Document revision date: 19 July 1999

The following configuration guidelines apply to all DSSI clusters:

• Each DSSI interconnect can have up to eight nodes attached; four can be systems and the rest can be storage devices. Each of the following counts as a DSSI node:
  • DSSI adapter
  • Any member of the HSDxx family of DSSI and SCSI controllers
  • Any RF, TF, or EF integrated storage element (ISE)
  
  In some cases, physical cabling and termination limitations may restrict the number of systems to two or three that may be connected to a DSSI interconnect. For example:
  • Some variants of the DSSI adapter terminate the bus; for example, the N710. For this reason, only two DEC 4000 systems can be configured on a DSSI interconnect.
  • The size of a DEC or VAX 10000 system generally limits the number of systems that can be connected to a DSSI interconnect.
• Each DSSI adapter in a single system must be connected to a different DSSI bus.
• Configure VAX 6000, VAX 7000, and VAX 10000 systems with KFMSA adapters.
• Configure DEC 7000 and DEC 10000 systems with KFMSB adapters.
• Configure PCI-based AlphaServer systems with KFPSA adapters. EISA adapters (KFESA/KFESB) adapters can also be configured on most AlphaServer systems, but use of KFPSA is recommended whenever possible.
• Dual porting of devices between HSD controllers is supported as long as they are connected to the same or separate DSSI interconnects. Dual porting of devices between HSD controllers and local controllers is not supported.
• All systems connected to the same DSSI bus must have a common power or ground.

Reference: For more information about DSSI, see the DSSI OpenVMS Cluster Installation and Troubleshooting Manual.

4.11 Ethernet, Fast Ethernet, and Gigabit Ethernet Interconnects

The Ethernet, Fast Ethernet, and Gigabit Ethernet interconnects provide single-path connections within an OpenVMS Cluster system and a local area network (LAN). Multiple Ethernet adapters or multichannel Ethernet adapters can be used to provide multiple paths.

Ethernet (including Fast Ethernet and Gigabit Ethernet) and FDDI are LAN-based interconnects. See Section 4.12 for information about FDDI and for general LAN-based cluster guidelines.

4.11.1 Advantages

The Ethernet, Fast Ethernet, and Gigabit Ethernet interconnects offer the following advantages:

• Typically, depending on implementation, the lowest cost of all OpenVMS Cluster interconnects.
• Support for 96 nodes on a single interconnect.
• Enable extended physical distribution of nodes.

The Ethernet technology offers a range of baseband transmission speeds:

• 10 Mb/s for standard Ethernet
• 100 Mb/s second for Fast Ethernet
• 1 Gb/s for Gigabit Ethernet

Ethernet adapters do not provide hardware assistance, so processor overhead is higher than for CI or DSSI.

Consider the capacity of the total network design when you configure an OpenVMS Cluster system with many Ethernet-connected nodes or when the Ethernet also supports a large number of PCs or printers. General network traffic on an Ethernet can reduce the throughput available for OpenVMS Cluster communication. Fast Ethernet and Gigabit Ethernet can significantly improve throughput. Multiple Ethernet adapters can be used to improve cluster performance by offloading general network traffic.

Reference: For extended LAN configuration guidelines, see Section 10.7.7.

4.11.3 Multiple Ethernet Load Balancing

If only Ethernet paths are available, the choice between which path the OpenVMS Cluster software uses is based on latency (computed network delay). If delays are equal, either path can be used. Otherwise, the OpenVMS Cluster software chooses the channel with the least latency. The network delay across each segment is calculated approximately every 3 seconds. Traffic is then balanced across all communication paths between local and remote adapters.

4.11.4 Supported Adapters and Buses

The following are Ethernet adapters and the internal bus that each supports:

• DEFTA-xx (TURBOchannel)
• DGLTA (TURBOchannel)
• DE2xx (ISA)
• DW110 (ISA)
• DE42x (EISA)
• DE435 (PCI)
• TULIP (PCI)
• KZPSM (PCI)
• DE4xx (PCI)
• DE500-xx (PCI)
• DE600-xx (PCI)
• DE602-xx (PCI)
• DEGPA-xx (PCI)
• DGLPB (PCI)
• 3COM (PCMCIA)
• DEMNA (XMI)
• TGEC (embedded)
• COREIO (TURBOchannel)
• PMAD (TURBOchannel)
• DE422 (EISA)
• DEBNI (VAXBI)
• DEBNA (VAXBI)
• SGEC (embedded)
• DESVA (embedded)
• DESQA (Q-bus)
• DELQA (Q-bus)

Reference: For complete information about each adapter's features and order numbers, see the DIGITAL Systems and Options Catalog.

