TECHiNSIGHT - Server Technology: A Platform Approach Story

Home > Server Technology: A Platform Approach Story


	> Whitebook War II > Raid Controllers > Hardware Security > Torture On The Rack > What Matters

For over a decade, the PC industry has puzzled over how to plant computing in the living room. Designs and marketing ploys from vendors big and small have floated through the channel, and none of them have stuck. Microsoft’s Windows XP Media Center Edition (MCE), now in its third incarnation, has at last managed to nail the right mix of power, ease of use, and playback quality to deliver a satisfying PC-meets-TV experience, but not even Microsoft can conquer the living room in a solo effort.

Dependability is the hallmark of a well-built white box, whether you're talking about desktop PCs, media center systems, notebooks, or servers. Without it, performance loses relevancy. But the customer who buys a living-room machine will define stability differently than an IT manager, and the hardware in each product should testify to that divide.

For example, home theater PCs are expected to dutifully play back movies, tune into television programming, and maneuver through the occasional game. Servers, on the other hand, must deliver Web pages, traverse massive databases, and play postal worker to entire corporations, all without dropping offline, succumbing to heat, or going down with the power.

Easier said than done, right? It turns out that in the most extreme cases, the dependability equation is directly proportional to your customer's budget. We're talking enterprise email servers, which are up for all but an hour each year (99.99% availability), or telephone systems that might go down for five minutes during the same stretch (99.999%). Small businesses without the cash for redundant technologies are more likely to see somewhere in the neighborhood of 99% availability--less than 3.7 days per year of downtime.

Take it from a small business consultant who has had to trek across California several times in the dead of night to troubleshoot ailing servers--guaranteeing your customers even 99% availability is difficult. The challenge is worth confronting, though. Billing for the monthly maintenance of a healthy server is much more pleasant than emergency break-fix visits. You'll also have an easier time retaining good customers. Of course, that makes it doubly important to deploy platforms architected with enterprise-class functionality and an industrial strength software infrastructure at every price point.

It's In the Platform

In many areas of computing, performanceis the catch-all by which vendors sling their newest wares. It's an understandable phenomenon. Performance is easy to measure through readily available and easily comparable tests. However, because such metrics only run briefly and often without applying the full rigors of daily use, they fail to take other factors into account, such as reliability. Therein lies the problem.

Without empirical data demonstrating the long-term robustness of your servers, it's difficult to sell into a market reliant on stability. Sammy Wong, sales manager at Universal Systems, Inc., gets around that obstacle by designing platform-oriented servers instead of focusing on any one component.

"VARs want to build the most solid systems possible. They don't want to wonder if they picked the right processor or RAID controller. And most resellers don't have the resources to validate an off-the-shelf motherboard with another vendor's chassis."

In response, barebones kits are becoming increasingly popular as server board manufacturers find value in mating their products to compatible enclosures. The combination takes one significant load off of the VAR's shoulders. However, there are other factors to consider beyond picking a motherboard/chassis combination. For example, certain processor features require explicit BIOS support. It would be a shame to build a fancy, high-end server only to discover your third-party board was missing support for the latest power optimization. According to Wong, Intel's own kits are popular due to the tight integration between all of the server components.

"Intel's platform--the processor, chipset, motherboard, and chassis--is validated by Intel. We know that all of the pieces will work together and our enterprise customers who cannot afford downtime are confident in the infrastructure. This isn't such a big concern on the desktop, but it's very important to our server business."

By focusing on the entire platform, VARs are able to effectively minimize post-sale problems they might otherwise encounter with products sourced from different vendors. And because Intel is heavily vested in server technology, low-level components such as chipsets and network controllers, along with more prominent processors, motherboards, and chassis, are all validated for interoperability by the same vendor.

"Enterprise customers are intolerant of server downtime yet still demand maximum performance," says Sammy Wong. "They are the ones who most immediately see value in an all-Intel infrastructure, which effectively addresses both points."

