Document revision date: 15 July 2002 | |
Previous | Contents | Index |
The RECALL/ALL command now accepts a parameter that allows you to display multiple occurrences of an individual command within the recall buffer.
For example:
$ RECALL/ALL SHOW 1 show system 4 show users 5 show intrusion 6 show error 7 show time |
You can now use the new logical, CDU$TABLE_SPACE, to increase the value of virtual memory allocated by the SET COMMAND/OBJECT command.
Prior to Version 7.3--1, when compiling very large command definition (CLD) files, you may have encountered the following error:
%CDU-F-INTNODESPACE, Internal error: node space exhausted |
The default value of virtual memory allocated by SET COMMAND/OBJECT is 384KB. This value may not be sufficient to process large CLD files.
To increase this size, use the CDU$TABLE_SPACE logical name to specify a larger value and reissue the command. For example:
$ DEFINE CDU$TABLE_SPACE 2048 $ SET COMMAND/OBJECT MYCLD.CLD $ DIR *.OBJ Directory $1$DKA600:[GUY.CDU] MYCLD.OBJ;1 |
You can now use the SHOW DEVICES command to display free space in bytes to accommodate large disk sizes.
Prior to Version 7.3--1, the SHOW DEVICES command displayed free space only in blocks.
Beginning with Version 7.3--1, you can use the SHOW DEVICES/SIZE=BYTES command to display free space in bytes. For example:
$ SHOW DEVICES DKB/SIZE=BYTES Device Device Error Volume Free Trans Mnt Name Status Count Label Space Count Cnt $2$DKB0: (IPG32) Mounted 0 ENG_IMAGES 5.41GB 1 3 $2$DKB100: (IPG32) Online 0 $2$DKB300: (IPG32) Online 0 $2$DKB400: (IPG32) Online 0 (remote shadow member) $2$DKB500: (IPG32) Mounted 0 GUY 621.37MB 1 4 $2$DKB600: (IPG32) Mounted 0 DATA 11.54GB 1 4 $ |
Note that the SHOW DEVICES/FULL command still displays free space in
blocks.
3.5 WWPPS Adds Support for GB18030 Codeset
The World-Wide PostScript Printing Subsystem (WWPPS) now supports the new, compulsory standard, codeset GB18030-2000 for simplified Chinese and traditional Chinese. Support for codeset GB18030 is required for all vendors who want to do business in the People's Republic of China (PRC).
For more information about WWPPS, refer to the OpenVMS User's Manual.
This chapter provides information about new features, changes, and
enhancements for system managers.
4.1 ACCOUNTING Utility Has New Qualifier
The /WIDE qualifier in the ACCOUNTING utility changes the width of Buffered I/O and Direct I/O fields in a report from 8 to 10 characters. Without the /WIDE qualifier, these fields print asterisks (*****), if the field overflows.
For more information, refer to the OpenVMS System Management Utilities Reference Manual.
4.2 AlphaServer DS25
OpenVMS Alpha Version 7.3--1 supports the AlphaServer DS25, the newest
addition to the AlphaServer DS20 family. With a processor speed of 1
GHz and support for the latest Gigabit Ethernet adaptor architecture,
the AlphaServer DS25 provides superior performance in a high-powered
computing environment.
4.2.1 Broadcom 5700 Series Gigabit Ethernet
The AlphaServer DS25 supports the 3COM Gigabit Ethernet adaptor (DEGXA)
as a LAN device and a cluster interconnect device. This support will be
included in future AlphaServers as well.
4.2.2 Use New CDRECORD.COM to Create a CD-ROM
Read and write-once support for CD-R and CD-RW drives is introduced in OpenVMS Alpha Version 7.3--1 on the AlphaServer DS25 system. On these systems, and future AlphaServer systems as well, you can create your own CD-ROMs for distributing or backing up files.
To create a CD-ROM, you need a CD-Recordable (CD-R or CD-RW) drive and a blank CD-R disk. CD-R and CD-RW drives use a laser beam to write (or burn) data to a blank CD-R disk. A CD-R disk is "write once." This means you can write data on it one time only. It is not rewritable.
The write process creates a CD-ROM in Files-11 format. Any supported CD-ROM reader on a computer running OpenVMS will be able to read the CD-ROMs you create.
You can create a CD-ROM that contains data files, but audio recording is not supported at this time. |
OpenVMS supports only qualified CD-R and CD-RW drives. For more information on Alpha systems and the drives they support, refer to the appropriate page at the following Web site:
http://www.compaq.com/alphaserver/configure.html |
You cannot use COPY commands to transfer files from a hard drive to a CD-R disk. You use a program called CDRECORD.COM.1 The sources for CDRECORD.COM are provided on the Open Source Tools for OpenVMS CD-ROM (see Section 6.8.) For more detailed information on using CDRECORD.COM, refer to the OpenVMS System Manager's Manual, Volume 1: Essentials.
