High Availability Storage Options and Their Impact on Performance

ManualsBrandsHP ManualsSoftwareMPE/iX 7.0 Operating System

Another (Off) White Paper on High Availability Storage Options

and Their Impact on Performance on the HP e3000

By Walter McCullough

Version Date January 20, 2004

This is an update to a white paper I wrote back in 1997 when we were receiving a number of questions

regarding the introduction of storage arrays and their use on MPE systems. It seems we continue to have

some confusion regarding this subject despite the passage of time and improvement in the storage

technology.

My goal for this (short) paper is to lay out a somewhat high level overview of what makes MPE different

from other operating systems and importantly why MPE goes to disk to satisfying an I/O request. Then I

would like to describe the available HA storage solutions for your HP e3000 and just a little behind the

curtain view on how that new technology works with MPE.

You can view the older paper at:

http://jazz.external.hp.com/mpeha/papers/off_white_paper.html

Architecture

The current (and last) incarnation of MPE and its lowest (machine dependent) layer was specifically

designed for the PA-RISC architecture. This thin layer allowed the MPE lab to create an operating system

that had very little shielding from the hardware layer. While the HP-UX approach was to create a (thicker)

layer which allowed for greater hardware independence, MPE's approach allowed operations to move

more expeditiously through the computer, thus giving it the ability to do more (and generate more I/O).

The disadvantages (for the HP e3000) of this thinner machine dependent hardware layer or shielding is

that MPE is limited to the PA-RISC architecture while HP-UX's thicker shielding allowed it much more

freedom and flexibility to move to newer faster computer architectures.

MPE I/O Behavior

First of all MPE/iX is different from HP-UX and Unix (like) systems on how it relates to file access and

how it utilizes the combination of disk storage and server memory. While Unix (like) systems are usually

referred to as using a paging system to access files and data to and from a swap area, the MPE operating

system uses the file's disk location as the file swap area and only needs to allocate process/user state

information/data which is "swapped" out to areas of the system volume set. This technique greatly

reduces the number of times that the operating system needs to go to disk to satisfy an application’s I/O

requirements.

Summary of content (9 pages)

PAGE 1
hp Another (Off) White Paper on High Availability Storage Options and Their Impact on Performance on the HP e3000 By Walter McCullough Version Date January 20, 2004 This is an update to a white paper I wrote back in 1997 when we were receiving a number of questions regarding the introduction of storage arrays and their use on MPE systems. It seems we continue to have some confusion regarding this subject despite the passage of time and improvement in the storage technology.
PAGE 2
To further reduce the need to go to disk, MPE employs a cache as part of its design. This design uses available main memory to keep active parts of the file accessible by an approach that is similar to direct memory access. As active applications require access to their data those requests are satisfied through MPE's memory cache and are handled by the MPE/iX memory manager facility. The next technique employed by MPE/iX to increase performance is that of reducing the need to post data to disk.
PAGE 3
Storage Performance Features One of the biggest features that all the venders tout as their answer to performance improvements is the use of large amounts of memory in the array used as a cache. Some even give the impression that this cache will improve performance over JBOD and that arrays are faster because of the cache. Not always so. The right side of the picture below shows an arrow that indicates the (relative) time it takes to do either a read or write to disk.
PAGE 4
MPE Performance Environment Lets spend a little time painting the picture of an environment where MPE was best designed to give optimum performance. We will break this part up into a number of categories: memory, connection technology and Ldev (spindle count/size). The order of the categories also indicates its impact on performance, where the fastest access is memory and the slowest is disk. We will then go through the list of available array products for the HP e3000.
PAGE 5
Ldev Count & Spindle Size The term Ldev in this paper referrers to an MPE configured disk or volume.
PAGE 6
MPE Mirrored Disk/iX Though mirrored disk is not an array it is the lowest level entry into high availability data protection. This technology is based on (SCSI) JBOD and relies on having a single piece of data (file) written to two different disks that are on two separate paths. This method protects the data from any single point of failure along the I/O path. The greatest advantage of this technology is its small cost to implement.
PAGE 7
Business Copy and Continuous Access are not supported and there are no plans to release these features for use on the HP e3000. Enterprise Virtual Array (EVA) Not supported for the HP e3000. Early indications are that it is much faster than the VA Family of storage and has many of the same features as the VA Family but is a higher storage capacity and connectivity solution. But is not supported on the HP e3000.
PAGE 8
The XP512/XP48 is no longer available from HP but may be still available in the used product market. HP StorageWorks Disk Array XP1024/XP128 The XP1024 (and smaller XP128) are HP's latest entry into 2Gbit per second fibre channel connectivity. The XP1024/XP128 again supports multi-host heterogeneous environments and along with HP's latest (last) released version of MPE/iX 7.5 on PCI based hardware support native fibre channel (no router) to the portfolio of HP FC switches and FC hardware.
PAGE 9
"Rules of Thumb" for Improving MPE/iX Disk Performance What I would like to do now is list some guide lines to help make the migration to an array a better experience. These are not hard rules and some flexibility is needed as you incorporate these "Rules of Thumb" into your environment. The most important thing is to help set realistic expectations as storage technology changes and improves: 1. MPE/iX prefers many small Ldevs over a few large Ldevs.