, the list should contain the host's name multiple times.

A filter function has the following signature

packets_in is a reference to a vector of Packets serving as input to the filter function. packets_out is a reference to a vector into which output Packets should be placed. In the rare case where Packets need to be sent in the reverse direction on the Stream, packets_out_reverse should be used instead of packets_out. local_storage may be used to define and maintain state specific to a filter instance. Finally, config_params is a reference to a PacketPtr containing the current configuration settings for the filter instance, as can be set using Stream::set_FilterParameters.

For each filter function defined in a shared object file, there must be a const char * symbol named by the string formed by the concatenation of the filter function name and the suffix "_format_string". For instance, if the filter function is named my_filter_func, the shared object must define a symbol const char *my_filter_func_format_string. The value of this string will be the MRNet format string describing the format of data that the filter can operate on. A value of "" denotes that the filter can operate on data of arbitrary value.

This topic currently pertains to use with the GNU C++ compiler only.

Since we use the C facility dlopen() to dynamically load new filter functions, all C++ symbols must be exported. That is, the symbol definitions must fall with the statements

extern "C" {

and

}

The file that contains the filter functions and format strings may be compiled with the GNU compiler options "-fPIC -shared -rdynamic" to produce a valid shared object.

A front-end that will dynamically load filters must be built with the GNU compiler options "-Wl,-E" to notify the linker export global symbols externally.