1 The CDRECORD software internals were developed outside of Compaq, and are protected by the GNU General Public License Version 2. A copy of the sources and the GNU license are on the Compaq OpenVMS Freeware Web page: http://www.openvms.compaq.com/openvms/freeware/ |
4.3 ANALYZE/DISK_STRUCTURE Utility Has New Qualifier
The /LOCK_VOLUME qualifier in the ANALYZE/DISK_STRUCTURE utility provides a way to prevent file system activity on a volume while you are analyzing that volume with the ANALYZE/DISK_STRUCTURE utility.
For more information, refer to the OpenVMS System Management Utilities Reference Manual.
4.4 Authentication and Credentials Management Extensions (ACME) Subsystem
The Authentication and Credentials Management Extensions (ACME) subsystem provides authentication and persona-based credential services.
Applications can use these services to interact with the user to perform one or more of the following functions:
ACME supports standard OpenVMS authentication and external authentication policies; therefore, applications utilize the same mechanisms as used by the system's LOGINOUT and SET PASSWORD components.
For general information on the ACME subsystem, refer to OpenVMS Guide to System Security.
Information about the ACME subsystem components is located in the following manuals:
ACME Component | Documentation Location |
---|---|
SYS$ACM system service | OpenVMS System Services Reference Manual and OpenVMS Programming Concepts Manual |
ACME_SERVER process and ACME policy-provider agents | OpenVMS Guide to System Security |
SET and SHOW SERVER ACME DCL commands | OpenVMS DCL Dictionary |
The BACKUP command /DENSITY qualifier allows you to specify the recording density of an output magnetic tape. The following table shows the new /DENSITY keywords that are supported beginning in OpenVMS Version 7.3--1:
Keyword | Description |
---|---|
AIT1 | Sony Advanced Intelligent Tape 1 - Alpha only |
AIT2 | Sony Advanced Intelligent Tape 2 - Alpha only |
AIT3 | Sony Advanced Intelligent Tape 3 - Alpha only |
AIT4 | Sony Advanced Intelligent Tape 4 - Alpha only |
DLT8000 | DLT 8000 - Alpha only |
8900 | Exabyte 8900 - Alpha only |
SDLT | SuperDLT1 - Alpha only |
SDLT320 | SuperDLT320 - Alpha only |
Refer to the OpenVMS System Management Utilities Reference Manual for more information.
4.6 Batch Job Support for NUMA Resource Affinity Domains
This section describes updates to the OpenVMS batch processing subsystem in support of Resource Affinity Domains (RADs) in a NUMA environment.
System managers now have the ability to assign batch queues to a RAD, and users can specify a RAD on which to run a batch job.
These new features are restricted for use on batch execution queues and batch jobs.
For information about the system services updates to $GETQUI and
$SNDJBC, see Section 5.15. For information about DCL commands, refer to
the OpenVMS DCL Dictionary.
4.6.1 Batch Queue Level RAD Support
A new qualifier, /RAD, is available for use with the following DCL commands: INITIALIZE/QUEUE, SET/QUEUE, and START/QUEUE. The system manager specifies the RAD number on which to run batch jobs assigned to the queue.
The RAD value is validated as a positive integer between 0 and SYI$_RAD_MAX_RADS. The SHOW/QUEUE/FULL command now displays the RAD in its output, and the F$GETQUI lexical function now accepts a new RAD item.
This section describes a sequence of commands and their effects on a batch queue. A SHOW command is included in each example to illustrate the batch queue modifications.
$ INITIALIZE/QUEUE/ON=QUEBID::/BATCH/RAD=0 BATCHQ1 $ SHOW QUEUE/FULL BATCHQ1 Batch queue BATCHQ1, stopped, QUEBID:: /BASE_PRIORITY=4 /JOB_LIMIT=1 /OWNER=[SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S) /RAD=0 |
$ START/QUEUE/RAD=1 BATCHQ1 $ SHOW QUEUE/FULL BATCHQ1 Batch queue BATCHQ1, idle, on QUEBID:: /BASE_PRIORITY=4 /JOB_LIMIT=3 /OWNER=[SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S) /RAD=1 |
$ SET/QUEUE/RAD=0 BATCHQ1 $ SHOW QUEUE/FULL BATCHQ1 Batch queue BATCHQ1, idle, on QUEBID:: /BASE_PRIORITY=4 /JOB_LIMIT=3 /OWNER=[SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S) /RAD=0 |
$ SET/QUEUE/NORAD BATCHQ1 $ SHOW QUEUE/FULL BATCHQ1 Batch queue BATCHQ1, idle, on QUEBID:: /BASE_PRIORITY=4 /JOB_LIMIT=3 /OWNER=[SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S) |
$ WRITE SYS$OUTPUT F$GETQUI("DISPLAY_QUEUE","RAD","BATCHQ1") -1 |
The new qualifier /RAD is added for use with the following commands: SUBMIT and SET ENTRY.
