HP OpenVMS Systems Documentation

This chapter describes the virtualization features of the OpenVMS operating system.

3.1 OpenVMS as a Guest Operating System on Integrity VM

OpenVMS for Integrity servers Version 8.4 is supported as a guest operating system on HP Integrity Virtual Machines (Integrity VM). Integrity VM is a soft partitioning and virtualization technology within the HP Virtual Server Environment, which enables you to create multiple virtual servers or machines with shared resourcing within a single HP Integrity server or nPartition.

Each virtual machine hosts its own "Guest" operating system instance, applications, and users. Integrity VM runs on any Intel VT-i enabled HP Integrity servers including blades. On HP Integrity servers, the Integrity VM Host runs under HP-UX, while OpenVMS can run as a guest.

3.1.1 Licensing Requirements

OpenVMS Guest operating system on Integrity VM is supported on VT-i (Intel Virtualization Technology for the Intel Itanium architecture) enabled Intel Itanium processors. Currently Intel Itanium 9000 and 9100 series support VT-i.

For more information on Integrity VM, see: http://h71028.www7.hp.com/enterprise/us/en/os/hpux11i-partitioning-integrity-vm.html

3.1.3 Installing OpenVMS as a Guest Operating System

To install OpenVMS as a guest operating system, see Chapter 3 in HP OpenVMS Version 8.4 for Integrity Servers Upgrade and Installation Manual.

3.1.4 OpenVMS Features as a Guest Operating System

• The OpenVMS guest OS is SMP enabled and supports up to 64 GB physical memory.
• The OpenVMS guests support virtualized disk drives and network interfaces provided by Integrity VM. Integrity VM presents disks and logical volumes as SCSI disks (DK devices on OpenVMS guests) and virtual network interfaces as Intel Gigabit Cards (EI devices on OpenVMS guests) regardless of the physical network card or mass storage connection for the Host system.
• Limited support for online migration - supports only the stand-alone guest configurations.
• Supports Accelerated Virtual IO (AVIO) LAN and SCSI drivers.
• Supports management and monitoring of OpenVMS guest operating system using the VSE suite of products.
• OpenVMS guest systems are cluster enabled and supports LAN and cluster over IP.

For more information, see the HP OpenVMS Version 8.4 Release Notes.

3.2 ID-VSE for OpenVMS

The HP Insight Dynamics - Virtual Server Environment (ID-VSE) is an integrated suite of multi-platform products that helps you to continuously analyze, and optimize physical and virtual server resources. It helps you to reduce the cost associated with capacity and energy planning, provisioning, upgrades, and making changes in your data center.

ID-VSE integrates with HP Systems Insight Manager (HP SIM) running on a central management station (CMS), and manages one or more managed nodes in your network.

The following suite of ID-VSE products are supported on OpenVMS Version 8.4:

HP Virtualization Manager

The Virtualization Manager software provides a framework for visualizing your virtual server environment (VSE) at different levels of detail. All the systems and workloads are displayed in a graphical view. The hierarchical relationships between systems and their current utilization are displayed on a single screen. It also allows you to access additional VSE Management software for management and configuration of systems and workloads. Virtualization Manager collects utilization data for processor, memory, network, and disk from OpenVMS managed nodes through the Utilization WBEM provider.

HP Capacity Advisor

The Capacity Advisor software provides capacity analysis and planning to help optimize the workloads across VSE for the highest utilization of server resources. It also provides scenario analysis to optimize the current server resources and plan for future workload expansion and server consolidation. Capacity Advisor collects utilization data for processor, memory, network, and disk from the OpenVMS managed nodes through the Utilization WBEM provider.

HP Global Workload Manager

HP Global Workload Manager (gWLM) is a multi-system, multi-OS workload manager that serves as an intelligent policy engine in the VSE software. It simplifies the deployment of automated workload management policies across multiple servers and provides centralized monitoring and improved server utilization to meet the service-level objectives.