Selling Trust

Architecting a solid server is only HALF of the battle. Medium-sized businesses need your support when it comes to hardware issues. And smaller organizations without IT departments require an even closer working arrangement in order to keep their technology running smoothly. Such reliance puts you in a trust relationship that extends long after the initial sale.

Being tied to the servers you built a year or more ago doesn't have to be a burden, fortunately. Universal Systems' Sammy Wong is as bullish on Intel's reseller support structure as he is on the server platforms themselves.

"As a VAR, Intel's overnight replacement policy is huge. If one of my customers has some sort of hardware failure, Intel will send an advanced replacement with paid FedEx shipping both ways. The server could be 18 months old--it doesn't matter. Getting my customer up and running with a new component, covered under warranty, the very next day is instrumental in preserving our relationships."

Erik Logan, CTO at Pogo Linux, likes to focus on the convenience of working with one vendor for support issues.

"Standardizing on components from one vendor helps improve interoperability. If we build a server with an Intel chip, an Intel motherboard, and Intel networking, there is less chance of having trouble with that solution. Once you start sourcing parts from a number of different vendors, the support landscape starts getting more complicated. Customers appreciate that level of foresight from a VAR."

Value Through Innovation

Approaching the server market with a platform perspective may help you build the most dependable systems possible. It might also help improve the trust relationships between you and your customers, compliments of fewer hardware issues and faster response to outages. But once you've covered the basics, you need a real ace to help set your offerings apart.

Your standard white box server might dish out plenty of value as it sits. Just think of the power-saving processor technologies, performance advantages of PCI Express, and backwards compatibility of SAS. But beyond those attributes inherent to many modern servers armed with enterprise hardware, it's also helpful to toss in the proverbial croutons that'll give your products more zest than competing solutions.

Take Intel's Active Airflow Control Technology as an example. All servers incorporate robust cooling fans designed to efficiently circulate air, enabling dual-core chips in tight 1U spaces. A majority operate at a constant speed, optimizing circulation at the expense of fan life, power consumption, and, most noticeably, noise. Consequently, many servers--especially those mounted in racks--must be tucked into closets or noise-dampening enclosures. A platform equipped with Active Airflow Control addresses the acoustic issue by monitoring ambient temperatures and adjusting fan speeds dynamically in response to unsafe environmental conditions. As temperatures cool off, the fans scale back.

Now here's where the platform focus starts yielding benefits beyond basic stability. Active Airflow Control, a chassis feature enabled through system temperature readings, is enhanced by DBS (Demand-Based Switching), an Intel processor technology intended to scale back on power consumption and heat output during periods of light load. The two work together, saving energy and minimizing acoustic output. According to Sammy Wong of Universal Systems, the pair also lay a foundation for denser racks and better power utilizations.

"With servers running at 100% around the clock, we can only populate 24U of a 42U rack. Intel's power optimizations, which include SpeedStep and Active Airflow Control, allow us to fill the whole enclosure without cooling or power delivery complications."

Drive Stabilization Technology, another Intel innovation, further illustrates the attractiveness of tapping a well-developed server infrastructure. The patented system secures each hard drive with a tensioned spring to prevent the imperceptible rattling of myriad disks in an enclosure, thereby improving mechanical reliability and augmenting data protection. When customers start investing in expensive storage subsystems, protective measures such as Drive Stabilization Technology become instrumental in maximizing ROI.

"Intel's Drive Stabilization Technology also has an effect on performance," Wong claims. "We found that in some cases, improperly securing a hard drive would cause a dip in disk throughput. Now it's basically impossible to mess up the installation."

Other value-adds built right into Intel's platforms make life easier for small consultants and off-site IT managers at larger organizations. The Rapid Recovery Toolkit, supported by the Intel Management Module, simplifies remote management even when a system is powered down. Exposure of a command-line interface gives administrators access to low-level functionality such as power cycling when problems with an operating system preclude a clean boot. That might not sound like much, but it might mean the difference between troubleshooting from your office and making an emergency site visit.