The user specifies in the qualifier value the RAD number on which the submitted batch job is to execute. The SHOW ENTRY and SHOW QUEUE/FULL commands are enhanced to list the RAD setting on batch jobs.
When a job is submitted to a batch queue that does not have a RAD setting, the job will execute using the RAD specified on the SUBMIT command.
The following command submits TEST.COM to queue ANYRADQ. There is no RAD setting on queue ANYRADQ:
$ SUBMIT/HOLD/QUEUE=ANYRADQ /RAD=1 TEST.COM Job TEST (queue ANYRADQ, entry 23) holding $ SHOW ENTRY/FULL 23 Entry Jobname Username Blocks Status ----- ------- -------- ------ ------ 23 TEST SYSTEM Holding On idle batch queue ANYRADQ Submitted 24-JUL-2001 14:19:37.44 /KEEP /NOPRINT /PRIORITY=100 /RAD=1 File: _$1$DKB200:[SWEENEY.CLIUTL]TEST.COM;1 |
When a job is submitted to a batch queue that does have a RAD setting, the job executes using the RAD specified on the queue, regardless of the RAD specified on the SUBMIT command. This behavior is consistent with other batch system features.
The queue, BATCHQ1, is defined with /RAD=0. The following SUBMIT command creates a job that runs on RAD 0, even though the user specified RAD 1 on the submission:
$ SUBMIT/HOLD/QUEUE=BATCHQ1 /RAD=1 TEST.COM Job TEST (queue BATCHQ1, entry 24) holding $ SHOW ENTRY 24/FULL Entry Jobname Username Blocks Status ----- ------- -------- ------ ------ 24 TEST SYSTEM Holding On idle batch queue BATCHQ1 Submitted 24-JUL-2001 14:23:10.37 /KEEP /NOPRINT /PRIORITY=100 /RAD=0 File: _$1$DKB200:[SWEENEY.CLIUTL]TEST.COM;2 |
When you specify a RAD on a batch job, the job controller creates the process with the $CREPRC home_rad argument set to the RAD value on the job.
If the RAD specified on the job is invalid on the target system, the
job fails to execute and the job controller sends a message to the
operator console indicating that a bad RAD was specified. If the bad
RAD value matches the RAD setting on the batch queue, the batch queue
is stopped and the job remains in the queue.
4.6.3.1 Error Processing
The following is an example of run-time error processing:
SYSTEM@QUEBID> SUBMIT/NONOTIFY/NOLOG/QUEUE=BATCHQ1 TEST.COM Job TEST (queue BATCHQ1, entry 30) started on BATCHQ1 OPCOM MESSAGES SYSTEM@QUEBID> START/QUEUE BATCHQ1 %%%%%%%%%%% OPCOM 25-JUL-2001 16:15:48.52 %%%%%%%%%%% Message from user SYSTEM on QUEBID %JBC-E-FAILCREPRC, job controller could not create a process %%%%%%%%%%% OPCOM 25-JUL-2001 16:15:48.53 %%%%%%%%%%% Message from user SYSTEM on QUEBID -SYSTEM-E-BADRAD, bad RAD specified %%%%%%%%%%% OPCOM 25-JUL-2001 16:15:48.54 %%%%%%%%%%% Message from user SYSTEM on QUEBID %QMAN-E-CREPRCSTOP, failed to create a batch process, queue BATCHQ1 will be stopped $SYSTEM@QUEBID> WRITE SYS$OUTPUT - _$ F$message(%x'F$GETQUI("DISPLAY_ENTRY","CONDITION_VECTOR","30")') %SYSTEM-E-BADRAD, bad RAD specified |
When you change the RAD value on a batch queue, the jobs currently in the batch queue are not dynamically updated with the new RAD value specified on the queue.
Any executing jobs will complete processing using the original RAD
value. Jobs in the pending, holding, or timed execution state will
retain the old RAD value on the job; however, when such a job becomes
executable, the job is updated with the new RAD value and runs on the
RAD specified on the queue.
4.7 DECram Support for Galaxy
DECram is a disk device driver that enables system managers to create logical disks in memory to improve I/O performance. Data on an in-memory DECram disk can be accessed at a faster rate than data on hardware disks. DECram Version 3.1 runs on OpenVMS Alpha systems, including Galaxy configurations, running OpenVMS Alpha Version 7.2--1H1 or higher.
DECram Version 3.1 takes advantage of the advanced capabilities of the newest Alpha systems, such as Galaxy shared memory. Additionally, the change to I/O post processing described in Section 4.8 is a significant performance enhancement. Version 3.1 allocates memory across Resource Affinity Domains (RADs). It moves the virtual device addressing from the S1 address space into the S2 address space, allowing for the creation and addressing of devices larger than 2 gigabytes (GBs).
