The time is now to move storage from servers

At this moment, no aspect of enterprise computing deserves more attention than maintaining and storing corporate data. CTOs are increasingly abandoning traditional storage methods for more sophisticated approaches -- primarily SANs (storage area networks) and NAS (network attached storage) devices -- to contain the mountains of data their companies generate.

In 1999, only US$2.2 billion of the $30 billion spent worldwide on disk storage was spent on SAN or NAS devices, according to IDC, but this will shift dramatically in the next few years. IDC predicts that by 2003 combined NAS and SAN acquisitions will exceed one-third of the estimated $46 billion disk storage market.

When assessing your organization's needs, you'll have to answer many questions and consider several possible solutions before determining the enterprise storage method that best meets those needs. Finding that solution will depend on how tied you are to your current storage approach and on a careful assessment of future requirements.

InfoWorld recently surveyed more than 500 readers involved in managing and purchasing their organization's storage solution. The participants responded to an exhaustive storage-related questionnaire and revealed aspects of current implementations, plans for the future, and perceived problems and priorities.

The results of the InfoWorld Enterprise Storage Survey are impressive and sometimes surprising. They indicate that 96 percent of participants will increase their storage capacity in the next 12 months, and four out of 10 respondents expect to increase it by 50 percent or more in the next year.

If you are planning to add more than half of your company's current capacity to your storage pool, you should re-evaluate your requirements and explore possible alternatives that offer flexibility and are less costly to manage than current implementations.

If you've maxed out the capacity of your servers or storage cabinets, you are facing a more challenging task than simply adding a few more disks to the pool. You should also factor in the cost of managing the additional storage on a unit-by-unit basis, as legacy storage solutions demand. You are in for a big dent in your IT budget.

Resist the popular solution

An old IS axiom makes CPU, memory, and storage essential components of a computer. That probably trained us to purchase server boxes that included disk storage with increasingly sophisticated features such as RAID, which offers performance, resilience, and reliability improvements over JBOD (just a bunch of disks).

Offerings from top-notch vendors include life-saving options such as hot-swappable disks, redundant cooling fans and power supplies, multiple and increasingly faster SCSI controllers, and disks to absorb peak loads.

Server-hosted storage remains the most popular storage solution, at least in the client/server arena. If you're like 95 percent of our survey respondents, you probably already have a few working examples in your company.

Despite its popularity, this solution suffers from several disadvantages. Its expansion capability is limited by the server's architecture. Adding more disk space can require installing an external storage unit or replacing the media with higher-capacity disks.

Either option shuts down the server and initiates a lengthy and risky re-engineering process, rendering server-hosted storage inflexible.

Free space on one server cannot easily be allocated to a different server bulging with data.

Strike three against bundling server and storage in the same box: Maintenance activity on one will stop the other. CPU technology also progresses significantly faster and more frequently than do disk drives; your server's CPUs may become obsolete when the storage component is still adequate.

With server-hosted storage, you manage the disk space on each server as an isolated island that shares resources poorly, and you put components with very different requirements in a single box. Our heartfelt recommendation: If critical business applications are involved, run away from this configuration as fast as you can.

Storage quits the server closet

A significant improvement over server-hosted storage is installing a SCSI array, a separate enclosure filled with disk drives. You can buy the quantity of disk space to meet your current requirements -- varying from gigabytes to terabytes -- adding more as your company's needs grow.

Typically, SCSI arrays feature an embedded computer to manage configuration and to monitor performance. They can be connected to multiple, heterogeneous server hosts. Maintenance on your servers affects only the applications on that machine, an obvious advantage over server-hosted storage.

Although a SCSI enclosure raises the bar on storage growth significantly higher than server-hosted solutions, the problem remains that, after you fill up your cabinets, further expansion is expensive and messy. You may have to weigh the pros and cons of installing an additional array and connecting all your servers to the new unit, against acquiring one with more capacity and moving all your data to the new unit. Either choice imposes significant cost and poses the threat of extended downtime.

Although SCSI arrays do not readily forgive capacity misjudgments, they have been the state-of-the-art enterprise storage solution for several years. In fact, 67 percent of our survey respondents have SCSI arrays, and 39 percent plan to buy more during the next 12 months. These devices, however, may not be able to contain future growth.

Storage goes network

Without a doubt, the easiest and most economical way to add storage is to implement an NAS device. In most cases, you get a preconfigured box with an OS streamlined for handling data, with storage capacity varying from a few gigabytes to several terabytes. And you can replace faulty components such as disk drives, cooling units, and power supplies without service interruptions.

Because NAS units are OS-biased, be sure to choose one that will support your mix of server OSes. You can select rack-mounted or floor-standing units that come in different form factors. Expect to pay more for ultracompact units.

One benefit common to all NAS solutions is worry-free connectivity to your LAN. You attach the unit with a standard network cable (for the technically minded, that's an RJ45 connector), power up, and start using it. You should allot some time to define network access rights or disk quotas for each user. You may also want to reconfigure the RAID settings, easily accomplished from the unit's management panel or browser-based GUI.

NAS introduces an unprecedented degree of simplicity into storage management because it allows for easy, almost unlimited expansion, and it is applicable to most client/server platforms. The storage can be seamlessly shared among different servers or clients, and you don't have to power-off your servers or make many configuration changes when attaching a new unit. The additional storage immediately appears to your servers and clients as a new network share.

NAS' advantages are more and more widely appreciated: According to our survey, 39 percent of respondents use NAS in combination with other storage solutions. This is likely because NAS units' simplified installation, easy manageability, and flexible access make it easy to forgive their potential pitfalls.

By definition, NAS delivers network-based storage access, which can add a significant burden to your already burdened LAN.

