With the explosion of data and the decentralization of hardware computing resources, storage management in corporate environments becomes more complex and poses new challenges for administrators and IS managers. Storage costs have escalated to such astounding levels that it is easy to predict that many organizations are heading to a crisis.
Many of the concepts developed to lower storage cost through new cost effective media or increased automation processes, have yet to gain the popularity they deserve because they are perceived as being too complex. Somehow the plug-and-play approach that proved so successful promoting the utilization of printers on networks, was never applied to storage.
A new concept has quietly
surfaced: Network Ready Storage (NRS). Originally developed to
address the connectivity problem that plagued the introduction
of optical drives and jukeboxes, this concept is has now expanded
to solve a more basic concern of how to add disk drives to heterogeneous
environments. It is an elegant solution whose prime objective
is simplicity. It relieves the administrator from the burden of
configuring, testing, and ensuring that independently selected
modules integrate harmoniously. We submit that simplicity and
reliability are the gating characteristics for market acceptance
of automation.
Network disk storage capacities are growing at rates exceeding 60% per year.The average yearly cost of managing storage is $350,000 for a PC LAN and $750,000 per network in the UNIX environment. While every megabyte of disk added to a PC LAN costs less than one dollar to purchase, the company incurs an additional $7 per megabyte each year in labor and lost productivity to manage this storage. These costs are primarily the result of incoherent, mostly manual storage management processes. Quite often administrators are oblivious to a pending storage crisis because they are too busy installing new hardware and software, reconfiguring and keeping the network up and running.
The average UNIX site has 14 servers and adds 3 servers and 35 GB of disk storage every year, as shown on Figure 1. On the average, once a month disks are added or reconfigured. Installation and maintenance of new hardware and software occupies an important portion of an administrator’s time. They shy away from new automated techniques that promise increased productivity because they are difficult to understand, install and maintain. For example, automatic file migration and hierarchical storage management have been available for some time, but have not succeeded to penetrate the market. Administrators are also skeptical about vendors’ claims of compatibility. Incompatibilities often prevent host adapters, drivers, libraries and jukeboxes from integrating with existing hardware and software.
STORAGE EXPANSION
1993 1993
Individual Total Server Total Server Capacity
Server/Site Server Capacity Capacity/Site Annual Growth Rate
------------------------------------------------------------------------------
UNIX 14 7.1 GB 99.4 GB 37%
PCLAN 7 2.6 GB 18.2 GB 85%
The overall complexity and difficulty of installing additional
storage devices and HSM software create an opportunity for a storage
solution that has not only the required performance and reliability,
but ensures easy product installation and maintenance.
A solution must be optimized to each particular set of requirements because of the complex nature of the problem which is bounded by emotional, psychological and business challenges and constraints, as shown in Figure 2. Generally speaking, solutions must be simple and reliable, and involve storage automation. NRS meets the challenge. It fills in a market need and significantly contributes to the overall solution. But, first, what is NRS?
NRS is a fully integrated and dedicated storage solution that can easily and quickly attach to a network backbone, becoming immediately and transparently available as a network resource to all clients. It is platform and operating system independent, and appears to any application as another server (Figure 3). It can be brought on-line without shutting down the network, and requires no changes to existing file servers.
It consists of a simplified server that performs only one function, that of controlling disk storage, but performs it well, not having to meet the conflicting requirements of a general purpose system. While it cannot run applications or databases, it can store files used in applications and databases and its ease of use and price/performance are unmatched.
The challenges in network storage expansion are to provide an easy- to-install, easy-to-use and easy-to-maintain product to meet the heterogeneous needs of a constantly changing environment at a reasonable cost and without compromising the performance. There are two choices today for expanding storage: add storage to an existing server, or add a new server.
In the first instance you have to bring the server down, power it off, open the box, install a host adapter, attach the device, bring the server up, install a driver to communicate with the host adapter, and run tests to ensure that there are no incompatibilities. This operation takes 1 to 5 hours, must be performed off prime working time, and prevents clients from accessing data on the server. In addition, adding disk to an existing server decreases the performance of the server.
Adding a new server involves costs for hardware and software, but primarily costs associated with reconfiguring the system, connecting to the SCSI ports, providing or attaching drivers. This operation, on the average, occupies one skilled administrator anywhere from 2 hours to 2 days. If you need to update your operating system, you have to worry about your storage solution that might be dependent on the previous version.
A third choice is now possible: NRS. In comparison to the above two approaches, installing NRS takes minutes: one plug to the LAN, one plug to the power supply. By avoiding the multiple choice approach, configuring it involves very few steps made easy through a simple administrative interface that is often menu-driven or graphical. Furthermore, NRS is not affected by upgrades of servers, operating systems and applications.
The user needs a solution that protects data accessibility, offers scaleability for modular expansion, and improves administrator and user productivity.
Data accessibility is key to reducing downtime and improving user productivity. Inaccessible data means lost productivity. Most networks experience regular downtime as servers are taken offline for planned maintenance or unplanned outages. An NRS approach reduces network downtime by maintaining access to data during server outages. NRS, in effect, externalizes storage from the server. The server does not have to be online for users to get to their data. The addition of NRS does not affect the performance of existing servers. It may actually increase server performance by sharing the load.
