Quick Start Guide
|
Ultra Messaging (Version 6.16.1)
This document provides step-by-step instructions on how to rapidly get started using Ultra Messaging high performance message streaming. It gets you started using our pre-compiled evaluation binaries to test performance, and then walks you through the steps of writing minimal source (publisher) and receiver (subscriber) applications.
For policies and procedures related to Ultra Messaging Technical Support, see UM Support.
(C) Copyright 2004,2023 Informatica Inc. All Rights Reserved.
This software and documentation are provided only under a separate license agreement containing restrictions on use and disclosure. No part of this document may be reproduced or transmitted in any form, by any means (electronic, photocopying, recording or otherwise) without prior consent of Informatica LLC.
A current list of Informatica trademarks is available on the web at https://www.informatica.com/trademarks.html.
Portions of this software and/or documentation are subject to copyright held by third parties. Required third party notices are included with the product.
This software is protected by patents as detailed at https://www.informatica.com/legal/patents.html.
The information in this documentation is subject to change without notice. If you find any problems in this documentation, please report them to us in writing at Informatica LLC 2100 Seaport Blvd. Redwood City, CA 94063.
Informatica products are warranted according to the terms and conditions of the agreements under which they are provided.
INFORMATICA LLC PROVIDES THE INFORMATION IN THIS DOCUMENT "AS IS" WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING WITHOUT ANY WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND ANY WARRANTY OR CONDITION OF NON-INFRINGEMENT.
See UM Glossary for Ultra Messaging terminology, abbreviations, and acronyms.
Ultra Messaging software consists of the documentation package and one or more binary packages. These instructions help get you started evaluating Ultra Messaging quickly. You can run pre-compiled commands from an Ultra Messaging binary distribution to get some quick performance numbers in your environment. A subsequent section guides you through creation of simple programs for local compilation.
A binary package consists of:
Binary versions of source code examples, in the bin directory.
Header files, in the include directory.
Link libraries, in the lib directory.
The Ultra Messaging code and documentation are copyrighted and confidential information owned by Informatica and are covered under the terms of our Software License Agreement or Non-Disclosure Agreement as appropriate. Use of this code and documentation without a valid Non-Disclosure Agreement or Software License Agreement with Informatica Corporation is strictly prohibited. If this code and documentation is being supplied under the terms of a Non Disclosure Agreement, all copies, in any form, must be returned or destroyed at the end of the evaluation period or as requested by Informatica.
The following steps assume that the windows package is installed on all test machines in the standard place:
C:\Program Files\Informatica\
rel-id\
platform-id
where rel-id is the release identifier and platform-id describes the ABI (Application Binary Interface). For example:
C:\Program Files\Informatica\UMQ_6.0\Win2k-x86_64
where rel-id is the release identifier. It is also assumed that the \bin
directory is included in the windows PATH environment variable. This is needed so that the .EXE
and .DLL
files can be found.
Open a command prompt window on the machine you want to use for receiving messages and enter the command lbmrcv topic to start a receiver. Note that topic can be any string. You should see output that looks something like this:
A new line will be printed about once per second showing the elapsed time, messages received, and data received. As long as there are no sources yet running on topic, the number of messages received will continue to be zero.
Open a command prompt window on the machine you want to use for sending messages and enter the command lbmsrc topic to start sending messages. The receiver will automatically discover the source, at which time its output will change to something like this:
With no options given, the source will send 10,000,000 small (25 byte) messages. If you would like to test different size packets or number of packets sent, you can set options. Enter lbmsrc -h at the command line, you will get a list of options you can control.
Ctrl-C
to kill the source or receiver.
The following steps assume that the Unix package is installed on all test machines under a normal user account:
/home/
user-id/lbmeval/
rel-id/
platform-id
where rel-id is the release identifier and platform-id describes the ABI (Application Binary Interface). For example:
/home/jsmith/UMQ_6.0/Linux-glibc-2.5-x86_64
It is also assumed that the /bin
directory is included in the PATH environment variable and the /lib directory is in the appropriate loader library search path environment variable (e.g. LD_LIBRARY_PATH for Linux; see http://bhami.com/rosetta.html for equivalences in other flavors of Unix).
