×

SUBSCRIBE TO TMCnet
TMCnet - World's Largest Communications and Technology Community

CHANNEL BY TOPICS


QUICK LINKS




 
February 1998
Desktop Or Commercial-Grade PCs - Which Is The Right Platform For CTI?

BY COSTANTINO CAROPPO

Spurred by advances in computing technology and squeezed by downsizing, original equipment manufacturers (OEMs) have been developing more and more applications that run on outsourced, openarchitecture computers, as opposed to the proprietary computing platforms created in-house by their computer divisions. A whole spectrum of computing platforms is available — from inexpensive desktop-grade PCs to highly rugged VME systems capable of delivering deterministic data processing for complex real-time applications.

While requirements for rugged systems in demanding CTI applications are often satisfied with VME platforms or the up and coming CompactPCI systems, many CTI applications are currently running on commercialgrade, passive backplane systems or desktopgrade motherboard PCs. The decision between desktop-grade and commer-cialgrade systems may get more com-plicated by the fact that, over the next few years, motherboard systems may become so inexpensive that some of the current reasons for basing several CTI applications on passive backplane systems may be defeated.

However, CTI system designers and decision makers must understand that there are crucial differences between low-end and high-end passive backplane systems. In fact, while the price premium of low-end passive backplane may not buy features much superior to those of motherboard systems, highend passive backplane systems offer advantages that will not be offset by the low price of motherboard-based systems.

PRICE-PERFORMANCE SPECTRUM
Depending on the degree of ruggedness, PC-compatible computing platforms can be classified as desktopgrade, commercial-grade and industri-algrade. Desktop-grade computers are motherboard-based desktop PCs and workstations; commercial-grade computers are passive backplanebased computers utilizing ISA and PCI/ISA buses; and industrial-grade computers are mostly passive backplane computers based on more hardy buses such as VME, STD32, Multibus I and II, Futurebus, PC/104, and CompactPCI.

If we classified the most common OEM computers — i.e., motherboard, ISA and PCI/ISA, CompactPCI, and VME systems — according to price and performance criteria (where performance does not consist of ruggedness alone, but includes bus flexibility, bandwidth, and support of advanced features), a spectrum showing motherboards as the lowest price/lowest performance system and VME or CompactPCI as the highest price/highest performance system would result.

Over the next few years, the line between desktop-grade and commer-cialgrade computers will likely shift due to some dramatic improvements in desktop computing. Indeed, as desktop computing economies continue to improve and reliability increases, these two seemingly distant worlds will come together. And, as desktop systems become feasible alternatives to commercial-grade systems, OEMs will face decisions as to which platforms to utilize — decisions that go well beyond engineering requirements, and involve the entire economics of a product. It is, therefore, fundamental that OEMs understand how these trends may make desktop-grade PCs viable alternatives to industrial-grade computers in order to make strategic design decisions for their next-generation CTI products.

DESKTOP PCs
On the lower end of the price/performance spectrum, desktop PCs represent a $400-billion market that enjoys vast economies of scale and steady progress in computing capacity. Since the same desktop PC used for typing this article can be used to control a simple food packaging station or to run a small company’s voice mail, it is easy to understand why commercial applications with minimum system requirements are built on desktop PCs. Desktops are motherboard-based computers, with limited capabilities for expansion (depending on the number of expansion slots available on the motherboard) and low- to mediumquality components: metal, fans, power supplies, motherboard, and peripherals.

COMMERCIAL-GRADE PCs
Technologically similar to desktop computers, commercial-grade computers are mostly larger, higher quality computers that employ a passive backplane architecture and are more prone to custom configurations than desktop PCs (Figure 2). Most of the passive backplane systems currently used as low-end CTI platforms are legacy systems that employ the 16-bit ISA bus, while higher-end CTI systems have been employing PCI/ISA backplanes that take advantage of the 32-bit PCI bus. By utilizing passive backplanes instead of motherboards, commercialgrade computers can support as many as 20 expansion slots, which make them highly suitable for CTI applications that require a high port count in as little rack space as possible.

DECISIONS, DECISIONS...
Depending on the application and, perhaps even more so, on the competitive pressures of their marketplace, OEMs may prefer desktop-grade computers over commercial-grade computers as platforms for their applications, or vice versa. For example, simple voice mail applications for use on customer premises might prove to be more competitive on a desktop computer due to the low cost of the system and the intense price pressure voice mail software vendors and integrators are experiencing in their marketplace. On the other hand, commercial-grade computers may be preferable to desktops when applications require a higher degree of availability, as is the case with IVR (interactive voice response) servers used by corporations to deliver rev-enuegenerating services like bank account balances or stock quotes.

