HP 3000 Manuals HP 3000 Manuals
HP-FL Disk Arrays: C2252B, C2252HA, C2254B, and C2254HA
by Bob Berliner
Computer Systems Division
Hewlett-Packard introduces support on MPE/iX Release 4.0 of four models
of HP-FL Disk Array devices.
The HPC2252B (2.7 GB), HPC2252HA (2.7 GB High Availability), HPC2254B
(5.4 GB), and HPC2254HA (5.4 GB High Availability) are high-speed,
high-capacity, high-density disk storage devices for use on mid-range and
high-end HP 3000 systems.
As shown in the following table, the improved storage density and lower
cost per byte of data stored permits consolidation of many previous
generation disks into only a few HP-FL disks, with dramatic savings in
overall cost and computer room space requirements. The table compares
the new C2254B Disk Array device with the C2204A disk, which until now,
has been the highest capacity HP-FL disk supported.
Storage Cost and Space Comparison
-------------------------------------------------------------------------------------
| | | | |
| Disk Subsystem | Using C2254B | Using C2204A | Savings |
| Description | | | |
| | | | |
-------------------------------------------------------------------------------------
| | | | |
| Number of Devices | 5 | 20 | 75% |
| | | | |
| Storage Capacity | 27 GB | 26.8 GB | +200 Mbytes |
| | | | |
| Num. of Cabinets | 1 | 2.5 | 60% |
| | | | |
| Cabinet Volume | 31.5 cu.ft. | 70 cu.ft. | 55% |
| | | | |
| Floor Space | 6 sq.ft. | 16 sq.ft. | 62% |
| | | | |
| Device Costs | $226.5 K | $345.K | 34% |
| | | | |
| Device Cost $/Mbytes | $8.39 | $12.89 | 34% |
| | | | |
-------------------------------------------------------------------------------------
DEVICE FEATURES AND BENEFITS
In addition to their great storage density and lower cost, the new disks
have innovative value-added features derived from their use of disk array
technology. This technology provides several key benefits in expanding
systems to very large disk storage capacities and in keeping systems
continuously available for use.
All of these devices use the HP HP-FL high-speed fiber-optic link for
connection to the host HP 3000 computer system, and all use a 5.25 inch
diameter, 1.36 GB capacity disk mechanism as the basic storage unit.
Using newly developed disk Array technology, called Data Striping, these
products combine multiple disk mechanism building blocks into one fast,
high-capacity disk array device.
The High Availability (HA) models are distinguished by their ability to
tolerate an outright failure of any single disk mechanism within the
device without losing any previously stored data and without stopping
their own or the host computer system's operation.
The following sections describe the significant features of the new
devices, explain the benefits these features provide, and summarize the
software features in the MPE/iX Release 4.0 that support the disk array
devices.
Fiber-Optic Interface
The fiber-optic interface used by these devices for connection to the
host system provides several significant benefits:
* High speed: Has a 5 MB per second burst data transfer rate.
* Long distance communication: Disk devices can be located up to
500 meters away from the SPU.
* Device cluster support: With the HP-FL disks, up to five disk
array devices can be placed in one 1.6-meter-high cabinet,
connected to the SPU with one fiber-optic cable.
* Thin, lightweight cable: Is flexible, easy to route, and easy to
install.
* Data integrity: All data transmission is protected by cyclic
redundancy checking codes, which detect errors, and automatic
retransmission of data, which corrects errors.
* Excellent noise immunity: Fiber-optic transmission is impervious
to interference from electromagnetic and radio frequency noise
sources.
Disk Array Technology: Modular Packaging
The disk array device package combines plug-in disk drive mechanisms with
a disk array base unit that contains the disk array controller,
fiber-optic interface, power supply, and cooling fans. The basic disk
drive mechanism is small (13.5 " x 4.5 " x 7 "), lightweight (11 pounds),
and of high capacity (1.36 GB). The modular package design provides these
benefits:
* Easy installation: The empty base unit is first installed in the
cabinet, and then the drive mechanisms are plugged in. One person
can handle the lightweight components.
* Incremental growth: Holds up to five disk array devices per
cabinet.
* High capacity, high density storage: Stores up to 27 GB, using
five C2254B or C2254HA devices in a single 1.6-meter high cabinet
whose footprint is 24 inches by 36 inches. This yields a storage
density of 865 Mbytes per cubic foot, including cabinetry with
power distribution, and gives a floorspace usage of about 4.5 GB
per square foot.
* Ease and speed of repair: Replacement of a failed drive mechanism
is simple and quick. Without removing power from the device, the
failed mechanism can be unlocked and unplugged, and a replacement
mechanism inserted, locked, and spun-up in about a minute. The HA
models incorporate Data Recovery and Data Rebuilding features that
prevent having to suspend access to the device while data are
regenerated onto the replacement drive mechanism. With HA models,
no reload of data from backup storage is needed to recover from
any single failed drive mechanism. (If two or more drive
mechanisms within a device fail at or near the same time, then a
reload from backup storage will be needed, or a mirrored disk
configuration can be used to avoid a reload.)