Open a command prompt window on the machine you want to use for receiving messages and enter the command lbmrcv topic to start a receiver. Note that topic can be any string. You should see output that looks something like this:
A new line will be printed about once per second showing the elapsed time, messages received, and data received. As long as there are no sources yet running on topic, the number of messages received will continue to be zero.
Open a command prompt window on the machine you want to use for sending messages and enter the command lbmsrc topic to start sending messages. The receiver will automatically discover the source, at which time its output will change to something like this:
With no options given, the source will send 10,000,000 small (25 byte) messages. If you would like to test different size packets or number of packets sent, you can set options. Enter lbmsrc -h at the command line, you will get a list of options you can control.
There are many other example programs available. See:
For general information on measuring performance, see LBM Reference Performance Tests and UMP Performance Testing Sample Applications.
For information on testing the LBT-RM (multicast) transport, see the following UM Knowledgebase articles:
Microsoft has batching issues in its networking stack, so the performance results can vary. We have a work around if this is an issue for you.
The Linux version was compiled on Fedora Core 1.
The C programs below contain the minimum code and supporting material. Their purpose is to verify that the user's build and run-time environments are set up correctly. They also give a basic introduction to the Ultra Messaging API.
We also have equivalent Java and C# programs available in source form. See MinSrc.java and MinRcv.java or MinSrc.cs and MinRcv.cs. We also have an example of how application callbacks are coded in C++ programs. See minrcv.cpp.
(Most browsers let you right-click on a link and use the "save link target" function, or some variation.)
Note that these programs do not allow the user to override any of the default configuration values. As a result, operation is fixed according to the normal LBM defaults; for example TCP is the transport protocol, topic resolution is performed using multicast, etc. See the Ultra Messaging Configuration Guide.
The Source Code Examples tab on the left panel provides a much richer set of source files that use a wide variety of features. However, those programs double as performance testing tools, so they tend to be more complex than just demonstrating the features. We recommend to first build and run these minimal examples.
This source code example is provided by Informatica for educational and evaluation purposes only.
Error handling in these programs is primitive. A production program would want to have better error handling, but for the purposes of a minimal example, it would just be a distraction. Also, a production program would want to support a configuration file to override default values on options.
When building on Windows, the following notes are applicable.
Make sure the preprocessor variable "WIN32" is defined.
Add the '...\include\lbm
' (under the package install directory) as an additional include directory.
Add lbm.lib and wsock32.lib as Object/library modules.
Add the '...\lib
' as an additional library path.
lbm.dll
can be found when a program is run.
When building on Linux, the following notes are applicable.
Sample build command:
(For programs that use more features, additional UM libraries might be required. See the contents of your installed UM "lib/" directory.)
The appropriate library search path should be updated to include the installed UM "lib/" directory. For example:
Alternatively, the shared library can be copied from the LBM lib/
directory to a directory which is already in your library search path.
Using other flavors of Unix can introduce differences. For example, not all Unixes use "LD_LIBRARY_PATH". For a great resource for learning how Unix flavors differ, see: http://bhami.com/rosetta.html.
Starting in UM version 6.14, the necessary Java JAR files are included in the main package under "java".
See Using UM Java on Unix and Using UM Java on Windows.
For example: "UMP_6.14/java"
For .NET Core on Linux, the necessary .NET files are included in the main package under "bin/dotnet".
For example: "UMP_6.14/Linux-glibc-2.17-x86_64/bin/dotnet"
For .NET Framework on Windows, the necessary .NET files are included in the main package under "bin/dotnet".
For example: "UMP_6.14\Win2k-x86_64\bin\dotnet"
This is a source code listing of a minimal source (sender) program. Examples also include equivalent Java and C# programs available in source form. See MinSrc.java and MinSrc.cs. (There are build .NET instructions in the initial comment blocks of MinSrc.cs and MinRcv.cs which are oriented towards the Windows development environment. For information related to .NET on Linux, see UM .NET on Linux.)
Notes:
Create a context object. A context is an environment in which LBM functions. Note that the first parameter is a pointer to a pointer variable; lbm_context_create() writes the pointer to the context object into "ctx". Also, by passing NULL to the context attribute parameter, the default option values are used. For most applications only a single context is required regardless of how many sources and receivers are created.
Allocate a topic object. A topic object is little more than a string (the topic name). During operation, LBM keeps some state information in the topic object as well. The topic is bound to the containing context, and will also be bound to a source object. Note that the first parameter is a pointer to a pointer variable; lbm_src_topic_alloc() writes the pointer to the topic object into "topic". Also, by passing NULL to the source topic attribute, the default option values are used. The string "Greeting" is the topic string.