Of course, common sense has it that desktop PCs are chosen for two reasons: impressive cost savings and acceptable reliability. Today, a fully-loaded 200 MHz Pentium system does not cost more than $1,500. Compared with a 200 MHz Pentium commercial-grade system, the desktop system may be as much as $500 to $1,000 cheaper. From a reliability standpoint, it is commonly accepted that a desktop system will probably operate for three to five years before a failure occurs. Factored together, low cost and acceptable reliability could be powerful reasons against a more reliable but costlier commercial-grade PC. On the other hand, commercialgrade PCs have enjoyed a growing reputation as the choice solution for mission critical applications. Passive backplane commercial-grade systems generally offer better quality and reliability characteristics than desktop PCs, in addition to being more serviceable, scalable, upgradable, fault-tolerant, and stable. Indeed, by design, the passive backplane architecture (whether it’s ISA, PCI/ISA, or the more recent CompactPCI) minimizes mean time to repair (MTTR) by moving all active system components from the motherboard to separate insertion cards residing on the backplane, all of which can be quickly substituted in case of failure.

In addition, several commercialgrade PC manufacturers have stringent requirements on CPU board quality and subject their CPU boards to rigorous environmental tests such as the Dynamic Environmental Screening Test, where CPU operations are monitored under varying degrees of temperature. Even more mundane components such as power supplies — which are prone to high failure rates — are subject to extensive scrutiny.

COST OF OWNERSHIP
In fact, the most compelling argument in favor of commercial-grade PCs is the total cost of ownership, as compared to the up-front investment cost of desktop PCs. Even though a desktop PC may initially cost as little as half of a passive backplane PC, there are other costs to be accounted for. These include all costs associated with equipment downtime, the cost of servicing equipment, the cost of the inability to upgrade or scale-up, and the cost of frequent changes. When fully accounted for, these downstream costs may very well end up offsetting the considerable up-front savings of a desktop-based PC platform and tip the scale in favor of the more robust commercialgrade computer.

INEXPENSIVE PCs FOR YOUR CTI APPS
What if desktop PCs were onethird cheaper than commercial-grade PCs? This is a very real possibility, as some trends seem to indicate. Several key components — memory, storage devices, and other I/O functions — have seen dramatic price declines and will most likely continue on this track, provided that the growth of the desktop market does not top off and that there is no supply undercapacity. More importantly, the recently emerged NC (network computer) will likely make inroads into the vast desktop market by means of its extremely lowcost, stripped down hardware which, in turn, will put even more competitive pressure on the full-blown version of a PC. At the recently-concluded Comdex ’97, several NC manufacturers launched basic products at prices as low as $500 and highend NCs as expensive as $1,000.

A crucial consequence of this trend is the possible expectation of lower PC hardware prices and the subsequent price decreases to match that expectation. If the trend holds true, OEMs, VARs, and system integrators may soon ask themselves whether a desktop PC is better suited for their applications on the basis of up-front price alone. In fact, several telecommunications applications — simple voice processing systems for service bureaus or voice mail/audiotex systems for customer premises — are already utilizing motherboard systems.

CLUSTERING
Even more intriguing is the possibility of networked PCs with relatively inexpensive fault tolerant software, such as the upcoming clusterenabled version of Windows NT, codenamed Wolfpack. By transforming each networked PC into a cluster node, Microsoft’s new operating system will enable any one server in a clustered environment to automatically take over operations in the event that another server goes down. Clustering has the obvious advantage of improved availability, and also offers good manageability by allowing the centralized movement of applications from one cluster node to another via point-andclick network configuration. As a result, costs linked to system failures and MTTR are reduced and hardware costs become negligible, operating system software may end up accounting for most of the system cost. Although the current version of Wolfpack has limited capabilities (2node only), its price is significantly less than UNIX-based clustering — the two-node Wolfpack costs about $23,000 vs. $50,000 for UNIX. Ultimately, as more third-party vendors develop Wolfpack-enabled solutions and Microsoft lowers prices, the choice between a desktop-based system and a passive backplane-based system becomes less straightforward, unless the application necessitates hardware features that only high-end commer-cialgrade systems can offer.