To access the most recent DIGITAL Systems and Options Catalog on the World Wide Web, use the following URL:

http://www.digital.com:80/info/soc/

4.11.5 Ethernet-to-FDDI Bridges

You can use transparent Ethernet-to-FDDI translating bridges to provide an interconnect between a 10-Mb/s Ethernet segment and a 100-Mb/s FDDI ring. These Ethernet-to-FDDI bridges are also called "10/100" bridges. They perform high-speed translation of network data packets between the FDDI and Ethernet frame formats.

Reference: See Figure 10-21 for an example of these bridges.

4.11.6 Configuration Guidelines for Gigabit Ethernet Clusters

Use the following guidelines when configuring systems in a Gigabit Ethernet cluster:

• Two-node Gigabit Ethernet clusters do not require a switch. They can be connected point to point.
• Most Gigabit Ethernet switches can be configured with Gigabit Ethernet or a combination of Gigabit Ethernet and Fast Ethernet (100 Mb/s).
• Each node can have a single connection to the switch or can be configured with multiple paths, increasing availability. The AlphaServer models GS140, 8400, 8200, and 4x00 support up to four adapters each. The AlphaServer models 1200 and 800 support up to two adapters each.
• A limited number of functions are not yet supported in an OpenVMS Cluster system:
  • Jumbo frames
    The frame size supported for cluster communications is the standard 1518-byte maximum Ethernet frame size.
  • Optimal path selection
    Because optimal path selection is not provided in this release, you cannot rely on the cluster software for this function.
  • Satellite booting with the DEGPA as the boot device
    Although the DEGPA cannot be used as the boot device, satellites can be booted over standard 10/100 Ethernet network adapters configured on a Gigabit switch.

FDDI is an ANSI standard LAN interconnect that uses fiber-optic cable. FDDI supports OpenVMS Cluster functionality over greater distances than other interconnects. FDDI also augments the Ethernet by providing a high-speed interconnect for multiple Ethernet segments in a single OpenVMS Cluster system.

4.12.1 Advantages

FDDI offers the following advantages:

• Supports 96 nodes on a single interconnect.
• Combines high throughput and longest distance between nodes.
• Supports a variety of topologies.

The FDDI standards define the following two types of nodes:

• Stations --- The ANSI standard single-attachment station (SAS) and dual-attachment station (DAS) can be used as an interconnect to the FDDI ring. It is advisable to attach stations to wiring concentrators and to attach the wiring concentrators to the dual FDDI ring, making the ring more stable.
• Wiring concentrator --- The wiring concentrator (CON) provides a connection for multiple SASs or CONs to the FDDI ring. A DECconcentrator 500 is an example of this device.

FDDI limits the total fiber path to 200 km (125 miles). The maximum distance between adjacent FDDI devices is 40 km with single-mode fiber and 2 km with multimode fiber. In order to control communication delay, however, it is advisable to limit the maximum distance between any two OpenVMS Cluster nodes on an FDDI ring to 40 km.

4.12.4 Throughput

The maximum throughput of the FDDI interconnect (100 Mb/s) is 10 times higher than that of Ethernet.

In addition, FDDI supports transfers using large packets (up to 4468 bytes). Only FDDI nodes connected exclusively by FDDI can make use of large packets.

Because FDDI adapters do not provide processing assistance for OpenVMS Cluster protocols, more processing power is required than for CI or DSSI.

4.12.5 Supported Adapters and Bus Types

Following is a list of FDDI adapters and the buses they support:

• DEFPA (PCI)
• DEFPZ (integral)
• DEMFA (XMI)
• DEFAA (Futurebus+)
• DEFTA (TURBOchannel)
• DEFZA (TURBOchannel)
• DEFEA (EISA)
• DEFQA (Q-bus)

Reference: For complete information about each adapter's features and order numbers, see the DIGITAL Systems and Options Catalog.