In addition, no matter how fast your disk drives are, access speeds can be only as fast as your network bandwidth. At 10Mbps this could translate into sluggish performance, a handicap you can avoid by dedicating a high-bandwidth network segment (such as gigabit Ethernet) to storage traffic.

Finally, you must administer each NAS unit individually, so management costs will increase proportionally as you add more NAS boxes to the network.

A NAS solution can elegantly and inexpensively offer relief from storage starvation, but only if you do not require top-notch performance. For business-critical applications, be sure to use a dedicated, high-bandwidth network segment.

But if your company's growth forecasts indicate an excessive proliferation of NAS devices or demands bandwidth greater than gigabit Ethernet, you should consider a SAN solution, the highest end of high-end enterprise storage solutions.

The ultimate storage network

The No. 1 reason to implement a SAN is to concentrate all of your storage administration under a common management interface. When you purchase a SAN, you also acquire a suite of management software that allows you to administer data using an easy GUI, regardless of platforms and media considerations

In the SAN world, storage is taken to the next level beyond NAS and above physical media and platforms, becoming an abstract pool that can satisfy the mutable demands across your company.

Imagine SAN storage as a basin of soft clay that you can easily mold to match your varying business requirements and as easily return to the pool when the demand changes.

Interestingly, this common pool will meet storage requirements for an IBM S/390 as well as a Unix or Windows server. Welcome to the world of open, unbiased storage and say goodbye forever to separate storage devices for each OS.

As the name implies, a SAN is a dedicated network that connects servers and storage devices. As in traditional Ethernet networks, SAN devices and servers are connected with hubs and switches. A minimum configuration consists of one server and one storage device interconnected using a switch or a hub. As your organization's requirements grow, you can add more devices or servers to the network. There are no theoretical limits to the number of storage devices and servers in a SAN; however, bandwidth is an obvious practical limitation.

The typical physical media in a SAN solution is fiber cable, but the network can handle copper as well. Using fiber is the smart choice because it allows a higher throughput, from 1Gbps to 4Gbps, and a greater distance between the devices it supports.

In fact, future implementations promise a fiber WAN that can span the globe. Currently you can stretch a fiber connection up to 10 kilometers, which opens some interesting possibilities for disaster recovery and physical security managers.

To be part of a SAN, servers and storage devices must have proper adapters or protocol translators. Connecting your servers is relatively easy: Install an adapter -- the equivalent of the familiar Ethernet NIC (network interface card) -- and the proper drivers for your OS and connect to the switch using the fiber cable. Connecting storage devices is somewhat more difficult. Unless your IT staff likes building things themselves, you should buy fiber-ready storage devices such as disk arrays or tape libraries, which are available from many vendors.

Choosing the right vendor is probably the most difficult decision you'll make in your SAN project. Unfortunately, SAN solutions from different vendors do not necessarily work together. Therefore, you should plan to stick with a single vendor, at least until the standard committees of the Storage Networking Industry Association complete their work. Be warned: Do not expect to have complete vendor interoperability anytime soon.

Nevertheless, a SAN presents an opportunity too good to pass up. It's fast, easy to manage, offers practically unlimited expansion capabilities, and can accommodate storage for just about any OS, including mainframes. Its tiny, 1/8-inch pair of fiber cables provides freedom from distance constraints and from managing SCSI cables. It offers a storage-specific network that currently rivals gigabit Ethernet performance, and it's open to further enhancements. What's more, you can easily integrate existing storage solutions into a SAN for a simplified migration.

Although purchasing a SAN device is more expensive than, for example, a SCSI array with equivalent capacity, its simplified and less expensive management can easily offset the higher acquisition cost. Furthermore, vendors are offering SAN units with capacity and price ranges suitable for a department or even a small company or branch office.

Moving data storage outside

With such an impressive resume, the SAN architecture should be a hands-down winner in the storage context. Any company planning to increase storage significantly should consider a SAN.

Interestingly, this is not what our survey indicates. In fact, only 14 percent of the IT leaders InfoWorld surveyed said they had a SAN installed, and only 10 percent planned new implementations. Our participants indicated that the main reasons for not implementing a SAN were lack of internal knowledge, high cost, and the perception that it is a solution for larger companies.

If you recognize yourself in this picture, outsourcing your needs to a service provider with a SAN could be the right solution. Without the need to master new technologies or to make huge, up-front investments, you can achieve the benefits of having a SAN without actually having a SAN.

An SSP (storage service provider) will analyze your requirements and deliver the storage capacity that your applications need to a wall outlet on your premises just as easily as you get water or electricity. In our Test Center Analysis, on page 54, we discuss SSPs and the ways in which they can benefit your company.

Every company faces challenging choices, and choosing the right storage architecture tops the list. As indicated in the InfoWorld Enterprise Storage Survey, a huge amount of your IT budget will be spent on storage in the next 12 months. Consider it an opportunity to give your company a more agile and reliable storage infrastructure that will grow as your business does.

Mario Apicella has managed the storage requirements at several companies. Share your plans for future storage with him at mario_apicella@infoworld.com.

THE BOTTOM LINE

Enterprise storage

Business Case: The fast pace of e-business will not allow you to close your shop to upgrade storage capacity. A carefully selected enterprise storage architecture will support your company better and minimize disastrous stops.

Technology Case: Hardware compatibility is granted by standards, but differences in storage administration software make mixing various NAS and SAN solutions risky.

Pros:

+ Many solutions available from reputable vendors

+ Evolving technologies with significant future improvements in performance and manageability

+ Standards committees working to define interoperability standards

Cons:

- Evolving market

- Incompatible management software

This story, "The time is now to move storage from servers" was originally published by InfoWorld.

From CIO: 8 Free Online Courses to Grow Your Tech Skills
Join the discussion
Be the first to comment on this article. Our Commenting Policies