For decision-makers involved with CTI products, the implications of the above trends could be farreaching, depending on the type of application. Some applications lend themselves to taking advantage of the future cost-saving potentials from low-cost hardware. A couple of years from now, a nofrills Pentium-based PC under $1,000, with storage, I/O, and OS included, may be a more suitable platform for entrylevel customer premise equipment. An alternative solution could be a no-frills, lowend passive backplane system which may end up costing almost twice as much. Indeed, $250 for a low-end enclosure with a passive backplane, $750 for a Pentium CPU board, and probably an additional $400$800 for I/O, OS and storage quickly add up to $1,400–$1,800.

Is a $400–$800 additional expense worth the assurance that a passive backplane system provides against a possible system failure? Maybe not, if a quick component replacement is possible, and if the component to be replaced (i.e., a motherboard) is relatively inexpensive. After all, even with a superior MTBF (mean time between failures) granted by its architecture, the passive backplane solution may end up being too expensive over the life-span of a product. In other words, the limited performance difference between a motherboard and a low-end passive backplane system may end up being offset by the cost savings of a desktop-grade PC. It is then important to distinguish between a low-end and a high-end passive backplane computer, since the former differs little from a motherboard (except from its passive architecture), while the latter differs quite dramatically.

NOT ALL PASSIVE BACKPLANE SYSTEMS ARE EQUAL
Most commercial-grade PCs provide a robust computing platform that is more suitable to business productivity than desktop PCs. However, what is currently misunderstood about commercialgrade passive backplane systems is the fact that they encompass two price/performance tiers — one lower and one higher.

Low-End Systems
Low-end passive backplane systems are mostly systems that employ commoditized passive backplane CPU boards, which are often produced overseas. These CPU boards offer attractive combinations of I/O functions at a low price. At the end of the day, a lowend 10-slot passive backplane chassis, a Pentium CPU with on-board SCSI and VGA and a few peripherals may cost as little as $1,500. Yet, they often do not have buffers to drive 20 expansion slots, a true commercial-grade design (for example, obtrusive memory SIMMS do not allow for optimal cooling), and — most importantly — are subject to design changes that are beyond the control of their vendors. Similarly, low-end passive backplane systems take advantage of lowcost enclosures with commercial grade-like features such as thicker metal, larger fans, and larger power supplies. However, lower enclosure prices may mean sub-optimal chassis design, barely acceptable cooling and EMI performance, and most notably little or no room for customization. Does all this sound familiar? It should, because it’s not much different from a motherboard-based approach. With the exception of the passive versus active motherboard architecture, lowend passive backplane systems offer the cost savings typical of off-theshelf desktop systems, yet remain affected by the limitations of desktopgrade systems.

High-End Systems
By contrast, high-end passive backplane systems are more expensive com-mercialgrade solutions (Figure 3). Most of the time, they employ enclosures designed to enhance cooling, minimize EMI (electromagnetic interference), and optimize system maintenance — elements that can make or break the effectiveness of certain applications. For instance, a unified messaging equipment company may want to offer the possibility of upgrading the storage capability of its servers with the most recent high-capacity hard drives. If the servers are based on highend commercial-grade systems with optimal cooling, the transition from a 2.1 GB SCSI drive to a 9 GB Ultra-2 SCSI is not a cause for concern with regard to thermal issues. However, a server design based on a cheaper, lowend passive backplane system may call for a chassis redesign which will support increased heat dissipation requirements, or it may cause the selection of a different chassis if the thermal issue can’t be solved due to the limitations of a low-end passive backplane platform.

From an ease-of-upgrade standpoint, if your company is planning to upgrade the IVR servers currently deployed at customer sites with new Dialogic or Natural MicroSystems PCI voice cards, the ease with which a technician unplugs devices and substitutes an ISA backplane for a PCI backplane may have a large cost impact on the labor each unit requires. Easy maintenance and upgrades of a system are the fruit of high-quality hardware design, where details like the placement and number of screws used to hold a fan tray are not a matter of compromise. If attention to details is important, low-end commercialgrade systems may not be fit for the job.

MORE DECISIONS...
Most importantly, high-end passive backplane manufacturers offer design and revision control over lowerend commercial-grade systems. Most of the time, lower-end passive backplane systems are manufactured abroad in large lots by companies whose main emphasis is cost-reduction. Because of their business model, such vendors elect to source whatever is cheapest at build time, regardless of a customer’s application. If a lower-cost chipset comes about, the vendor may just re-lay out the CPU board and modify the BIOS, which spells change for the CTI application unless your purchasing department bought enough supply to cover the entire program. On the other end, higherpriced passive backplane systems may supply the assurance that no changes will need to be implemented because their vendor will continue to design, source, manufacture, and support older components.