Create the source object. A source object is used to send messages. It must be bound to a topic. Note that the first parameter is a pointer to a pointer variable; lbm_src_create() writes the pointer to the source object to into "src". Use of the third and fourth parameters is optional but recommended in a production program - some source events can be important to the application. The last parameter is an optional event queue (not used in this example).
Need to wait for receivers to find us before first send. There are other ways to accomplish this, but sleep is easy. See Avoiding or Minimizing Delay Before Sending for details.
Send a message to the "Greeting" topic. The flags make sure the call to lbm_src_send() doesn't return until the message is sent. Note that while this ensures low latency, flushing every message carries a heavy efficiency cost. See Intelligent Batching for more information.
For some transport types (mostly UDP-based), a short delay before deleting the source is advisable. Even though the message is sent, there may have been packet loss, and some transports need a bit of time to request re-transmission. Also, if the above lbm_src_send() call didn't include the flush, some time might also be needed to empty the batching buffer.
This is a source code listing of a minimal receiver program. Examples also include equivalent Java, C#, and C++ programs available in source form. See MinRcv.java, MinRcv.cs, and minrcv.cpp. (There are build .NET instructions in the initial comment blocks of MinSrc.cs and MinRcv.cs which are oriented towards the Windows development environment. For information related to .NET on Linux, see UM .NET on Linux.)
Notes:
LBM passes received messages to the application by means of a callback. I.e. the LBM context thread reads the network socket, performs its higher-level protocol functions, and then calls an application-level function that was set up during initialization. This callback function has some severe limitations placed upon it. It must execute very quickly; any potentially blocking calls it might make will interfere with the proper execution of the LBM context thread. One common desire is for the receive function to send an LBM message (via lbm_src_send()), however this has the potential to produce a deadlock condition. If it is desired for the receive callback function to call LBM or other potentially blocking functions, it is strongly advised to make use of an event queue, which causes the callback to be executed from an application thread. See the example tool lbmrcvq.c for an example of using a receiver event queue.
Note - printf can block, which is normally a bad idea for a callback (unless an event queue is being used). However, for this minimal application, only one message is expected.
Create a context object. A context is an environment in which LBM functions. Note that the first parameter is a pointer to a pointer variable; lbm_context_create() writes the pointer to the context object into "ctx". Also, by passing NULL to the context attribute parameter, the default option values are used. For most applications only a single context is required regardless of how many sources and receivers are created.
Lookup a topic object. A topic object is little more than a string (the topic name). During operation, LBM keeps some state information in the topic object as well. The topic is bound to the containing context, and will also be bound to a receiver object. Note that the first parameter is a pointer to a pointer variable; lbm_rcv_topic_lookup() writes the pointer to the topic object into "topic". Also, by passing NULL to the source topic attribute, the default option values are used. The string "Greeting" is the topic string.
One of Ultra Messaging's great strengths is its use of the network hardware and protocols to achieve very high performance and scalability. By default, Ultra Messaging uses multicast for topic resolution so that data sources and receivers can find each other.
In general, we recommend the use of multicast whenever possible because it provides the best performance and scalability. However, we recognize that it is not always possible to provide multicast connectivity between the machines. For those cases, we support unicast-only operation.
There are two parts to unicast-only operation. One is to use a unicast form of transport. Ultra Messaging uses TCP by default for transport, but LBT-RU is also available and has some performance advantages. For initial "quick start" testing, we recommend not changing the default TCP.
The second part of unicast-only operation is to use the lbmrd (UM Resolver Daemon) for topic resolution. (NOTE: this does not route message data through a daemon. This is a helper process that lets data sources and receivers find each other.) To enable unicast topic resolution, do the following:
Choose one machine to host the resolver daemon and enter the lbmrd command (use lbmrd -h for full instructions).
Create a Ultra Messaging configuration file containing:
context resolver_unicast_daemon [Iface[:Src_Port]->]IP:Dest_Port
where:
Iface is the interface to use (previously set via resolver_unicast_interface).
Src_Port is the source port to use (previously resolver_unicast_port).
IP is the resolver daemon's IP address (previously resolver_unicast_address).
For more information, see resolver_unicast_daemon (context).