NRS ties to a network protocol and topology. This makes it available to the entire network supporting heterogeneous platforms and operating systems. As an example, NRS has access to 100% of the UNIX market if it supports NFS.
With NRS, you can integrate additional storage at any location without changing your overall storage strategy; buy and plug-in the exact amount and type of storage you need now and grow incrementally. The system platform independence and direct network connectivity allow corporations to retain their prior investments in their storage devices when they change environments, and optimize their utilization.
Though not possible in all implementations, storage can be expanded at will. More storage can be added at any time, given different addresses, and accessed transparently. Any number of devices can be added to the LAN, unlike server-based devices which have a limited number of SCSI addresses.
The solution needs to be simple, reliable and automated. We have discussed how simple NRS can be when it comes to installation and maintenance. The fact that its connection does not affect other components plugged on the network, removes the mystery aspect. It eliminates the overhead of the additional components that give a general purpose computer its flexibility with a very large number of patterns; the simpler hardware and software reduces the number of points of failures, and becomes easier to test, thus providing increased reliability.
The NRS concept is not necessarily
automated in itself, since it can apply to a simple addition of
disk drives. However, it paves the way to a simple integration
of automated tools. When connected with a library or a jukebox,
a disk cache keeps frequently used data blocks on hard disk so
repetitive requests are handled quickly. Integrating NRS with
hierarchical storage management products that manage other connected
storage devices is one of the foreseen evolutions of this concept
The principle is simple. An NRS product contains the basic modules of a server that interfaces to one or more network protocols on one end, and to a SCSI port for attaching devices or library components at the other end. To applications running on the network, NRS looks like an ordinary server. To any client, it simply looks like a storage resource.
When used in conjunction with an optical jukebox, an internal storage management hierarchy may use a combination of RAM, hard disk cache and optical technology to migrate data upwards and downwards within the system, depending on data access patterns.
The implementations differ according to the characteristics that the developer wants to emphasize, and the breadth of storage technologies and devices that are targeted for support. The primary characteristics are:
Functionality: Many implementations start with a full fledged server and limit its functionalities to satisfy the Plug-and-Play concept. The potential for greater flexibility is inherent in the design, and the vendor may be tempted to allow the user to benefit from it, essentially providing a workstation functionality to the product. The drawback is obvious; this approach may defeat the purpose of simplicity that is the first objective of the product. This approach may not achieve the cost nor the additional reliability that one can expect to obtain from a stripped down implementation version.
Performance: By restricting the number of functions. the implementation may build in the product a high degree of performance related to speed, simultaneity and/or additional availability through fault tolerant components.
Cost: The safest way to minimize cost is to implement from the ground up a design optimizing the sole function of attaching storage devices. This implementation puts more burden on the product developer.
Implementations vary also with the supported storage technologies and devices. Some may restrict it to CD-ROM, other to any optical technologies, while others may provide access to any storage technology including tape and magnetic disk.
The smaller the configuration, the greater the sensitivity to cost. Adding a few gigabytes of disk drive, or a small jukebox configuration may greatly benefit from an implementation that focuses on cost.
The system’s platform
independence and direct network connectivity allow corporations
to retain their prior investment in storage devices. The NRS sits
on the network like any other network device but does not significantly
increase administrative overhead or incur the network costs typically
associated with primary on-line storage.
The cost of managing storage far outstrips the acquisition costs of hardware and software. Administrators spend enormous amounts of time installing and maintaining storage on the network, oblivious to a pending crisis. The market is in crying need for solutions that simplify their task and improve productivity.
The model for managing storage is undergoing great changes, stimulated by new architectural concepts, a growing need for a high level of data accessibility, and the constraint of heterogeneous platforms and operating systems.
The NRS concept represents a logical evolution in the storage paradigm. It can provide needed additional capacity through a direct and transparent connection anywhere on a network. It enables workgroup-centered storage to be placed next to its users, as opposed to centralized storage confined to one corporate location and one host platform. Its plug-and-play ability will start a new era in storage, one that stimulates the utilization of automated hardware and software which were judged too complex. The credibility of NRS has been demonstrated through its adoption by system vendors such as IBM and Hewlett-Packard, as well as storage vendors such as Maxoptix, QStar, and Symmetrical Technologies.
NRS will also prove effective
for simple addition of disk drives to the network, a cost-effective
alternative to adding new multi-purpose servers. Multi-purpose
server based storage is powerful and configurable. NRS is easy,
independent and network ready.
This report was
produced by:
Peripheral Concepts, Inc.
and
Strategic Research Corp.
Peripheral Concepts, Inc. - Farid Neema,
President of Peripheral Concepts, Inc., has thirty years of experience
in the computer industry. Peripheral Concepts was formed in 1991.
It specializes in market research and consulting in the industries
related to data storage and printers, with emphasis on the systems,
the applications, and the end user requirements, The company’s
services include product, strategic, and business planning as
well as competitive analysis and product development.
Strategic Research Corporation - Michael Peterson, President and senior analyst of Strategic Research Corp., has twenty-two years of experience in the software, drive, head, and media industries. Strategic Research publishes reports on the storage software, hardware, and network markets, conducts user and channel research, and consults in the areas of strategic planning and product and market development. The company also sponsors and facilitates industry forums and trade associations.