On OpenVMS with Global Workload Manager, all the capabilities of iCAP, TiCAP can be automated based on the defined business policies. For example, if performance goals are not met, additional processors can be automatically turned on using TiCAP or usage rights can be dynamically moved from a partition.

Prerequisite

To use the Global Workload Manager, the gWLM agent must be running on the OpenVMS managed nodes.

This chapter describes new features relating to performance enhancements in this version of the HP OpenVMS operating system.

4.1 RAD Support (Integrity servers Only)

OpenVMS Version 8.4 has been enhanced to support RAD for cell-based Integrity server systems. This feature enables OpenVMS to utilize the advantages of cell-based systems configured with cell local memory (CLM). On systems with both CLM and interleaved memory (ILM) configured, OpenVMS allocates process memory from the CLM within a cell and schedules the process to run on a CPU within the same cell. The overall memory latency and bandwidth for the process is improved by reducing the frequency of a CPU in one cell referencing memory in another cell.

Prerequisite

To use the RAD support, CLM must be configured on the operating system using the Partition Manager software. The Partition Manager provides the system administrators with a graphical user interface (GUI) to configure and manage nPartitions on HP server systems. The Partition Manager is supported on HP-UX, Microsoft Windows, Red Hat Enterprise Linux, and SUSE Linux Enterprise Server. This software interacts with the user through a web browser running on a client system. The client system can be the same as the server system, or it can be a separate workstation, or a PC. Note that the Partition Manager does not run on OpenVMS. For more information and for software downloads, see:

Recommendation

OpenVMS recommends configuring systems with a combination of both CLM and ILM. Initially, configure 50% of the memory in each cell as CLM. For best performance, follow the hardware guidelines for configuring systems with combinations of CLM and ILM. For cell-based systems, the number of cells and the amount of ILM must be in power of 2.

By default, on a system configured with CLM, OpenVMS boots with RAD support turned on. RAD support can be turned off by setting the RAD_SUPPORT system parameter to 0. The recommended method of turning RAD support off is to configure all the memory on the system as ILM.

When ILM and CLM are both present on an Integrity server system, the ILM is seen as an additional RAD. Because all CPUs on the system have a similar average memory latency when accessing this memory, all CPUs are associated with this RAD. Note that there is no Alpha hardware that has both RADs and ILM, and thus this extra RAD never appears on Alpha.

For example, consider an rx7640 system with 2 cells, 16 GB of memory per cell, and 8 cores per cell. If you configure the system with 50% CLM per cell, this system boots and OpenVMS configures the system with 3 RADs. RAD 0 contains the CLM and cores from the first cell. RAD 1 contains the CLM and cores from the second cell. A third RAD (RAD 2) contains the ILM and all cores.

Although there are 3 RADs, processes are assigned only to the first 2 RADs as home RADs. The RAD in which a core first appears is the RAD with the best memory access for the core. The $GETSYI system service and F$GETSYI lexical function enumerate this for the RAD_CPUS item code.

During system boot, the operating system assigns a base RAD from which shared and operating system data is allocated. The base RAD that the operating system assigns is the RAD with ILM because all CPUs have similar access to this memory. Non-paged pool is allocated from the base RAD. By default, per-RAD non-paged pool is turned off.

By default, global page faults are now satisfied with pages from the base RAD compared to the RAD of the CPU where the fault occurred. Because global sections can be accessed by many processes running on all RADs, pages are allocated from the RAD with ILM.

For more information about RAD support, see the HP OpenVMS Alpha Partitioning and Galaxy Guide.

4.1.1 Page Zeroing for RAD based Systems (Integrity servers and Alpha)

Within the idle loop, CPUs can "zero" the deleted pages of memory to satisfy the future demand-zero page faults. The ZERO_LIST_HI system parameter indicates the maximum number of zeroed pages that the operating system must keep zeroed. For systems with multiple RADs, ZERO_LIST_HI specifies the maximum number of zeroed pages per RAD.

