Reference <-

transport_session_maximum_buffer (source) <-

: Value used to control the maximum amount of data buffered in UM for the transport session used for the topic.

: For the normal multiple receiver behavior, this value represents the total buffered by all TCP receivers. For the bounded_latency and source_paced multiple receiver behavior, this value represents the individual receiver buffered amount. This option affects the transport session underlying the source rather than the source itself. The transport session uses the value from the first source created on the session and ignores subsequent sources' configuration.

: Refer to Source Object for additional information.

Scope:	source
Type:	lbm_ulong_t
Units:	bytes
Default value:	65536
When to Set:	Can only be set during object initialization.

transport_tcp_activity_method (receiver) <-

: The type of timeout method to use for TCP receivers to detect "half-open" TCP connections.

: For TCP sessions only.

: This defines how transport_tcp_activity_timeout (receiver) is interpreted. Note that transport_tcp_activity_timeout (receiver) defaults to 0 (disabled), meaning that half-open TCP connections may not be detected in a timely way.

Scope:	receiver
Type:	int
When to Set:	Can only be set during object initialization.
Version:	This option was implemented in LBM 3.3.8/UME 2.0.6.

String value	Integer value	Description
"timer"	LBM_RCV_TOPIC_ATTR_TCP_ACTIVITY_TIMEOUT_TIMER	Timer method that requires new TCP session data to be sent to determine if the connection is alive. Default for all.
"SO_KEEPALIVE"	LBM_RCV_TOPIC_ATTR_TCP_ACTIVITY_TIMEOUT_SO_KEEPALIVE	Set SO_KEEPALIVE on the TCP connection which uses the TCP keepalive support in the operating system to determine if the connection is alive. When you use the SO_KEEPALIVE method, UM uses transport_tcp_activity_timeout (receiver) value to set the idle and probe times for SO_KEEPALIVE. The idle time is 90% of the timeout value at most. The probe time is 10% with 10 seconds as the minimum.

transport_tcp_activity_timeout (receiver) <-

: A timeout used by a receiver to close a TCP transport session that has no activity.

: For TCP sessions only.

: Normally, when a source transport session is deleted by the sending application, the TCP connection is closed, which the receiver detects within a few milliseconds. However, there are unusual situations where a temporary network outage prevents the receiver from detecting the closing of the connection, resulting in a "half-open" connection. This situation can prevent the receiver from detecting the closed connection for an unbounded time.

: This timeout can be used to detect and close half-open connections.

: This option has two different meanings, depending on the setting of transport_tcp_activity_method (receiver). See that option for details.

: A value greater than zero turns the timer on.

Scope:	receiver
Type:	lbm_ulong_t
Units:	milliseconds
Default value:	0
When to Set:	Can only be set during object initialization.

transport_tcp_activity_timeout (source) <-

: This timeout option enables a source to use SO_KEEPALIVE to detect when a receiver does not cleanly disconnect or is no longer reachable from the source.

: For TCP sessions only.

: When the timeout expires, a DISCONNECT source event is delivered. This option affects only Linux or Windows platforms. Outstanding TCP retransmit timers must expire before this timer starts. With a default Linux or Windows system configuration, TCP retransmit timers may take minutes or even hours to expire. Therefore the total time taken to detect an unreachable receiver may be significantly higher than the value configured for this timeout. A value of zero (0) defers TCP timeout to OS settings.

Scope:	source
Type:	lbm_ulong_t
Units:	milliseconds
Default value:	0
When to Set:	Can only be set during object initialization.

transport_tcp_coalesce_threshold (source) <-

: UM passes implicitly batched messages to the Operating System sendmsg() as a set unless the size of the set exceeds the coalescing threshold at which point the messages are coalesced and passed to the O/S as one copy.

: This option accommodates the different number of iovecs supported by different O/Ss. Tuning this option balances the efficiency of less iovecs handled by the OS vs. the expense of an additional copy operation of the messages before sending. The default values are also the maximum allowable values.

Scope:	source
Type:	int
Units:	number of individual messages
Default value:	1024 for Linux, Microsoft Windows, Darwin; 16 for Solaris, AIX, HPUX
When to Set:	Can only be set during object initialization.
Version:	This option was implemented in LBM 3.6/UME 2.3.

transport_tcp_datagram_max_size (context) <-

: The maximum datagram size that can be generated for a TCP transport session. While TCP does not use UDP datagrams, this option limits the size of the UM message which is given to the underlying transport type, including all UM headers and overhead. It does not include TCP, IP, or packet overhead added by the network stack. The default value is 65535, the minimum is 500 bytes, and the maximum is 65535.

: See Message Fragmentation and Reassembly for more information.

: Informatica does not recommend setting datagram max size options to the network MTU. See Datagram Max Size and Network MTU.

Warning: When the DRO is in use, it is recommended that all UM applications and components (including the DRO and Persistent Store) share the same maximum datagram size setting. See Protocol Conversion.

: Users of kernel-bypass drivers should also see Dynamic Fragmentation Reduction.

Scope:	context
Type:	lbm_uint_t
Units:	bytes
Default value:	65535
When to Set:	Can only be set during object initialization.
Version:	This option was implemented in LBM 4.1/UME 3.1/UMQ 1.1

transport_tcp_dro_loss_recovery_timeout (receiver) <-

: For TCP transport sessions originating from a DRO endpoint portal, delay declaring as unrecoverable a lost message.

: Message streams traversing a DRO can have the message order changed. If the DRO's outgoing transport session uses the TCP protocol, these out-of-order messages will normally trigger immediate unrecoverable loss. This timeout allows an opportunity for the messages to be re-ordered properly.

: The value 0 disables this delay (i.e. receivers immediately declare unrecoverable loss).

: See DRO Reliable Loss for more information.