To access the most recent DIGITAL Systems and Options Catalog on the World Wide Web, use the following URL:

http://www.digital.com:80/info/soc/

4.12.6 Configuration Guidelines for FDDI-Based Clusters

FDDI-based configurations use FDDI for node-to-node communication. The following general guidelines apply to FDDI configurations:

• The HS1xx and HS2xx family of storage servers provide FDDI-based storage access to OpenVMS Cluster nodes.
• Alpha and VAX systems can be configured with any mix of FDDI adapters.
• All FDDI paths used for OpenVMS Cluster communication must operate with a minimum of 10 Mb/s throughput and low latency. You must use translating bridges when connecting nodes on an Ethernet to those on an FDDI. LAN segments can be bridged to form an extended LAN.
• Multiple, distinct OpenVMS Cluster systems can be configured onto a single, extended LAN. OpenVMS Cluster software performs cluster membership validation to ensure that systems join the correct LAN OpenVMS cluster.
  Reference: For information about multiple LANs, see Section 8.6. For extended LAN (ELAN) configuration guidelines, see Table 10-3.

Because FDDI is ideal for spanning great distances, you may want to supplement its high throughput with high availability by ensuring that critical nodes are connected to multiple FDDI rings. Physical separation of the two FDDI paths helps ensure that the configuration is disaster tolerant.

4.12.8 Multiple FDDI Load Balancing

If only FDDI paths are available, the OpenVMS Cluster software bases the choice between which path to use on latency (computed network delay). If delays are equal, either path can be used. Otherwise, OpenVMS Cluster software chooses the channel with the least latency. The network delay across each segment is calculated approximately every 3 seconds. Traffic is balanced across all communication paths between local and remote adapters.

This chapter describes how to design a storage subsystem. The design process involves the following steps:

1. Understanding storage product choices
2. Estimating storage capacity requirements
3. Choosing disk performance optimizers
4. Determining disk availability requirements
5. Understanding advantages and tradeoffs for:
   • CI based storage
   • DSSI based storage
   • SCSI based storage
   • Fibre Channel based storage
   • Host-based storage
   • LAN InfoServer

The rest of this chapter contains sections that explain these steps in detail.

5.1 Understanding Storage Product Choices

In an OpenVMS CLuster, storage choices include the StorageWorks family of products, a modular storage expansion system based on the Small Computer Systems Interface (SCSI--2) standard. StorageWorks helps you configure complex storage subsystems by choosing from the following modular elements:

• Storage devices such as disks, tapes, CD-ROMs, and solid-state disks
• Array controllers
• Power supplies
• Packaging
• Interconnects
• Software

Consider the following criteria when choosing storage devices:

• Supported interconnects
• Capacity
• I/O rate
• Floor space
• Purchase, service, and maintenance cost

One of the benefits of OpenVMS Cluster systems is that you can connect storage devices directly to OpenVMS Cluster interconnects to give member systems access to storage.

In an OpenVMS Cluster system, the following storage devices and adapters can be connected to OpenVMS Cluster interconnects:

• HSJ and HSC controllers (on the CI)
• HSD controllers and ISEs (on the DSSI)
• HSZ and RZ series (on the SCSI)
• HSG controllers (on the Fibre Channel)
• Local system adapters

Table 5-1 lists the kinds of storage devices that you can attach to specific interconnects.

Table 5-1 Interconnects and Corresponding Storage Devices

Storage Interconnect	Storage Devices
CI	HSJ and HSC controllers and SCSI storage
DSSI	HSD controllers, ISEs, and SCSI storage
SCSI	HSZ controllers and SCSI storage
Fibre Channel	HSG controllers and SCSI storage
FDDI	HS xxx controllers and SCSI storage

5.1.3 How Floor Space Affects Storage Choices

• Choose disk storage arrays for high capacity with small footprint. Several storage devices come in stackable cabinets for labs with higher ceilings.
• Choose high-capacity disks over high-performance disks.
• Make it a practice to upgrade regularly to newer storage arrays or disks. As storage technology improves, storage devices are available at higher performance and capacity and reduced physical size. For example, replacing an HSC95 and SA800 with an HSJ40 and SW800 increases capacity and reduces floor-space consumption.
• Plan adequate floor space for power and cooling equipment.

Storage capacity is the amount of space needed on storage devices to hold system, application, and user files. Knowing your storage capacity can help you to determine the amount of storage needed for your OpenVMS Cluster configuration.

5.2.1 Estimating Disk Capacity Requirements

To estimate your online storage capacity requirements, add together the storage requirements for your OpenVMS Cluster system's software, as explained in Table 5-2.