CONSIDERATIONS FOR PLANNING NEXT-GENERATION PLATFORMS
If the assumptions we’ve made about the future of desktops and commercialgrade computers hold true, what should you consider when designing a nextgeneration CTI product?

Have Clear Priorities
Is price your first priority, or is it slot count? Can you live with hardware changes, or do you need as much stability as possible? Are you willing to pay more for customization, or can you live with off-the-shelf PCs at a lower price? These are the questions you must ask yourself at the start of and throughout a project.

Up-Front Price and Total Ownership Price Are Different
If low price is paramount to the success of your application, you may want to consider using a motherboardbased system instead of a passive backplane system, especially if you are thinking about using a lower-end passive backplane system. You will probably see the advantages of selecting a motherboard system within the next two or three years, as prices decline even further. To be sure, a low-end passivebackplane system that costs $500 to $1,000 more than a motherboard system may provide a larger slot count and better MTBF. But it would not buy you stability of supply, custom modifications, optimal serviceability, low EMI emissions, and high cooling performance.

However, you should keep in mind that desktops and low-end passive backplanes change hardware often (sometimes more than once a year), and before discounting a more expensive high-end passive backplane solution, you should try to calculate how much two or three hardware changes (CPU, chipset, BIOS) over your product lifespan may cost your company, and then factor that into the financial projections of the product.

Slot Count Is Critical
If your CTI application requires more than 7–8 expansion slots, motherboard solutions are excluded from the start, unless you are able to find a manufacturer willing to build you a custom motherboard at the quantities required by your system. On the other hand, you may want to consider a passive backplane system for low slotcount systems, especially those that are able to support multiple segments within one enclosure. By virtue of their architecture, multiple-segment backplane systems are very advantageous for applications where rack space is critical. For example, a company that makes an enhanced service platform for one-number voice/data services using dense voice processing boards (multi-function, high port count) can literally fit two completely independent servers within one 19'' rackmount high-availability enclosure. Finally, if you select a passive backplane system from a low-end manufacturer, make sure that their low-cost CPU supports 20 fully loaded slots, as many of them only support up to 10 slots.

Hardware Stability Has Its Advantages
If you are willing to support hardware changes once or twice a year, motherboards and low-end commercialgrade systems offer the advantage of costing $500 to $1,000 less than high-end com-mercialgrade systems. However, if there will ever be a need for avoiding a specific hardware change, you will not be able to do so if you select an off-theshelf PC, and you will have a very hard time convincing your low-end passivebackplane vendor to run one more batch of CPUs with, for instance, a specific BIOS version.

Direct VendorCustomer Relationship Has Its Advantages On the same note as with hardware stability, you will most likely find it hard to communicate directly to a motherboard manufacturer and explain your problems, wishes, and concerns for the future, unless you buy thousands of motherboards each month. Commercialgrade PC manufacturers, who sell to low- to medium-volume clients, will be more accessible to your needs than motherboard vendors. However, remember that low-end commercial-grade system vendors normally shy away from custom work, unless you can buy in very large lots. If your CTI solution calls for application-specific customizations, most of the top high-end passive backplane system vendors will do the work.

THE DIFFERENCE IS IN THE DETAILS
Despite their cost advantages, if you choose a motherboard-based system or a low-end passive backplane for your application you will need to compromise on details that may be important to you, like optimal chassis cooling, low EMI emissions, easily removable cables and parts, low backplane noise, or DRAM SIMMS impeding air circulation close to the CPU chip.

Ultimately, it is critical to remain informed about the trends developing in desktop-grade, commercialgrade, and industrial-grade computing in order to make the right strategic decisions on computer platforms for your CTI applications. Armed with the right information and with clear objectives for your products and services, you will likely be able to leverage the breakthroughs occurring in open-architec-ture, PC-compatible computing.

Costantino Caroppo is a senior product marketing analyst at I-Bus, an ISO 9001 provider of custom and semi-custom PC-compatible computers to computer telephony and telecommunications OEMs. I-Bus designs and manufactures commercial-grade enclosures, CPU boards, passive backplanes, highavailability passive backplane platforms, and CompactPCI turnkey systems. For more information, call IBus at 800-382-4229 or visit the company’s Web site at www.ibus.com .







Technology Marketing Corporation

2 Trap Falls Road Suite 106, Shelton, CT 06484 USA
Ph: +1-203-852-6800, 800-243-6002

General comments: [email protected].
Comments about this site: [email protected].

STAY CURRENT YOUR WAY

© 2024 Technology Marketing Corporation. All rights reserved | Privacy Policy