Performance
In this chapter, we describe the ways in which the various components within IBM Content Manager OnDemand (Content Manager OnDemand) might be configured or tuned to enhance performance. In most cases, it is not possible to give specific parameter values; however, we provide broad concepts and recommendations in areas where tuning for performance is possible.
In this chapter, we cover the following topics:
13.1 Tuning Content Manager OnDemand to enhance performance
There are two components to performance: throughput and response time.
•Throughput: The number of transactions (Content Manager OnDemand requests) that can be satisfied per unit of time. The more transactions that are run per unit of time, the higher the throughput. Higher throughput implies that more users can be served concurrently and more load jobs can be run in parallel. If the throughput values are low, then the system might not be able to support the required number of users.
•Response time: The amount of time it takes to service a single transaction (Content Manager OnDemand request). Faster response times imply that the users are able to retrieve their data faster from the archive, which in turn leads to more satisfied users. If the response time is slow, then users are dissatisfied with the system.
A high performance system, such as Content Manager OnDemand, provides both a high throughput and a short response time.
The following sections describe the various components of a Content Manager OnDemand system and its architecture, and provide guidance about parameters and configurations that you can change to improve performance.
The ability to separate the object server from the library server has two
main advantages:
•The ability to share workload by dedicating machines to individual tasks
•The ability to reduce the impact of retrieving a large piece of data over a network that is either slow or overloaded
13.1.1 Content Manager OnDemand configuration
How reports are defined, indexed, and stored within Content Manager OnDemand greatly influences the speed at which Content Manager OnDemand can retrieve them. There are various hints and tips for the optimum way of defining reports within Content Manager OnDemand that are described in Chapter 3, “Administration” on page 53.
13.1.2 System logging
Content Manager OnDemand system logging can be used for usage monitoring, chargeback, or troubleshooting. Because system logging involves writing all of the selected log messages to disk, you incur an increase in both resource usage and response time. Logging increases both the amount of processor that is used and the amount of I/O to disk. For this reason, select only the types of logging that you want performed for a particular application group. Depending on your system usage requirements, you might decide to perform any of the
following tasks:
•Turn off all system logging.
•Record a minimal amount of information (only the information that is needed for reporting functions).
•Record all transactions.
•Record the log information to one or more external files using the System
Log exit.
•Turn on system logging only while troubleshooting the system.
•Turn on system logging once every time period to sample the system
usage patterns.
13.1.3 System management
For effective system management, set the appropriate value for ARS_NUM_DBSRVR and correct file systems for various Content Manager OnDemand components.
ARS_NUM_DBSRVR
The ARS_NUM_DBSRVR parameter is set in the ars.cfg file. This is the maximum number of threads that are concurrently opened between the Content Manager OnDemand library server and DB2. Typically, this is set to a number 4 - 30. This number must be large enough to support all of the concurrent database requests from all users / clients and Content Manager OnDemand commands and daemons, such as ARSLOAD, ARSDOC, ARSDB, ARSMAINT, and ARSADMIN. This number must not exceed the number of DB2 batch connections (MAXDBATS for z/OS and MAXAPPLS for MP). The number of DB2 batch connections should be greater than the ARS_NUM_DBSRVR plus all the other connections that are required by all DB2 applications that you have defined in your DB2 configuration.
For systems that are running several large load jobs in parallel, or for systems that have large numbers of active users, increase this parameter from the default of 4.
File systems on UNIX
During the installation and setup of Content Manager OnDemand, one of the tasks is to create the file systems that are required to contain the various Content Manager OnDemand components.
For performance reasons, when the Content Manager OnDemand file systems are created, the following components should not be on the same
physical media:
•The cache file system
•The database file system
•The primary logs file system
•The secondary logs file system
•The load / indexing file system
•The Content Manager OnDemand temporary space file system
13.1.4 Storage management
Regardless of the platforms, storage management with Content Manager OnDemand can be divided into two areas: cache storage that is managed by Content Manager OnDemand and archive media that is managed by an external product, such as Tivoli Storage Manager, object access method (OAM), Virtual Storage Access Method (VSAM), or Archive Storage Manager.