Click to see Server Technologies Product Matrix

Along the same lines, Intel's One Boot Update feature facilitates remote BIOS and firmware updates without taking a critical box offline. Depending on whom you talk with, server downtime costs businesses from hundreds of dollars per hour to hundreds of thousands of dollars. So even a 10-minute BIOS update hurts. One Boot Update instead applies updates while the server performs its duties. Changes take effect the next time it is deliberately restarted. Dense server clusters can also be updated through a scriptable command-line interface.

There are a couple of unique opportunities for VARs, as well. The primary purpose of Intel's Local Control Panel is to display monitoring and control options using the LCD integrated into compatible server chassis. However, resellers will find provisions for custom branding on the LCD. Additional customization is available through adaptive slots on Intel motherboards. Depending on the server chassis, VARs can choose between several combinations of PCI-X and PCI Express expansion. You'll only get one slot from a 1U form factor, but 2U chassis will take up to three--enough for RAID adapters and network cards.

Moreover, resellers often find themselves working extra hard for smaller companies with little or no in-house IT staff and employees who don't necessarily understand the technology. In those cases, it makes good sense to leverage these value-added extras. For example, Intel touts the Light Guided Diagnostics feature included on many of its platform kits. Special LEDs strategically positioned on certain server boards help troubleshoot problematic hardware.

"Before we even visit a remote client site, we can troubleshoot over the phone by having the customer check for illuminated LEDs," says Sammy Wong. "In many cases, Light Guided Diagnostics tells us exactly what went wrong. We can then get the customer running again in one visit."

The unique interaction between Intel processors, motherboards, mechanical chassis components, and specialized software really touch at the heart of why the platform message is so persuasive. It takes relatively little effort on the reseller's part to enable technologies missing even from many tier-one servers. In a market sorely lacking powerful differentiators, those technologies are compelling sales points.

Best of all, you can pick and choose the Intel technologies to deploy, right-sizing each build according to customer demands. That's the approach Pogo Linux's Erik Logan takes with his servers.

"Intel's server technologies help resellers differentiate their offerings, but we don't necessarily use them all. However, by mixing and matching the features where appropriate, we help ensure Pogo Linux products are optimized for each customer's needs."

Putting it All Together

In the world of servers, reliability is king. And while you might be challenged to deliver greater than 99% availability, certain best practices may help improve your batting average.

Taking a platform approach to server construction is the first step. Matching one vendor's processor with another's chipset on a third-party motherboard might be a great way to optimize for performance on the desktop, but that angle could prove problematic at the enterprise level. Instead, look to barebones kits as a means to leverage the validation efforts of larger vendors. You can go a step further by adopting Intel's branded platforms, which combine an Intel chipset with an Intel motherboard in an Intel chassis. You and your customers will both benefit from Intel's in-house Validation Stress Test suite--the heavy-duty qualification every server should go through.

"Because of Intel's historically platform-oriented approach, many enterprise customers will tend to go that route," says Pogo Linux's Erik Logan. "Validation often takes us less time on an Intel platform, improving our time to market, which is especially important for anticipated features like dual-core."

When it comes to mixing in value-added extras, Intel offers a veritable buffet from which you can pick and choose. So by deploying complete platforms, you have an opportunity to offer solid hardware, a support structure rivaling any tier-one, and the cool, little features that'll show your customers you mean business when it comes to high-end machines.

You can think of Intel's server technologies as value-adds that keep you competitive against tier-one vendors, make your life easier as a reseller, and save your customers money on support. Some of them are specific to Intel chassis and others are the domain of Intel motherboards. By combining the two, you can maximize the functionality of your white box servers.

The Server Technologies Product Matrix (page 50) should give a better idea of which boards and enclosures offer specific features. An "n" indicates full support, "s" means partial support, and "l" specifies that you'll need an optional upgrade controller in order to enable the feature. For more information on the server technologies, check out Intel's Web site (http://developer.intel.com/design/servers/technologies), where you'll find animated demos and posters.