4.1.2 SYS$EXAMPLES:RAD.COM (Integrity servers and Alpha)

The SYS$EXAMPLES:RAD.COM command procedure provides an example for using the RAD related F$GETSYI item codes, RAD_CPUS and RAD_MEMORY. This procedure has been updated to produce a more concise view of the RAD configuration for Integrity servers and Alpha systems.

$ @SYS$EXAMPLES:RAD 
 
Node: SYS123 Version: V8.4  System: HP rx7640  (1.60GHz/12.0MB) 
 
RAD   Memory (GB)   CPUs 
===   ===========   =============== 
  0        3.99     0-7 
  1        3.99     8-15 
  2        7.99     0-15 

This procedure is run on an rx7640 system with 16 GB of memory, with each cell configured to have 50% CLM. A portion of the CLM from each cell and the ILM memory may be allocated for the console, and thus not available for use by the operating system. As a result, you do not see 4 GB for the first 2 RADs and 8 GB for the RAD with ILM.

4.1.3 RAD Memory Usage (Integrity servers and Alpha)

To determine the memory usage per RAD, use the SDA command SHOW PFN/RAD. This command reports the number of free and zeroed pages per RAD. The CPU is extensively used when you execute the SHOW PFN/RAD command. Hence, use SHOW PFN/RAD only on occasions if you want to check the memory usage per RAD.

This chapter describes new features relating to disaster tolerance and clusters of the OpenVMS operating system.

5.1 Cluster over IP

OpenVMS Version 8.4 has been enhanced with the Cluster over IP feature. Cluster over IP provides the ability to form clusters beyond a single LAN or VLAN segment using the industry standard Internet Protocol. This feature provides improved disaster tolerance.

Cluster over IP enables you to:

• Form a cluster between nodes in data centers that are in different LAN or VLAN segment.
• Form geographically distributed disaster tolerant cluster with IP network.
• Improve total cost of ownership.

The cluster over IP feature includes:

• PEdriver to use UDP protocol in addition to IEEE 802.3 LAN for system communication services (SCS) packets.
• Reliable delivery of SCS packets by PEDRIVER using User Datagram Protocol (UDP).
• IP multicast and optional IP unicast to discover nodes in an IP only environment.
• Ability to load TCP/IP services during boot time to enable formation of cluster in IP only environment.

HP TCP/IP services for OpenVMS 5.7 is required to use the cluster over IP feature.

For more information, see the Guidelines for OpenVMS Cluster Configurations and HP OpenVMS Cluster Systems guides.

5.2 Volume Shadowing for OpenVMS Enhancements

OpenVMS Version 8.4 supports a maximum of six-member shadow set compared to the previous three-member shadow set. This is aimed at multi-site disaster tolerant configuration. With three member shadow set, a three site disaster tolerant configuration will have only one shadow member per site. In this case, during failure of two sites, the member left out in the surviving site becomes a single point of failure. With six-member shadow set support, you can have two members of a shadow set in each of the three sites providing high availability.

5.2.2 New DISMOUNT Keyword for HBMM

All the 12 write bitmaps are used by shadowing as multiuse bitmaps, thus removing the single point of failure of single minicopy master bitmaps. To invoke this feature, a new keyword is added to the SET SHADOW/POLICY command: DISMOUNT=n

where; n specifies the number of HBMM bitmaps to convert to multiuse bitmaps when a member is dismounted from a shadow set with the $DISMOUNT/POLICY=MINICOPY command.

5.2.3 Fast Minicopy and Minimerge

Shadowing has been enhanced to increase the performance of shadow minicopy and minimerge using "looking ahead" of the next bit that is set in the write bitmap.

The number of QIOs between SHADOW_SERVER and SYS$SHDRIVER is drastically reduced using this method, thus allowing minicopy and minimerge to complete faster.