For effective storage management, one of the key performance features with Content Manager OnDemand is its ability to load data to archive media, but simultaneously retain a temporary cached copy of the most recent archived data on fast access media (such as the hard disk drive). The expiration and management of this cached copy of the data is done by Content Manager OnDemand. After a certain predefined period elapses, the data is removed from cache and the only remaining copy is held on the much slower archive media that is managed by either Tivoli Storage Manager, OAM, VSAM, or Archive Storage Manager, depending on the platform.
If performance problems are encountered at the storage manager level, the issue is almost always related to the inherent qualities of the slower media types (such as optical platters and tape volumes) or how the archive media manager
is configured.
Some of the parameters that affect storage management are ARS_NUM_OAMSRVR, ARS_NUM_OAMSRVR_SLOW_RETRIEVE, and ARS_OAM_SLOW_RETRIEVE_THRESHOLD.
ARS_NUM_OAMSRVR
This parameter specifies the maximum number of concurrently attached threads between the Content Manager OnDemand object server and OAM for z/OS. Typically, this is set to a number 4 - 30, depending on client access patterns and object storage locations (disk versus tape). This parameter has a maximum value of 30. Any value larger than 30 result in a U0039 abend.
ARS_NUM_OAMSRVR_SLOW_RETRIEVE
This parameter determines the number of task control blocks (TCBs) that the Content Manager OnDemand server starts to handle connections to OAM for retrievals from objects with a slow retrieval time as defined by the ARS_OAM_SLOW_RETRIEVE_THRESHOLD parameter. The ARS_NUM_OAMSRVR_SLOW_RETRIEVE parameter applies to all object servers. If the value specified for this parameter is zero, no TCBs are dedicated for slow retrievals and all retrievals are processed by the TCBs that are associated with the ARS_NUM_OAMSRVR parameter. The default is zero. The ARS_NUM_OAMSRVR_SLOW_RETRIEVE TCBs are in addition to the ARS_NUM_OAMSRVR TCBs and use additional DB2 connections.
ARS_OAM_SLOW_RETRIEVE_THRESHOLD
This parameter specifies the threshold at which OAM retrievals are processed by the TCBs that are associated with the ARS_NUM_OAMSRVR_SLOW_RETRIEVE parameter. If the estimated retrieval time for an object (as indicated by QELQERRT) is greater than or equal to the value of the ARS_OAM_SLOW_RETRIEVE_THRESHOLD parameter, the OSREQ RETRIEVE is processed by an ARS_NUM_OAMSRVR_SLOW_RETRIEVE TCB. The default value is 12000. For other valid QELQERRT values, see the Object Access Method Application Programmer's Reference, SC35-0425-08. An ARS_OAM_SLOW_RETRIEVE_THRESHOLD value of zero along with a non-zero ARS_NUM_OAMSRVR_SLOW_RETRIEVE value causes all OAM retrieve requests to be processed by the ARS_NUM_OAMSRVR_SLOW_RETRIEVE TCBs, while the ARS_NUM_OAMSRVR TCBs process store, query, and delete requests.
13.2 Data loading performance
The data loading process is illustrated in Figure 13-1. The process begins with the Content Manager OnDemand Administrator Client defining the application group and application parameters for the reports to be loaded. These parameters are stored in the Content Manager OnDemand system tables on the
library server.
Figure 13-1 Data loading
During the load process, in addition to any command-line parameters that are supplied, the application group and application parameters are retrieved from the library server, and then, based on the parameter definitions, the load process completes the following steps:
1. Selects the indexer to be used for indexing the report data and retrieves the indexing parameters.
2. Reads in the report data from the identified source location. The input report data can be of any data type.
3. Indexes the report data based on the defined indexing parameters.
4. Segments the report into “documents”.
5. Compresses the documents.
6. Stores the compressed documents in storage objects (10 MB by default).
7. Sends the storage objects to the object server where they are stored in the identified archive (storage node).
8. Sends the index data for the stored objects to the library server where the indexes are stored in the appropriate application group data table.
13.2.1 Factors that affect the load performance
There are many factors that affect load performance:
•Hardware: The quantity and speed of processors, disks, I/O channels,
and memory
•Network bandwidth and throughput
•The speed and capacity of the available hardware (processors, memory, disks, network, and so on)
•Operating system tuning components: DB2, TCP/IP, and
Language Environment
•Content Manager OnDemand tunable components
•Storage management tunable components: UNIX System Services, zFS, HFS, OAM, Tivoli Storage Manager, and ASM
•Data components:
 – Report file size, document file size (or in the case of large objects, report segment size), and number of documents per report.
