Printer Friendly

Storage: the outer limits; Take a look at storage 2012.

The data storage industry has been based on a model of delivering ever-increasing device capacity with an ever-decreasing price per unit purchased. How many times have you heard someone say how cheap storage has become? Clearly, most of the dialogue centers around the price and capacity of storage but the third key element for data storage is less often discussed and is now beginning to elevate its level of importance: performance. The question of how long capacities can increase without corresponding increases in performance leads to a discussion of just exactly what are the real limits of storage. Are the real data storage limits technology limits or practical limits? From the latest technology roadmaps, both disk drive and tape cartridge capacities are growing faster than their corresponding data rates and are projected to do so through the year 2012.

Disk Limits

Disk roadmaps are growing at 40-60% annually and now indicate a clear path to 5 terabytes or more per drive by 2012 at the latest; however, disk drive performance capabilities have not kept pace with the growth in disk capacities. Disk drive performance is improving well below 10% annually. Raw disk drive performance is normally measured in total random I/Os per second and can surpass 100 I/Os per second as the average access time (average seek, latency and data transfer) is about 10ms per I/O on newer drives. Continual increases in capacity, without corresponding performance improvements at the drive level, create a performance imbalance that is defined by the ratio called Disk Access Density. Access Density is the ratio of performance, measured in I/Os per second, to the capacity of the drive, usually measured in gigabytes (Access Density = I/Os per second per gigabyte). The primary focus for disk manufacturers remains on increasing drive capacity and lowering the purchase price per gigabyte but is much less focused on disk performance and disk capacity and performance tradeoffs. The "price and capacity is everything" approach is good for storing data but what about for accessing and retrieving data? Disk cache provides a major performance benefit and is mandatory for most moderate to high activity subsystems. As disks get larger and the access density problem mounts, more cache is needed and costs increase since cache often ranges from 5-10 times more per gigabyte than raw disk storage.

If capacity doubled and performance doubled, the access density would remain unchanged. Unfortunately this is not happening. In reality, the access density has steadily declined as the capacity has increased much faster than performance. As time passes, very-high-capacity disks are not well suited for applications that have a large number of concurrent applications or users as increased contention for the single actuator results in elongated response times and high queue depth. For these reasons, the disk industry will likely consist of two levels or classes of disk drives in the future: lower-capacity, higher-performance drives and lower-activity, high-capacity drives.

Tape Limits

Tape cartridge capacities are also growing at unprecedented rates but are not faced with the super-paramagnetic effect like disk drives. Increasing tape capacities, without providing any offsetting performance improvements, lowers the effective retrieval capability of a given tape technology much as is the case with disk drives. Tape roadmaps project native, uncompressed cartridge capacities to reach 8TB or more by 2012, at the very latest. Similar to the key disk drive metric, Tape Throughput Density defines the ratio of tape cartridge capacity in gigabytes and tape data rate in megabytes per second (cartridge capacity divided by data rate). The throughput density formula can be applied to any tape format to determine if the tape is better suited for high-performance applications or more archival applications that demonstrate write-once-read-seldom-if-ever (WORSE) access profiles. As a guideline for vendor evaluations, a throughput density of 10:1 or less indicates higher-performance tape with good retrieval capability, while greater than 10:1 indicates high-capacity and archival tape.

From a technology standpoint, the limits of device capacity or even data rates are not yet evident and are now expected to continue indefinitely but at varying rates of improvement in the years ahead. Other limits may come into play however and need to be considered. The practical limits of storage technologies are based on the usability of the devices and include a new set of metrics. Utilization levels, throughput, performance, energy consumption and response times all represent the practical limits. The access density challenge has contributed to a steady decline in device allocation levels (as a percentage of device capacity) as the capacity of a disk increases. Storage administrators have become increasingly cautious about allocating too many applications or over-allocating storage on a disk drive in order to maintain acceptable levels of response time and device performance. As a result, disk allocation levels are declining as capacity increases. For non-mainframe systems, the average disk allocation levels are below 50% of the overall disk capacity. Lower allocation levels mean more disks must be acquired to meet the storage demand therefore increasing the overall cost of storage as well as the operational expenses needed to keep the storage functional. Most businesses have clearly indicated that they are not interested in spending more than they should or in purchasing capacity that they can't use.

Virtual Tape

Tape cartridge systems have benefited from the advent of virtual tape architectures and this may be one the few technology solutions that improves utilization, improves performance and lowers overall cost when installed. Virtual tape concepts, pioneered and popularized by IBM and Storage Tek for mainframe computers and virtual tape subsystems, first appeared in 1997. The benefits of virtual tape implementations are well documented and virtual tape is now becoming available for the faster growing non-mainframe computer systems market. Virtual tape combines both disk and tape into hybrid storage architecture. By embedding disk arrays as a front-end to a tape library, the disk storage serves as a cache reservoir for the larger-capacity and lower-cost tape library much like standard disk caching does. The device image presented to the operating system appears as a tape drive rather than the physical disk drive which it really is, therefore "virtualizing" the disk making it look like something other than it really is. As 2005 begins, the majority of disk storage vendors are offering some type of tape image on their disk subsystems as tape virtualization enters its second life for non-mainframe systems.

Pre-established policies such as file size, the value of data, object and file naming conventions and usage patterns are used to determine when the data is moved directly between the disk and the automated tape library without moving the data through the server. In this case, the policy-based functionality resides outboard of the server, directly controlling bi-directional data movement between disk and tape storage and multiple virtual tape files are allocated and written on a single cartridge. The major benefits for virtual tape are significant and include a much higher tape-cartridge utilization level over non-virtual tape, higher performance (as data can frequently be accessed from the disk cache) by defining more tape drives than physically exist, and a financial savings resulting from the reduction in the physical number of tape drives and media required.

Once virtual tape is implemented and the benefits are realized, the challenge of keeping up with tape technology progress begins to reappear. Virtual tape enables numerous files to reside on the same cartridge. Concurrent access to more than one file on a tape cartridge is impractical. Therefore, the access density issues that disk faces today by having too many concurrently active files on a single disk negatively impacting performance potentially awaits the tape market also, though in a slightly different form.

Conclusion

The bottom line for storage remains price in the eyes of most consumers; however, this can be very misleading. The declining utilization levels resulting from rapidly increasing capacities without corresponding performance increases are quickly becoming a major concern for savvy storage administrators. Though the price is initially appealing, other side affects are lurking underneath this appetizer of lower prices per gigabyte as a result of ever-increasing device capacities. These issues are moving the limits of storage from its technological limit to its practical limit. Are you riding the wave of seemingly limitless technology or preparing to face the oncoming practical limits in your business?

www.ibm.com

www.storagetek.com
COPYRIGHT 2005 West World Productions, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2005, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Storage Management
Author:Moore, Fred
Publication:Computer Technology Review
Geographic Code:1USA
Date:Jan 1, 2005
Words:1388
Previous Article:DVD Rot: the horror story that won't die.
Next Article:Storage finds a home ... at home: the couch potato runs the show.
Topics:


Related Articles
Storage Management--The Grand Vision Unfolds.
Winick Realty Group.
Bush, Putin sign arms reduction treaty.
The need for simplification: open storage management for complex storage environments. (Storage Networking).
Switch-fabric requirements for next generation storage directors. (Storage Networking).
Pro-active storage management on 2004 horizon.
Networked storage systems break the boundaries.
Innovation, litigation place increased demand on e-mail archiving.
When should I use DAS instead of NAS?
Fujitsu Wins Order from Institute for Cosmic Ray Research of University of Tokyo for Data Analysis System for Neutrino Research.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters