|only for RuBoard - do not distribute or recompile|
The SAN will not go the way of the 5.25 diskette. It is a robust, modular, scalable architecture. The SAN will respond to the two great, inevitable drivers of high technology: more and faster. A wonderful new driver is easier.
Faster, faster, faster. As mentioned earlier in this book, 2 Gbps Fibre Channel will earn end- user acceptance this year. In plain terms, as soon as manufacturers offer it, customers will buy it. This is startling, considering that 1 Gbps Fibre Channel is just growing in acceptance. Recall that 4 Gbps Fibre Channel is part of the ANSI standard, so can it be far behind? However, even at 2 Gbps some applications (such as moving uncompressed high-definition television data) become viable .
More distance, more distance, more distance. The SAN can reach greater distances in businesses and on campuses than SCSI. The fiber optic medium is relatively cheap and easy to install, so why shouldn t enterprises look for distance solutions to a variety of problems? If the speed of the FC connection will double or quadruple, do delays caused by distance really make a difference?
If you assume that fiber optic cable is no more difficult or expensive to lay than other connection media, then businesses, government, and schools can get more life out of existing buildings by expanding their interfloor and interbuilding links.
As telephone companies lay more fiber optic cable, businesses will take advantage of it for purposes of connecting to cross-town operations.
Further, nothing happens in a vacuum . Changes in distance and speed of data delivery over fiber will help drive changes to T-1, OC-3, and OC-12 wide area connection technology.
What are the social consequences of faster access over longer distances? Businesses, always conscious of the high price of real estate, might take two paths. First, they could locate data centers in declining urban districts, save some rent money, and help revitalize the neighborhood, too. The land is cheaper, the buildings are cheaper, and there s potentially no degradation in performance. Alternately, businesses might locate data centers on the prairies or in the foothills, helping those local economies. They d also be located away from the fault line, as those of us from earthquake-prone Los Angeles or Silicon Valley might say.
More capacity, more capacity, more capacity. It is the eternal cry of commercial Information Technology.
HP s disk capacity point, at this moment , is 47 GB, but that s just a passing figure. Higher-capacity disk drives with small form factors will continue to come. We would expect to see drives in the 100 GB range within a year, with more than 500 drives in large disk arrays.
The IT world abhors unused capacity, so as increased storage capacity becomes available, it will fill up. This will, of course, create a demand for greater capacity.
Bigger capacities for SANs will drive bigger capacities in Network Attached Storage and PCs. They will store more. Even at this moment, we know of a few PC implementations with 40+ GB of disk storage.
More interconnects. Switches and hubs shouldn t become the restraining component in building elaborate SANs. Given that the fabric switch is inherently more flexible than the FC-AL hub, we expect to see more switches with greater capabilities: more ports, more cascading options, more intelligence. It s inevitable that prices will come down as the number of units manufactured increases .
Hubs will not disappear. These devices will still have a place in smaller SANs, and public loops attached to switches. The hub might become a low-cost building block for implementing Fibre Channel in Small Office/Home Office (SOHO) installations.
Your stress level as a storage management professional will be lowered as the process of connecting new devices becomes easier.
Fabric switch technology already provides for cascadable switches. There should be plenty of ways to connect new devices to a SAN, and you will never feel strapped for ports. And we expect that switches will continue to develop good manners by discovering new devices and making them part of the fabric with a minimum of human interaction.
The worst-case scenario is that after connecting a new mass storage device, you ll simply walk over to your workstation, execute the Web-based management software, and bring the new device onto your SAN with just a few mouse clicks.
Of course (over the short term , until your operating system grows more intelligent ), you ll still have to partition, create file systems, and migrate data to new devices.
When it comes to backups, in many respects, the future is here. It arrived while this book was being written, and for many of us, it didn t arrive a moment too soon. Many of the features we d like to see in backup and restore are happening now. They include:
Native Fibre Channel tape drives. StorageTek and Exabyte have announced tape drives that don t require a Fibre Channel/SCSI bridge to connect.
Denser, faster tapes. The Ultrium Tape format and Sony DTF-2 tape have the capacity to meet tape backup needs for some years to come.
Bigger tape libraries. There s a better selection of large tape library solutions than ever, although the vast majority are SCSI-based and require a bridge to connect to the SAN. There is apparently no limit to the number of drives and number of tape slots that can be integrated into a single cabinet.
LAN-less and serverless backup. Hewlett-Packard s Omniback II software is the kind of utility that accomplishes this task.
Disk-to-disk backups. HP s Continuous Access XP software allows you to make copies from one XP256 disk array to another XP256. HP s Business Copy XP software lets you make copies to another disk in the same XP256.
The backup picture looks pretty good at the moment. However, we ll have to see how long it takes before increased online storage puts the backup process under more pressure.
Whether you are buying candy over the Internet, or lying on the operating table while your medical imaging data is being retrieved, the phrase the computer is down has become completely unacceptable. So it s necessary to maintain the health of the SAN as effectively as possible without stopping operations.
Fortunately, in several respects the future is here already. Consider the following factors:
Virtually every manufacturer of SAN equipment recognizes the fix on the fly operational model and has made the word robust a component of every product announcement.
Disk arrays have redundant fans, power supplies , and controllers. RAID technology helps overcome the problem of unavailable data caused by failed disk drives. Hot-swappable disk modules are sometimes featured.
Other devices, such as switches, have redundant components .
Smart hubs and switches bypass failed ports. The device may be unavailable, but the SAN continues to function.
A product announcement for a Brocade switch cites 70 fewer parts in the current model of the device. The reasoning is that with fewer parts , there s less to break.
Most devices report failures, using SNMP or a similar mechanism. Devices with the best manners have a phone home capability. The device calls the manufacturer s response center, and a customer engineer is on the way with replacement parts ”sometimes before the customer knows anything has gone wrong.
The smart SAN designer (you) has made no single point of failure a mantra. Every device is double-pathed. In The Book of Five Rings , written in 1643, Miyamoto Musashi points out, It is better to wield two long swords rather than just one, when facing a number of opponents alone.
As storage grows, we don t want storage management problems to grow. Especially, we don t want storage management stress to grow.
The goal is to eliminate the stress, as humankind prefers not to have problems bigger than its ability to manage them. We see the SAN of the future as being not more complicated to manage, but less complicated.
In this respect, the future can t get here fast enough. We expect to see a vast improvement in storage management software in just the next year.
Up to this point, there have been a number of methods for reporting problems at the device level: the ubiquitous HP-UX ioscan command, Event Monitoring Service, Storage Node Manager, and Fibre Channel Manager, just to name a few.
Complex storage devices, like HP s XP256 disk array, have a collection of monitoring and management software. The bag full of essential tools includes Logical Unit Size Expansion (LUSE) to make big LUNs, Custom Volume Size (CVS) to make little LUNs, Performance Manager XP to display graphs and adjust performance, Cache LUN to manage the cache, and Secure LUN to zone LUNs. And that s just for one device.
We expect that better and easier SAN management is on the immediate horizon. HP is justifiably proud of its SAN Manager DM and SAN Manager LM products. These applications provide broad control of key SAN elements, and once we feel comfortable with simple, comprehensive human management, we can look ahead to automated management.
AutoPath XP is a good example. This is a small application that performs automatic failover and load balancing over multiple paths. Of course, you have to define the paths, but we would expect that soon path management software will find the paths for you. Fixing a failed path will still require human intervention to replace a severed fiber optic cable or a failed GBIC.
Managing today s storage is still a little like starting a Model T Ford. It takes lots of little manual operations. This is soon to be replaced by a vastly more automated approach.
The operative term for the future of storage management is Policy Management. The simple definition: just tell the SAN what you want and it will do it for you. Ideally, those directives might be:
Add more storage
Allocate storage where and when it s needed
Manage my backups
Recover my data
Optimize your performance
Find the best solution and tell me
Find the best solution and don t tell me
|only for RuBoard - do not distribute or recompile|