Disk Array Technology: Data Striping
Data striping is a technique used by the disk array controller to operate
all the disk mechanisms within the device in parallel in a way that
boosts data transfer rate to and from the device. When writing data to
the device, the array controller spreads the data across the drive
mechanisms in a round-robin fashion; it reverses the process when reading
data back from the device. As implemented in the HP-FL disk arrays, data
striping benefits include:
* High data transfer rate: The C2252 model disks, using two drive
mechanisms per device (two-way striping), have a basic burst data
transfer rate twice that of a single disk. The C2254 disks, with
four-way striping across four drive mechanisms, have four times
the burst data transfer rate of a single disk. These high
data transfer rates can benefit those applications that
characteristically read and write rather large (several tens of
kilobytes) blocks of data in I/O operations.
* Basis for High Availability option: The data striping feature
provides the base on which the disk array controller implements
the HA feature, described below.
* Optimal striping performance, tolerant of disk speed variation:
For the fastest possible data transfers, the disk array controller
operates the drive mechanisms so that they spin synchronously at
identical rates and in step with one another. Yet the design is
able to tolerate a drifting apart of the disks' rotations. Should
drift occur, the devices continue to operate correctly, gracefully
decreasing their transfer rates to accommodate the drift.
Disk Array Technology: High Availability
In the HA models, C2252HA and C2254HA, the combination of modular
packaging and data striping disk array technology provides for an
outstanding high availability feature.
The HA models each add one extra drive mechanism to the basic striped set
of drive mechanisms, giving a total of three drive mechanisms in the
C2252HA and five mechanisms in the C2254HA.
The disk array controller uses the extra disk to record check data that
it derives from the application data on every write operation. The check
data are sufficient to reproduce the data originally written to the
striped disks in the event that any one of the drive mechanisms within a
device should fail.
In case of any single drive mechanism failure, the disk array controller
automatically switches the device into Data Recovery Mode. In this mode
the controller uses the check data together with the data from the
non-failed drive mechanisms to satisfy read requests from the computer
system. New data to be written simply bypass the failed drive mechanism,
and are stored on the combination of the non-failed drive mechanisms and
the check disk. (If it should be the check disk that failed, then the
controller simply uses the original, non-failed striped data disks to
store and retrieve data.)
Outstanding benefits of the HA feature are:
* High system uptime: In the event of a single mechanism failure,
both the disk device and the computer continue to operate
correctly. There is no suspension of system operation, and no
data loss.
* High performance: Entry into Data Recovery Mode is automatic in
the device controller, and there is no performance degradation
while operating in this mode. The device continues to operate at
its full speed.
* Graceful recovery: When a drive mechanism fails, it can quickly
and easily be replaced without powering-off the disk and without
stopping the system. After mechanism replacement, the device's
Data Rebuilding function reconstructs the data that belong on that
mechanism. During the rebuilding operation, Data Recovery Mode
continues to operate so that the host system can proceed with its
disk operations while, simultaneously, the disk array controller
performs data reconstruction. The rebuilding procedure uses the
HA mode check data to rebuild the old data that resided on the
failed disk, and it automatically takes care of merging in any new
data coming from the host computer as it continues to access the
device during the rebuilding. Note that data rebuilding is not
initiated automatically by the device. The FLEXDIAG disk
diagnostic program must be used to initiate data rebuilding on the
disk array device.
* Excellent price/performance: The HA mode devices provide a level
of protection against system stoppages and loss of disk data that
approaches that of Mirrored Disks, yet costs only incrementally
more than the cost of the basic disk device, rather than twice as
much.
High Availability Mode: Recovery from Disk Failure
In HA configurations, the combination of modular packaging and high
availability hardware and software design present in these devices
provides for rapid, convenient, and low-impact recovery from single-disk
failure.
Recovery is rapid and convenient because of the extraordinary ease and
speed with which a failed drive mechanism may be replaced with a new
drive mechanism. If a spare drive mechanism is available on-site, the
replacement can be made in only a few minutes. In addition, drive
mechanisms can be replaced hot. It is not necessary to turn device power
off, nor to suspend system operation, while removing a failed drive
mechanism and inserting a replacement mechanism.
Recovery is low-impact because the recovery process does not suspend
system operation nor does it inhibit the system's access to the device's
data. Two features contribute to the low-impact that recovery has on
system operation.
First, the ability of the device to operate in Data Recovery Mode means
that there is no system emergency to deal with. From the time of
mechanism failure up to the time of completion of replacement and data
rebuilding, the Data Recovery Mode keeps the device and system operating
normally, without interruption of service. Thus, from the viewpoint of
system uptime, there is no immediate requirement to replace a failed
mechanism. We stress, however, that to maintain continued protection
against data loss and system downtime in the event of a second mechanism
failure within the same device, it is important that failed drive
mechanisms be replaced without undue delay. This is because the device,
with one check disk, can not remain in operation if a second mechanism
fails before the first failed mechanism's repair is completed.
Second, the disk array controller has a self-contained Data Rebuild
function that, once a replacement drive mechanism has been installed and
the host system has commanded the device to begin rebuilding data,
automatically and autonomously reconstructs all of the data that used to
reside on the failed drive mechanism and writes it onto the replacement
drive mechanism. Data reconstruction takes place entirely within the
disk array device. The host system itself does not spend any of its own
processing time or memory resources on the task of data rebuilding.
Instead, the host system can continue to do its own work, including
reading and writing data on the disk device being rebuilt. The
rebuilding process in the disk controller is smart. It understands how
to handle new data coming from the host computer during rebuilding, so
that old reconstructed data are discarded and replaced by newly written
data.
During the data rebuilding process, a small decrease in disk performance
as seen by the host system is inevitable and does occur. After drive
mechanism failure and before rebuilding starts, while the disk is
operating in a pure Data Recovery Mode, there is no performance
degradation in salvaging disk data from the check data. But once
rebuilding commences, the disk device must divide up its time between
responding to ongoing host system disk I/O requests and rebuilding data
onto the replacement drive mechanism. Some degradation in system
performance is necessary in order to ensure that the data rebuilding
operation steadily makes progress. It is important that the rebuilding
be completed in a reasonable time period, and not be excessively delayed
by giving too much priority to satisfying ongoing system I/O requests,
because the HA disk device is not in a failsafe condition until
rebuilding is finished. By a failsafe condition, we mean the normal
high-availability condition in which a drive mechanism failure can be
handled gracefully, without stopping device operation.
A C2254HA device, at 5.4 GB storage capacity, can perform drive mechanism
rebuilding in about 20 minutes, so if the system is idle or lightly
loaded, and is not placing much demand on the disk for service, the
rebuilding operation completes and the normal failsafe condition is
achieved in about that time. Rebuilding takes longer if the system is
operating under a heavy workload that generates a high I/O rate to the
disk array device being rebuilt.
SOFTWARE FEATURES
MPE/iX 4.0 system software supports both the two-way striped (C2252) and
four-way striped (C2254) HP-FL Disk Array devices in both the Basic (B)
and High Availability (HA) configurations. Software features pertaining
to these devices include the following:
* Software supports attachment and operation of these devices on
HP-PA HP 3000 systems using either of the available HP-FL
Fiber-Optic Device Adapter host system interfaces. The older
HP27115A interface for CIO (Computer I/O) - based systems and the
new HP28616A interface for NIO (Native I/O) - based systems are
both supported for use with the new HP-FL Disk Array devices.
* Software permits and supports mixed configurations of both new
HP-FL Disk Array devices and prior HP-FL disk devices in a single
cluster, served by either of the HP-FL Device Adapters, within the
applicable hardware interconnection limitations dictated by the
device packaging and cabling specifications.
* Software supports configuration and operation of the supported
models of HP-FL Disk Arrays as either System Volumes or Private
Volumes.
_________________________________________________________________
NOTE Configuration limitation: In order to use HP-FL Disk Array
device as the System Master Volume, LDEV #1, on MPE/iX
Release 4.0, the device MUST be in either two-way striped
(C2252B) or two-way striped HA (C2252HA) configuration.
_________________________________________________________________
* HP-FL Disk Arrays are supported by the MPE/iX Mirrored Disk and
SPU Switchover software products.
* SYSGEN, the System Generation software mechanism, supports
configuration of these devices using their respective HP Product
Numbers: HPC2252B, HPC2252HA, HPC2254B, and HPC2254HA.
* The SHERLOCK Diagnostic System software and the FLEXDIAG HP-FL
disk family diagnostic program have been enhanced to support the
new HP-FL Disk Arrays. The FLEXDIAG program, in addition to its
diagnostic functions, supports the Data Rebuilding operation used
for recovery of the HA model disks.
* MPE/iX software provides support for the management of HA HP-FL
Disk Arrays. The software detects and reports to the System
Console those drive mechanism failures in the HA models that cause
the disk array device to commence Data Recovery Mode operation,
and that require failed drive mechanism replacement together with
a Data Rebuild operation. Detection and reporting occur promptly
at the time of failure, so that repair and rebuilding can be
initiated as quickly as desired.
* Software keeps records of drive mechanism failures, as well as all
other pertinent device error conditions, in the system's
Diagnostic Error Log file, where they are accessible for review by
system managers and for automatic analysis by HP Predictive
Support software. For system environments in which no operators
are present, the Diagnostic Error Log and Predictive Support
software provide automatic detection of drive mechanism failures
and automatic notification of the problem to HP Response Center
personnel.
_________________________________________________________________
NOTE HP Predictive Support software is bundled together with the
purchase of an HP Hardware Support contract.
_________________________________________________________________
ADDITIONAL INFORMATION
Further information about these devices is provided in the following
Hewlett-Packard publications, available from the Literature Distribution
Center:
* HP-FL Multiuser Disk Arrays data sheet (5091-2856E)
* HP Disk Arrays for Multiuser Computer Systems flier (5091-2858E)
* Disk Array Primer (5091-1396E)