Scope:	receiver
Type:	lbm_ulong_t
Units:	milliseconds
Default value:	0
When to Set:	Can only be set during object initialization.
Version:	This option was implemented in LBM 6.12

transport_tcp_exclusiveaddr (source) <-

: Indicate whether the TCP session should set SO_EXCLUSIVEADDRUSE or not before it binds.

: Applicable only to Windows. The default setting in Windows allows multiple binds to the same port. By default, UM will set SO_EXCLUSIVEADDRUSE to minimize port sharing. Refer to Microsoft's web site for more information on SO_EXCLUSIVEADDRUSE.

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.

Value

Description

1

Set SO_EXCLUSIVEADDRUSE. Default for Windows.

0

Do not set SO_EXCLUSIVEADDRUSE.

transport_tcp_listen_backlog (source) <-

: The backlog used in the TCP listen() call to set the queue length for incoming connections.

: If 20 or more receivers will be joining this source, it may be beneficial to increase this number.

Scope:	source
Type:	int
Units:	number of queued connections
Default value:	5
When to Set:	Can only be set during object initialization.

transport_tcp_multiple_receiver_behavior (source) <-

: This option determines the flow control behavior of a TCP source.

: In particular, it addresses the scenario where some receivers are consuming messages more slowly than the source is sending. With normal receiver-pacing, the slowest receiver connected to the source will limit the maximum send rate for the source. I.e. if the source is sending faster than the receiver can process and the socket buffers fill up, the next call to "send" will either block or will return -1 with the error code LBM_EWOULDBLOCK. (But see TCP Flow Control Restrictions.)

: With source-paced behavior, a source send will drop messages for receivers whose socket buffers are full. I.e. fast receivers might get all messages, but slow receivers can see unrecoverable loss.

: See Source Object for additional information.

: FYI - the purpose of this option has evolved over time, and its name is no longer very descriptive of its function.

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.

String value

Integer value

Description

"normal"

LBM_SRC_TOPIC_ATTR_TCP_MULTI_RECV_NORMAL

Source will block for full socket buffers. Default for all.

"source_paced"

LBM_SRC_TOPIC_ATTR_TCP_MULTI_RECV_SOURCE_PACED

Source will drop messages for full socket buffers, resulting in unrecoverable loss for the slow receiver.

"bounded_latency"

LBM_SRC_TOPIC_ATTR_TCP_MULTI_RECV_BOUNDED_LATENCY

The application sends as fast as the fastest receiver can consume data even if recent data headed for slower receivers must be discarded. DEPRECATED since UM 6.9.

transport_tcp_multiple_receiver_send_order (source) <-

: In the case of multiple receivers, this option determines whether datagrams are sent to each receiver in the established order of receivers, or if receivers are selected in random order for each datagram transmission.

: Using random ordering can avoid giving one receiver a consistent latency advantage, at the expense of slightly higher per-message processing (calculating the random number).

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.

String value

Integer value

Description

"serial"

LBM_SRC_TOPIC_ATTR_TCP_MULTI_RECV_SEND_ORDER_SERIAL

Select receivers to receive a datagram based on current established order. Default for all.

"random"

LBM_SRC_TOPIC_ATTR_TCP_MULTI_RECV_SEND_ORDER_RANDOM

For each datagram sent, select receivers in random order, for the sake of "fairness". Note that this option adds a small amount of CPU overhead.

transport_tcp_nodelay (source) <-

: Controls whether the context sets TCP_NODELAY before it binds to the source port.

: Setting TCP_NODELAY disables Nagle's algorithm, which somewhat decreases the efficiency and throughput of TCP, but decreases the latency of individual messages.

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.

Value

Description

1

TCP transport sockets should set TCP_NODELAY (disable Nagle). Default for all.

0

TCP transport sockets should not set TCP_NODELAY (leave Nagle enabled).

transport_tcp_receiver_socket_buffer (context) <-

: Value used to set SO_RCVBUF value of the TCP receivers for topics. In some cases the OS will not allow all of this value to be used.

: A value of 0 instructs UM to use the default OS values. See Socket Buffer Sizes for platform-dependent information.

Scope:	context
Type:	lbm_ulong_t
Units:	bytes
Default value:	0 (use TCP autotuning)
When to Set:	Can only be set during object initialization.

transport_tcp_reuseaddr (source) <-

: Whether the TCP session should set SO_REUSEADDR or not before it binds.

Warning: This option is not recommended for Microsoft Windows users because the SO_REUSEADDR socket option in Windows allows a socket to forcibly bind to a port in use by another socket. Multiple sockets using the same port results in indeterminate behavior.

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.

Value

Description

1

Set SO_REUSEADDR.

0

Do not set SO_REUSEADDR. Default for all.

transport_tcp_sender_socket_buffer (source) <-

: Value used to set the SO_SNDBUF value of the TCP session.

: In some cases the OS will not allow all of this value to be used. A value of 0 instructs UM to use the OS defaults. See Socket Buffer Sizes for platform-dependent information.

Scope:	source
Type:	lbm_ulong_t
Units:	bytes
Default value:	0 (use TCP autotuning)
When to Set:	Can only be set during object initialization.

transport_tcp_use_session_id (source) <-

: Control whether a session ID is used for TCP Transport sessions.

: This option should be set to 0 if a version 6.0 (and beyond) TCP source must interoperate with a version pre-6.0 receiver.

Scope:	source
Type:	int
When to Set:	Can only be set during object initialization.
Version:	This option was implemented in UM 6.0

Value

Description

1

Indicates the application desires TCP to use a session ID. Default for all.

0

Indicates the application does not desire TCP to use a session ID. For use when version pre-6.0 receivers must be used with TCP sources that are version 6.0 and beyond.