Due to the fact that the minimal source programs presented in this document (minsrc.c and minrcv.c) do not allow the use of a configuration file, it is not possible to configure them for unicast-only operation. If multicast operation is not possible on your network, then please use the binary test programs (described in Ultra Messaging Binary Quick Start) which which do allow unicast-only configuration.
The DRO bridges disjoint topic resolution domains by forwarding multicast and/or unicast topic resolution traffic. Before installing and starting a DRO daemon (tnwgd) clear objectives and proper planning are very important. Approach this planning with the consideration that a DRO condenses your network into a single process. You must be clear about the traffic you expect to forward through a DRO. The following highlights some other specifics.
Know your Topic Resolution Domains. See Topic Resolution Domains.
Consider the size and quality of network paths into and out of your DROs. For example, DROs cannot efficiently forward messages from a 1GB path to a 100MB path.
The UMM Daemon manages UM XML configuration information defined in the GUI as well as user and license information. It serves license and configuration information to UM applications. The UMM Daemon requires either a MySQL or Oracle database installation. Informatica does not supply MySQL or Oracle nor any licensing to use them. For additional information about the steps in this process, see the Ultra Messaging Manager Guide.
Install MySQL or Oracle according to the user documentation.
Create a database. (You provide the UMM Daemon with the database name, username and password at start up. This user name and password should be kept secure.)
Log into the database.
From the database's command line or home page, execute the /UMM/install_tables_oracle.sql or source /UMM/install_tables_mysql.sql script. (This script installs the tables required by UMM.)
From the database's command line or home page, load /UMM/oracle_templates.txt or /UMM/mysql_templates.txt. (This script loads the example configuration templates for low latency and high throughput. These templates appear in the UMM GUI under Templates in the object tree in the left pane.
For Oracle, load /UMM/oracle_templates.txt.
Modify the /UMM/umm.properties to indicate the database username, password and driver to reflect your configuration database.
Edit the ummd.bat or ummd.sh file that starts UMM Daemon. The appropriate driver jar file name must be added to the classpath. Specify your certificate file or Java keystore information if using SSL certificates. The ummd.bat or ummd.sh provided with installation appears below.
java -cp .;mysql-connector-java-5.0.8-bin.jar;UMMD_1.6.0_02.jar -Xms512m -Xmx1024m umm.ummd.Server
Start the UMM Daemon with /UMM/ummd.bat or /UMM/ummd.sh.
Edit the umm.bat or umm.sh file that starts the UMM GUI. The appropriate driver jar file name must be added to the classpath. The umm.bat or umm.sh provided with installation appears below.
java -cp .;mysql-connector-java-5.0.8-bin.jar;UMM_1.jar; -Xms256m -Xmx1024m umm.gui.MainFrame
Start the UMM GUI with /UMM/umm.bat or /UMM/umm.sh.
Log into the UMM GUI with username, default and password default. Either change the default user's password or create new administrative user and delete the default user. See the Ultra Messaging Manager Guide.
To enable your UM applications to be served configuration information by UMM Daemon, set the following environment variable for every application/user combination.
export LBM_UMM_INFO=
application_name:
user_name:
password@
ip:
port
For example:
export LBM_UMM_INFO=lbmrcv:lbmrcv:default:default@10.29.3.95:21273
The daemon, umestored, provides persistent store services. To start umestored, perform the following steps:
Create the cache and state directories.
$ mkdir umestored-cache ; mkdir umestored-state
Create a simple umestored XML configuration file or use a sample configuration, ume-example-config.xml. For full details on Persistence, see Ultra Messaging Guide for Persistence.
$ umestored config.xml
See also Umestored Man Page.
The Ultra Messaging Concepts Guide introduces the important fundamental concepts you will need to design and code your applications.
See C example programs for descriptions and source files for the binary tools used for evaluation and performance measurements. They can be a valuable resource for seeing how various features can be implemented. Also available are Java and .NET example programs.
The Ultra Messaging Guide for Persistence contains concepts, tutorials and configuration information for UMP, along with discussions about designing persistent applications. It also contains detailed configuration information for the umestored daemon.
The Ultra Messaging Guide for Queuing contains concepts, tutorials and configuration information for UMQ, along with discussions about designing queuing applications.
The C API, Java API, .NET API documentation provide reference guides for the Ultra Messaging libraries.