 – Number and distribution of triggers, fields, and indexes per document.
 – Data type and required data conversion (if any).
 – Resource collection for AFP and PDF.
 – Document compressibility, which is a function of document data complexity and data type. Text (such as Line Data or SCS) is typically more compressible than AFP, which is typically more compressible
than PDF.
 – Storage object size (10 MB default): Contains 100 KB compressed object, which contains a compressed document.
 – Exit routines/programs.
13.2.2 Recommendations
For the most optimal performance in loading, we recommend the
following practices:
•For Multiplatforms and z/OS, run parallel load jobs to take advantage of multi-processors, large memory pools, multiple data paths, and multiple
disk drives.
•Ensure that each parallel load is loading to a different application group.
•Ensure that you set up a different temp directory for each of the parallel loads. The -c indexDir indexer parameter (which specifies the directory in which the indexer stores temporary data) should always be specified for ARSLOAD and should be unique for each running ARSLOAD process.
•For IBM i, start multiple output queue monitors over a single output queue to improve throughput and take advantage of multi-processors, large memory pools, and multiple disk drives.
•Each Content Manager OnDemand process is limited by the performance of a single processor. For example, the OS/400 indexer uses only one processor when indexing a document. Having two or more processors in your system or LPAR does not improve the performance of the OS/400 indexer. However, having two or more processors in your system or LPAR might allow you to run multiple loads jobs simultaneously. You can start multiple output queue monitors over a single output queue to improve document
load performance.
•For IBM i, the usage of the Merge Spooled Files (MRGSPLFOND) command can provide significant performance improvements when you load SCS
spooled files.
•For IBM i, depending on your retrieval patterns and system hardware configuration, it might be advantageous to not store a duplicate set of documents in the Content Manager OnDemand cache when using ASM because ASM might already be using disk space. If the application group uses ASM, caches the data, and specifies the migration of data at load time, two copies of the data are stored during the load. One copy is stored in cache and one copy is stored in the ASMREQUEST directory.
To avoid storing a duplicate set of documents in cache for non-AFP data, change Cache Data to No on the Storage Management tab of your application group definition. To accomplish this for AFP data, you might change Document Data to No Cache but leave Resource Data in cache for faster retrieval.
•For IBM i, every user loading data should have a home directory. If they do not have a home directory, the temporary files are stored in the root directory of the IFS.
•If the data source is on a remote system, it is possible to either load the data into Content Manager OnDemand on the remote system and directly store the export data to the specified Content Manager OnDemand library and object server, or upload the data to the specified Content Manager OnDemand server through FTP and then load the data on the selected Content Manager OnDemand system.
•For Multiplatforms and z/OS, all file systems should be dedicated file systems that are mounted on their own mount points.
•For z/OS, when loading PDF reports (using the PDF Indexer), placing the input report in the HFS or zFS causes the load to run nearly 50 times faster as compared to the input report being placed in a VSAM file.
13.2.3 Load testing
The goal of load testing is to verify that, under stressful system conditions, the required amount of data can be loaded into the Content Manager OnDemand system within a time window.
Here is general approach to load testing a system:
•Parallel loads: Run a single load and measure the load throughput. If the throughput does not meet the requirements, then run two loads in parallel and measure the throughput. As the loads are run, collect system statistics to determine the system resources being used and any potential bottlenecks. Tune or acquire additional system resources as needed. Progressively increase the number of parallel loads until the required throughput is met.
 
Note: For most users, a single load process meets the ingestion throughput requirements.
•Data types and exits: A different data type, and whether an exit is invoked during the load process, affects the load throughput. Test samples of the different types that represent the general loads.
13.3 Data retrieval performance
All Content Manager OnDemand clients (such as the Windows client, CICS client, IBM Content Navigator, ODWEK APIs, and structured APIs) retrieve data from the Content Manager OnDemand server by using a standard proprietary Content Manager OnDemand protocol. From a Content Manager OnDemand server perspective, there is no difference between one client and another.
13.3.1 Data retrieval parameters
Various parameters affect data retrieval performance.