Table 5-2 Estimating Disk Capacity Requirements

Software Component	Description
OpenVMS operating system	Estimate the number of blocks 1 required by the OpenVMS operating system. Reference: Your OpenVMS installation documentation and Software Product Description (SPD) contain this information.
Page, swap, and dump files	Use AUTOGEN to determine the amount of disk space required for page, swap, and dump files. Reference: The OpenVMS System Manager's Manual provides information about calculating and modifying these file sizes.
Site-specific utilities and data	Estimate the disk storage requirements for site-specific utilities, command procedures, online documents, and associated files.
Application programs	Estimate the space required for each application to be installed on your OpenVMS Cluster system, using information from the application suppliers. Reference: Consult the appropriate Software Product Description (SPD) to estimate the space required for normal operation of any layered product you need to use.
User-written programs	Estimate the space required for user-written programs and their associated databases.
Databases	Estimate the size of each database. This information should be available in the documentation pertaining to the application-specific database.
User data	Estimate user disk-space requirements according to these guidelines: Allocate from 10,000 to 100,000 blocks for each occasional user. An occasional user reads, writes, and deletes electronic mail; has few, if any, programs; and has little need to keep files for long periods. Allocate from 250,000 to 1,000,000 blocks for each moderate user. A moderate user uses the system extensively for electronic communications, keeps information on line, and has a few programs for private use. Allocate 1,000,000 to 3,000,000 blocks for each extensive user. An extensive user can require a significant amount of storage space for programs under development and data files, in addition to normal system use for electronic mail. This user may require several hundred thousand blocks of storage, depending on the number of projects and programs being developed and maintained.
Total requirements	The sum of the preceding estimates is the approximate amount of disk storage presently needed for your OpenVMS Cluster system configuration.

Before you finish determining your total disk capacity requirements, you may also want to consider future growth for online storage and for backup storage.

For example, at what rate are new files created in your OpenVMS Cluster system? By estimating this number and adding it to the total disk storage requirements that you calculated using Table 5-2, you can obtain a total that more accurately represents your current and future needs for online storage.

To determine backup storage requirements, consider how you deal with obsolete or archival data. In most storage subsystems, old files become unused while new files come into active use. Moving old files from online to backup storage on a regular basis frees online storage for new files and keeps online storage requirements under control.

Planning for adequate backup storage capacity can make archiving procedures more effective and reduce the capacity requirements for online storage.

5.3 Choosing Disk Performance Optimizers

Estimating your anticipated disk performance work load and analyzing the work load data can help you determine your disk performance requirements.

You can use the Monitor utility and DECamds to help you determine which performance optimizer best meets your application and business needs.

5.3.1 Performance Optimizers

Performance optimizers are software or hardware products that improve storage performance for applications and data. Table 5-3 explains how various performance optimizers work.

Table 5-3 Disk Performance Optimizers

Optimizer	Description
DECram for OpenVMS	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 disks are capable of being shadowed with Volume Shadowing for OpenVMS and of being served with the MSCP server. 1
Solid-state disks	In many systems, approximately 80% of the I/O requests can demand information from approximately 20% of the data stored on line. Solid-state devices can yield the rapid access needed for this subset of the data.
Disk striping	Disk striping (RAID level 0) lets applications access an array of disk drives in parallel for higher throughput. Disk striping works by grouping several disks into a "stripe set" and by dividing the application data into "chunks," which are spread equally across the disks in the stripe set in a round-robin fashion. By reducing access time, disk striping may improve performance, especially if the application: Performs large data transfers in parallel. Requires load balancing across drives. Two independent types of disk striping are available: Controller-based striping, in which HSJ and HSD controllers combine several disks into a single stripe set. This stripe set is presented to OpenVMS as a single volume. This type of disk striping is hardware based. Host-based striping, which creates stripe sets on an OpenVMS host. The OpenVMS software breaks up an I/O request into several simultaneous requests that it sends to the disks of the stripe set. This type of disk striping is software based. Note: You can use Volume Shadowing for OpenVMS software in combination with disk striping to make stripe set members redundant. You can shadow controller-based stripe sets or you can host-based disk stripe shadow sets.
Virtual I/O cache (VIOC)	OpenVMS offers host-based caching in the form of VIOC, a clusterwide, file-oriented disk cache. VIOC reduces I/O bottlenecks within OpenVMS Cluster systems by reducing the number of I/Os from the system to the disk subsystem.
Controllers with disk cache	Some storage technologies use memory to form disk caches. Accesses that can be satisfied from the cache can be done almost immediately and without any seek time or rotational latency. For these accesses, the two largest components of the I/O response time are eliminated. The HSC, HSJ, HSD, and HSZ controllers contain caches. Every RF and RZ disk has a disk cache as part of its embedded controller.

Reference: See Section 10.8 for more information about how these performance optimizers increase an OpenVMS Cluster's ability to scale I/Os.

	privacy and legal statement
6318PRO_004.HTML