5.2.4 New Qualifiers for SET SHADOW

• /DISABLE=SPLIT_READ_LBNS - disables the "split read lbn" behavior and as a result the reads are alternated between the source shadow set members having the same read_cost and device queue length.
• /ENABLE=SPLIT_READ_LBNS - logically divides the shadow set members having the same read cost into equal groups of logical block numbers (LBNs). When reads are performed to the virtual unit, they are read from the corresponding LBN group disk. This results in the maximum usage of the controller read-ahead cache.
• /STALL=WRITES[=nnn] - where nnn equals the number of seconds to stall the write. This qualifier will be useful if a user wants to stall the write operations for "nnn" seconds. If no value is specified for "nnn" seconds, the lock is released after SHADOW_MBR_TMO seconds. The default is SHADOW_MBR_TMO.
• /NOSTALL=WRITES[=nnn] - releases the write lock after "nnn" seconds so that the write operation continues on the shadow set.

Write Bitmaps (WBM) is a feature used by OpenVMS during minimerge and minicopy operations of Shadowing minimerge and minicopy. Information, about which blocks on a disk are written, is transmitted to other nodes within the cluster. The following updates have been made in this release.

Asynchronous SetBit Messages

There can be multiple master bitmap nodes for a shadow set. Currently, SetBit messages are sent to the multiple master bitmap nodes synchronously. Only when the response for the SetBit message is received from the first remote master bitmap node, is the message sent to the next master bitmap node. When done with all of the remote master bitmap nodes, the I/O is resumed. SetBit messages are now sent to all multiple master bitmap nodes asynchronously. I/O operation is resumed when the responses from all the master bitmap nodes are received. This reduces the stall time of the I/O operation by the write bitmap code.

Reduced SetBit Messages for Sequential I/O

If sequential writes occur to a disk, it results in sending Setbit messages that set sequential bits in the remote bitmap. The WBM code will now recognize where a number of prior bits in the bitmap have already been set. In this scenario, the WBM code will set additional bits so that if sequential writes should continue, fewer Setbit messages are required. Assuming the sequential I/O continues, the number of Setbit messages will be reduced by about a factor of 10 and thus improve the I/O rate for sequential writes.

This chapter describes the support added for the Storage devices and I/O controllers in this version of the OpenVMS operating system.

6.1 8 Gb Fibre Channel PCIe Adapter Support

Support for 1-port 8 Gb Fibre Channel Adapter (AH400A) and 2-port 8 Gb Fibre Channel Adapter (AH401A) PCI-Express (PCIe) has been added. For more information, see: http://www.hp.com/products1/serverconnectivity/storagesnf2/8gbfibre/index.html

OpenVMS also supports 2-port 8 Gb Fibre Channel Mezzanine Card for HP BladeSystem c-Class (product number 451871-B21).

6.2 PCI Sound Card HP AD317A PCI Support

Limited support for HP AD317A PCI sound card on Integrity servers has been added.

6.3 Storage Devices and I/O Controllers Supported After the Initial OpenVMS V8.3--1H1 Release

The following list of storage devices and I/O controllers were released after OpenVMS V8.3--1H1 was shipped. V8.3--1H1 support was introduced through patch kits; all of these devices will be qualified and supported by OpenVMS Version 8.4.

• HP Smart Array P700m
• HP Smart Array P411
• HP StorageWorks MDS600
• HP StorageWorks D2D Backup Systems
• HP StorageWorks Ultrium Tape blades
• HP StorageWorks Secure Key Manager (SKM)
• MSL LTO4 Encryption Kit
• HP StorageWorks MSA2000fc Modular Smart Array (FC)
• HP StorageWorks MSA2000sa Modular Smart Array (SAS)
• 2xGigE LAN (Intel 82575), 1x10/100/1000 Management LAN
• EVA4400
• EVA6400/8400
• P2000 G3