Folder parameters: General tab
In the Content Manager OnDemand Administrator Client, under the Folder parameter and on the General tab, you have the following option:
•Note Search: If the Annotation flags in a document database are set to No in the Advanced tab of General window of the Application Group, then this option determines when Content Manager OnDemand searches the database for annotations and notifies users that annotations exist for the documents that match a query. Content Manager OnDemand provides three search and notification methods:
 – Hit List: Content Manager OnDemand searches for annotations when the user runs a query. When annotations exist for a document, the client programs display a note icon next to it in the document list. This method has a direct performance impact on the generation of the document list.
 – Retrieve: Content Manager OnDemand searches for annotations when the user selects a document for viewing. This is the default and recommended value.
 – Note: Content Manager OnDemand searches for annotations when the user chooses the Note option while viewing a document.
Folder parameters: Permissions tab
In the Content Manager OnDemand Administrator Client, under Folder parameters and on the Permission tab, you have the following option:
•Max Hits: Determines the maximum number of hits that are retrieved and transmitted to the client. By reducing the maximum number of hits, users are forced to enter queries that better match the documents that they are searching for. This results in a more optimum usage of the system resources both in performing the queries and in downloading the resulting document list.
TCP/IP considerations
There is a known Windows configuration setting that might affect performance when connecting to a Content Manager OnDemand server. During repeated searches and retrievals on a Content Manager OnDemand server, many Windows sockets are opened and closed. Two default Windows settings might impact heavy traffic between the client and the Content Manager
OnDemand server:
•When an application closes a Windows socket, Windows places the sockets port into TIME_WAIT status for 240 seconds; during this time, the port cannot be reused.
•Windows limits the number of ports that an application can use to 5000.
To avoid the problems that might result, change the values for the timeout wait time and number of ports by editing the Windows registry:
•Change the value of the timeout wait time from 240 seconds to a lower number (valid values are 30 - 300 seconds). The key's name is HKEY_Local_MachineSystem CurrentControlSetservicesTcpip ParametersTcpTimedWaitDelay.
•Increase the maximum port number from its default of 5000 to a higher number (valid values are 5000 - 65534). The key's name is HKEY_Local_MachineSystemCurrentControlSet servicesTcpipParametersMaxUserPort.
For more information about TcpTimedWaitDelay and MaxUserPort, consult your Windows documentation.
Verify with your network personnel that the values that are appropriate for your environment are being set correctly.
13.3.2 Factors that affect retrieval performance
Figure 13-2 shows the data retrieval performance testing, which is an illustration of the methodology that is used by the Content Manager OnDemand lab for its internal performance testing. On the client side (where the cTest program is), both throughput and response time are recorded. They are defined as follows:
•Throughput: The amount of work that is performed over a period of time (how many transactions can the Content Manager OnDemand server (CMOD SERVER) run at the same time).
•Response time: The time that is elapsed between when a request is submitted and when the response from that request is returned (how long does it take for a transaction to run).
Maximizing performance is a balancing act between optimizing throughput (which is based on keeping the computing resources busy) and optimizing response times (which requires the computing resources to be available when they are needed). As the throughput increases, so does the response time.
Figure 13-2 Data retrieval performance testing
The concepts that are shown in Figure 13-2 are described here for
your reference.
The retrieval performance is mostly limited by the resources that are available to the Content Manager OnDemand server.
For example, for disk and I/O capacity, each retrieve requires that the data be obtained from the archive (Tivoli Storage Manager, OAM, ASM, and cache). This data is on a disk or some other storage device. The storage device retrieval rate is part of the total response time that is observed at the client, and both of them are affected by the following resources and system demand:
•Real memory: The data that is retrieved from disk must be stored in memory in order for it to be processed. Virtual memory allows for large amounts of data to be swapped in and out of real memory, but does not remove the need for real memory.
•Processing: Any data transformations that are performed on the Content Manager OnDemand server require available processing capability. If the capability is not available, the server waits until it becomes available. This causes the total response time to the client request to be lengthened.
•Concurrent retrievals: Each retrieval requires resources on the server. The higher the number of concurrent retrievals, the larger the amount of resources that are needed to complete the work in an acceptable amount of time.
•Network bandwidth: The retrieved data is sent to the clients over the Internet Protocol network. If the network bandwidth is not wide enough to satisfy all the concurrent requests, then the response time to the clients is slower and data is queued up in the server buffers, further slowing down the system.