You can easily upgrade an Alpha-based system to DECram Version 3.1. With Version 3.1, you can either use the improved new DECram command interface or continue using the familiar SYSMAN commands for creating, initializing, and mounting DECram disks.
DECram Version 3.1 disks are created and formatted at the DECRAM> prompt and are initialized with the DCL INITIALIZE (INIT) command. If you are configuring DECram Version 3.1 on OpenVMS Alpha, you can generate a DECram startup procedure to set up the disk and copy any required files to it. Usually, this procedure is called from the system startup procedure SYS$MANAGER:SYSTARTUP_VMS.COM.
A DECram disk must have 516 bytes of free page list per block (512 bytes) of disk space allocated. |
DECram Version 3.1 and supporting documentation are included on the
Software Products Library and the Online Documentation Library CD-ROMs.
4.8 DECram, Mailbox Driver, and Volume Shadowing Performance Gains
DECram for OpenVMS Version 3.1 and Version 2.5, the Mailbox driver, and Volume Shadowing for OpenVMS have been modified to improve performance on multiprocessor systems. A change has been made in OpenVMS Alpha Version 7.3--1 that enables device drivers to specify that I/O post processing be done on the local CPU. DECram, Mailbox driver, and Volume Shadowing have all been modified to use this performance enhancement. The performance gains are greatest for DECram and can be significant for Mailbox driver. For configurations that use Volume Shadowing, if the primary CPU is saturated, such systems will derive modest benefits.
Prior to the introduction of this capability, two routines were called to complete an input-output (I/O) operation, IOC$REQCOM and IOC$ALTREQCOM. These routines cause non-fast-path I/O completion to be performed by the primary CPU, thus potentially degrading performance.
As of Version 7.3--1, a device driver writer can improve performance by setting a bit in the unit control block (UCB) initialization routine to indicate that I/O post processing should be done on the local CPU. This bit is in the UCB$L_STS field and is called UCB$[V,M]_IOPOST_LOCAL. It causes IOC$REQCOM and IOC$ALTREQCOM to queue the IRP to the CPU-specific post processing queue and do a SOFTINT #IPL$_IOPOST, if necessary.
Because of ordering requirements, network drivers should not utilize the IOPOST_LOCAL feature. |
Prior to OpenVMS Alpha Version 7.3--1, all hardware interrupts took
place on the primary CPU. Interrupts from Fast Path enabled devices
would have to be redirected from the primary CPU to a
"preferred" CPU. This redirection still involved the primary
CPU, and also incurred interprocessor overhead. Furthermore, the
algorithm for assigning preferred CPUs to ports did not take into
account the performance implications of Resource Affinity Domains
(RADs).
4.9.1 Distributed Interrupts
Starting with OpenVMS Alpha Version 7.3--1, hardware interrupts for a Fast Path port can be directed to an interrupt CPU, which need not be the primary CPU. Fast Path performance is greatly enhanced when the preferred CPU for a Fast Path port is also the interrupt CPU for that port. In this case, interrupts go directly to the interrupt CPU, eliminating all I/O processing on the primary CPU. This major Fast Path performance enhancement is known as distributed interrupts.
You can use SHOW DEVICES/FULL to display the interrupt CPU ID of a Fast Path port. (If a system does not support distributed interrupts, the interrupt CPU is the primary CPU and does not appear in the SHOW DEVICES display.)
Distributed interrupts are supported on AlphaServer DS20, ES40/45, and GS series systems on Fibre Channel, CI, and some SCSI ports. |
The Fast Path code now takes into account the performance implications of the port's Resource Affinity Domain (RAD) in assigning preferred CPUs and interrupt CPUs. Systems without RADs are treated as systems consisting of one RAD. In some systems that support distributed interrupts, not all CPUs can handle them. An interrupt CPU must either be the primary CPU or must be in the RAD of the port.
Although the Fast Path code attempts to take all this into consideration in assigning preferred CPUs, you can use the SET DEVICE/PREFERRED_CPUS command to limit activity to a subset of available CPUs. In a Galaxy configuration, its choice can be overridden by the movement of CPUs in and out of an instance. The default value for PREFERRED_CPUS is NOPREFERRED_CPU, which gives the Fast Path code the greatest latitude in optimizing for performance.
For More Information On | See |
---|---|
Fast Path | VMS I/O User's Reference Volume |
RADs | OpenVMS Alpha Partitioning and Galaxy Guide |
Fibre Channel, SCSI, and CI configurations | Guidelines for OpenVMS Cluster Configurations |
Previous | Next | Contents | Index |
privacy and legal statement | ||
6657PRO_002.HTML |