13.3.3 Retrieval testing
The goal of retrieval testing is to verify that, under stressful system conditions, the maximum number of concurrent users can still be served while meeting the business requirements. Here is a general approach to retrieval testing
the system:
•Transaction type: Different types of transactions present different types of workloads on the system. For example, logon, document query, and document retrieval all use different components of the Content Manager OnDemand system. For each transaction type, measure the throughput and response time for a number of concurrent users that exceed the maximum predicted number. Tune and add resources to the system as needed until the system exceeds the SLA requirements.
•Data types: The stored documents might be of different sizes and data types (and might invoke preview exits). Multiple document retrieval tests must be run to verify the performance for the various types of stored documents.
•User workloads: The users that access the system might all exhibit the same usage patterns or there might be two or more usage patterns. Here is an example usage pattern:
a. Log on.
b. Wait five seconds.
c. Issue a document query with a maximum hit list size of 12 documents.
d. Wait five seconds.
e. Retrieve a 10 KB document.
f. Wait 40 seconds.
g. Retrieve a 20 KB document.
h. Wait 60 seconds.
i. Log off.
There might be a total of 50 concurrent users that are following this pattern. There also might be other patterns that are being run at the same time. So, the user workload test should model this behavior and be able to meet the business requirements at peak loads.
•Test driver location: The code that is generating the retrieval workload can be installed on either of the following machines:
 – The same server on which the Content Manager OnDemand system
is installed.
If this is the case, it is possible to maximize the stress on the Content Manager OnDemand system by eliminating the network connection and by using system processing cycles to generate and measure the response time and throughput.
 – A network connected workstation.
This situation simulates either a web server that is connected to the Content Manager OnDemand server or a user that is connected to the Content Manager OnDemand server.
•Number of test drivers: The number of systems issuing the requests can be increased so that the number of concurrent requests reaching the Content Manager OnDemand server exceeds the maximum expected number
of requests.
•Test measurement: There are two sets of measurements. The first is at the test driver, which represents the user or Content Manager OnDemand client. At this location, both throughput and response time on a transaction basis should be collected. Also, it is important to check that the system that is issuing the retrieve requests is not overloaded and thus is not the performance bottleneck. In addition, at the Content Manager OnDemand server, request service times can be observed in the Content Manager OnDemand system log. System performance measurements should be collected using operating system-specific tools.
13.3.4 System testing
After the load and retrieval tests are performed individually, it is important to perform an overall system test. This test should include running everything in parallel up to the maximum expected system usage. Everything includes load, retrieval, expiration, migration, duplication, and backup operations. The goal is to ensure that under the most possible stressful condition, the system meets business requirements.
 
Note:
•The performance tuning process demands great skill, knowledge, and experience, and it cannot be performed by only analyzing statistics, graphs, and figures.
•The goal is to tune the Content Manager OnDemand Server. You can only “see” the bottlenecks in the server only if both the client and the network are clear of bottlenecks.
13.4 Performance issues based on data type
This section describes issues that are related to individual data types that can have significant effects on the overall performance of Content Manager OnDemand. Some of these issues can be addressed by selecting or clearing certain functions and features within Content Manager OnDemand. Some of the issues that we describe can be addressed only by changing the way in which the data is produced from the source.
13.4.1 PDF data
Portable Document Format (PDF) data is an increasingly common data type that can be archived within Content Manager OnDemand. Here are the key advantages of using this data type as a document format:
•It is a read-only format that does not require any external resources, such as images or fonts. It is self-contained.
•The viewer for PDF can be downloaded at no charge from the Adobe website and the browser plug-ins for PDF are also available at no charge.
During PDF document creation, resources such as images and custom fonts are placed in the data stream once and then referenced many times from within the PDF file. If a large report is produced from many small documents, then that report requires only one copy of the resources.
However, when the PDF is indexed, the PDF Indexer creates many PDF documents from the input file. Each of these documents requires a certain number of PDF structures, which define a document. These documents are concatenated together in the .out file, and then loaded into Content Manager OnDemand as separate documents. Because the resources are extracted and placed into a separate resource file, they are not included in each document. For an illustration of the process, see Figure 13-3.
Figure 13-3 PDF indexing
If no resource collection is done, then the size of the .out file, which contains all the individual documents, might be larger than the original file. For tips about how to reduce the size of the output file, see 7.3.3, “Reducing output file size with PDF documents” on page 203.
The size of the input and output file can create problems during the load process:
•The temporary space that is used during indexing can be too small and the load fails.
•The maximum input file size that the PDF Indexer can process is 4 GB, but the recommended maximum size for a single document (after indexing) is
50 MB. If this size is exceeded, the system might run out of disk space
or memory.
Create PDF data with the base 14 fonts, which do not need to be included in the PDF file. Because they are not included in the PDF file, they are not extracted during resource collection, which improves performance. For more information about the PDF data stream and fonts, see 7.3.1, “PDF fonts and output file size” on page 196.
13.4.2 Line data
Line data (ASCII or EBCDIC text-based reports) is the most common type of data that is stored in Content Manager OnDemand. The type of line data that we describe here is a special form of transaction style report, where it is necessary to search on a value that appears on every line of the report. This transaction data has a transaction number that appears on every line and must be sorted either by column or row and either ascending or descending.
When indexing transaction data, if each transaction number from each line of the report is treated as a database index, such as date or customer name, then the database becomes very large in a short period. Content Manager OnDemand has a special type of field for transaction data, which is illustrated in Figure 13-4 by the boxed data on the left of the window.
Figure 13-4 Transaction data in graphical indexer
The transaction data field selects the first and last values from a group of pages and only these group level values are inserted into the database. Content Manager OnDemand queries the database by comparing the search value that is entered by the user to two database fields, the beginning value and the ending value. If the value entered by the user falls within the range of both database fields, Content Manager OnDemand adds the item to the document list.
From a performance perspective, using the transaction data field for transaction style line data optimizes indexing performance by reducing the number of index values to be inserted into the database. This means that loading and retrieving these large reports is faster and the Content Manager OnDemand database is many times smaller.
13.4.3 AFP data
Advanced Function Presentation (AFP) data is a multi-part data type. This means that in addition to the variable data itself, there are also external resources, such as images, fonts, and logos, which are referenced by the AFP data stream. When Content Manager OnDemand stores AFP, the resources are also archived. When the data is viewed, the referenced resources are displayed.
It is a common misconception that if fonts are collected when the data is loaded, they are available for viewing in the Windows client. The fact is that Windows does not recognize AFP Fonts. It is not possible to use these fonts even if they are sent to the client as part of the resource. Windows clients require a mapping from AFP Fonts to ATM or TT fonts. Content Manager OnDemand provides this mapping for most standard fonts. For more information about mapping custom fonts, see IBM Content Manager - Windows Client Customization Guide and Reference, SC27-0837.
One possibly useful implementation of storing fonts with the resource group is when server reprint is necessary. If the fonts are stored with the resource group, they can be retrieved from Content Manager OnDemand and used by AFP printers. However, if fonts are collected, they are also sent to the client as part of the resources group and then discarded. Storing the fonts with the resource group serves only to increase network traffic when transferring the resource to the workstation. A more practical option for server printing is to store the font in a fontlib and to keep only the reference (path) to the fontlib. Although the font is accessible on the server, Print Services Facility (PSF) or InfoPrint does not need the font to be inline (stored in the resource group). Using this approach also allows all AFP data that references the font to use the single instance of the font without redundant inline storage.
Figure 13-5 shows the indexer information in the application where you can select the resources to collect with the Restype= parameter. Unless reprints to AFP printers with 100% fidelity is a requirement, do not collect the fonts.
Figure 13-5 Collecting AFP Fonts
The Content Manager OnDemand for i server does not collect the fonts and does not give the administrator that option. The Resource Information window (under Indexer Properties) is not available to the Content Manager OnDemand for i administrator. If you are reprinting to an AFP printer, the fonts must be available on the IBM i server, or font substitution is done.
13.4.4 Image data
To optimize performance with storing and retrieving image formats, such as TIFF, GIF, and JPEG, do not compress the data because the file sizes might increase. To turn off compression, select the Disable option from the Load Information tab within the application. See Figure 13-6.
Figure 13-6 Disabling compression
There are two ways to turn off data compression:
•Disable: Content Manager OnDemand does not compress the input data. Choose this option when the input data, such as PDF and compressed TIFFs, is already compressed. Documents are extracted by the appropriate viewer on the client (for example, Adobe Acrobat Reader).
•None: Content Manager OnDemand does not compress the input data when loading it into the system. When the user selects a document for viewing, Content Manager OnDemand compresses the document before transmitting it over the network and extracts the document at the client.
 
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset