How to sell servers to SMBs

The battle for server market share in small to medium business has always been fought on speeds and feeds. One vendor will play a trump set of benchmarks until another vendor's next hardware refresh. 

However, there are signs that vendors may focus less on who has the faster box than who presents the most comprehensive, well-integrated approach to designing an IT environment. The right server to fit an SMB now might depend on who has the best "ecosystem".

The most expansive interpretation of the ecosystem approach brings storage and networking under the same management umbrella as processing power. While this level of integration is more useful to the enterprise, server vendors are keen to sell the benefits of simpler management to smaller companies too.

Making servers and other IT infrastructure easier to manage means SMBs can avoid paying for engineering certifications for their IT staff - or having to call in expensive IT support for basic moves and changes.

Resellers chasing the ecosystem sale need to take a more customised approach than the one server, one application approach of the past. Sales strategies need to appeal as solutions to business problems rather than a straightforward server upgrade.

The payoff is the likelihood of a more heavily featured (and more profitable) server sale.
Analysts say the ecosystem approach will favour tier-one vendors that have invested in headline alliances, such as Microsoft and HP; or Cisco, VMware and EMC.

"A system builder can manufacture a system to work as well as a tier-one OEM can, but they may not have connections to the same enterprise technology," says Errol Rasit, principal analyst for data centre systems at Gartner.

"The big tier-one vendors are chasing the same target which is a fabric-style computing. This does resonate with a lot of large business today and it will filter through to small businesses in the future."

Aspects of fabric computing are already passing down the food chain.

For example, look at the takeup by small business of the blade server, an enterprise innovation which packs processor, memory, I/O and program storage into a single unit.

IBM and HP have released small form-factor chassis that plug straight into a 240-volt wall socket without a separate power supply. IBM's model in this size, the Bladecenter S, is marketed with the line, "Big benefits for the small office".

While not every SMB will be ready for blades, every business can benefit from one focus of fabric computing - easier management.

Undoing the hassles of admin
Many tier-one vendors are focusing on reducing the cost of ownership by improving ease of management through the single-vendor, ecosystem approach.

"Customers aren't just after a faster server, they're after a better ecosystem. We could just deliver the next processor and say, ‘Here it is, a bit shinier and faster'. But if it's not integrated and easier for a partner to sell then we are going to be wanting," says HP's marketing manager Raymond Maisano.

The goal of integrated components is to make it less complex to deliver leading edge technology, says Maisano. "We believe management is at the core of delivering service to the customer. Rather than having to spend more time learning, we are driving up the utilisation of what they've got," says Maisano.

Management is of course easier if all your network components come from the same vendor. HP and to a lesser extent Dell are encouraging their partners to sell their own servers, storage and networking.

Not only is it possible to have a single management tool for the network, it makes troubleshooting easier in theory because if something doesn't work, there's only one support phone number to call.

Cisco has tackled the issue of support through its alliance with VMware and EMC. There is a single support number for customers to call for the three vendors' products.

Cisco's server strategy is to simplify the architecture by running a unified network - called Unified Computing System (UCS) - to lower the cost and complexity from a networking perspective.

While not a strong point in the past, Cisco has received plaudits for the single-pane management interface built into its UCS environment.

The vendor is extending these benefits down the foodchain. By next quarter an SMB customer will be able to add a rack server to a blade environment and have it appear as an additional blade, says Dylan Morison, Cisco's data centre regional manager.

Dell is working on a similar theme with its server management console, based on Symantec's Altiris management software.

The console is integrated with virtualisation software from Microsoft, VMware and Citrix.

"It comes as a part of the server and gives you an engine to manage networking, storage and helpdesk," says Jon McBride, Dell's manager of SMB  enterprise solutions.

Another management feature is the lifecycle controller - a solid state drive that contains drivers to get a server up and running quickly. The controller connects back to Dell and uploads the latest versions. "It saves a lot of time putting CDs and DVDs in a slot and waiting for something to load up," says McBride. Dell is witnessing increased sales in appliances - virtual and physical.

Devices with pre-installed software which has been hardened and optimised at the factory dramatically cut down install times, says McBride.

KACE is one appliance which fits neatly into the renewed emphasis on easier administration. KACE is a systems management tool which gives IT managers a view of hardware and software from any vendor on a network. KACE helps SMBs manage their inventories, simplifying patching and eliminating unused software licences.

In its appliance form KACE can be up and running in a day, says Dell, compared to weeks for other products.
Management is becoming more important because servers are expected to do a lot more than in the past.
Dedicating a box to a single application is becoming less common in multi-server sites thanks to virtualisation, which is reinventing IT - and server buying patterns - for SMBs.

Everybody get virtual
While virtualising business applications may not seem a worthy goal in itself, a business owner is hardly likely to ignore the benefits if they are explained in simple business terms.

Concepts such as business continuity, disaster recovery and automated provisioning are understood by CIOs but are confusing jargon to small businesses, even though they share similar needs.

Framing these concepts as business solutions puts them in terms an SMB can understand. Would you like to reduce the chance of losing data through failing hard drives? Would you like faster, easier, multiple backups - or snapshots of your information throughout the day?

Would you like to reduce the effect of server crashes on critical applications such as email or CRM?
These propositions make sense to any business.

The take-up of virtualisation in Australia is roughly double that of the rest of the world, according to Gartner.

Peter Hedges, IBM's business development manager for System X servers, has seen anecdotal evidence of this trend from IBM's sales. As the man responsible for value solutions within IBM Australia and New Zealand, Hedges has contact with all IBM's sales reps.

"More and more we get these sorts of questions and solutions" about virtualisation, says Hedges.

"You can always tell [when a customer is rolling out virtualisation] because whenever there's a question about how a vendor licences their software that's a clear indication of what the customer wants to do."

Feature sets on entry-level servers have come to include enterprise-level capabilities in response to the virtualisation trend, says Hedges. This reflects the greater risk in losing a server. When a conventional server crashes, a company might lose a single application.

If a virtualised server goes offline, it could take out 10 or 20 applications with it.

"With virtualisation the impact of something going wrong is multiplied. Risk is the product not only of the probability of something going wrong but the impact of something going wrong," says Hedges. IBM and other vendors are designing servers to minimise that risk.

Newer servers have more features, higher specc'ed configurations and smarter management software - and higher margins.

Virtualisation requires more stable servers to handle the intensive workloads of running multiple virtual machines. Analysts quote the average usage of a traditional server processor at 18 percent; a virtualised server is closer to 80 percent.

Just as important is memory. The amount of memory plays a large role in determining how many virtual machines can be run on a server. It is important to know how many virtual machines your customers intend to run; and businesses tend to have higher counts of virtual machines than the servers they replaced because VMs are so much easier and cheaper to deploy.

Hedges points out IBM‘s entry-level X3200 supports up to 32GB of memory, more than double some competitors' equivalent products. "To run the same number of virtual machines [on competitors' servers] you need twice the number of hardware," he says.

Most server crashes are never diagnosed, but statistical data points to memory errors as a likely culprit. In the 1980s, computer systems had a couple of megabytes of memory and used parity protection, which could detect if a single bit went wrong and stopped the machine.

Now an entry-level sever can hold 32GB and support several virtualised workloads. You really need to provide the highest level of memory protection, says Hedges. Memory limits increase with the number of sockets, and servers with more sockets are sold on average with higher amounts of memory.

This is reflected in the average spend per server, which is heavily inflated by hikes in memory according to the number of cores. The average single-socket server costs US$1,600; a dual-socket is more than double at US$4,500; and a four-socket more than triple at US$14,900, according to Gartner.

Crash me not
While virtualisation software vendors have developed methods for improving uptime, these should not be relied upon, says IBM's Hedges.

One high-availability feature on IBM's entry-level X3200 server is the use of RDIMM memory, which claims to have better error correction and therefore greater reliability than the standard UDIMM memory.

"It means we have less opportunity for a data error to corrupt the system or the application in a virtualised world," says Hedges.

"Knowing the difference between memory based on RDIMM versus UDIMM can give an edge for a reseller, given that this is such a crowded market."

While slotting in the best memory for the job can lower risk, it comes at a cost. Prices vary with the size and type of memory. Cisco's memory extension lets an SMB use low-cost DIMMs so they appear as a bigger unit. For example, four 2GB DIMMs can look like an 8GB DIMM, which is 20-30 percent more expensive.

Another availability feature is predictive failure analysis. The server constantly monitors the health of major subsystems such as fans, processors and memory. IBM claims to monitor twice the number of subsystems as other vendors, which helps a server predict outages before they occur.

The "chip kill" feature isolates faulty chips in memory, so if a memory stick loses a whole chip the server will continue operating without memory loss or a crash.

IBM released a product towards end of last year called VM Control which provides a common management interface across Power VM, Hyper V and VMware. VM Control gives the customer the ability to create and manage virtual machines, says Andre Liem, who looks after IBM's RISC-based Power platform.

"Once a customer has confidence with virtualisation, the next step is to automate production of images. The third step is to optimise and build resource pools of processor and memory and storage. So all of a sudden you have a customer that is able to create their own mini-cloud," says Liem.

"At the smaller end there can be very, very little onboard technical skill within a client," says Hedges. "Virtualisation is a great way for reseller to build up the services alongside the hardware."

Servers capable of supporting virtual machines need to be much more capable, and this is reflected in the sales figures, says Liem. While the number of systems year on year is tending to flatten off, server revenues are continuing to improve, which indicates that each system is being sold in a richer configuration, he says.

Bigger, smarter servers can handle more complex applications - such as desktop virtualisation. This technology is becoming more attractive with the increase in hardware power, says Hedges.

"I think that 2010 is going to be a very interesting year" for desktop virtualisation, he says.

Working the cross-sell

Virtualising servers can lead to lucrative opportunities in other areas of the business, as the technology places higher demands on storage and networking devices. It also means resellers must change the sales pitch from a standard server upgrade.

Storage and networking ports as well as processing power need enough headroom to avoid becoming a bottleneck. However, it is not enough to just upgrade the capacity - customers also want them to be smarter too.

"As customers start to deploy virtualisation more they see they have to virtualise networks, storage and IO. And in that sense the channels and the go-to-market and the direct sales have to change from selling products on price and performance to selling solutions," says Gartner's Rasit.

McBride says SMBs are choosing virtualisation for scalability, protection and agility. But there is an expectation that they want to do the same thing with their data in the backend, which leads to upgrades in storage too.

"You want a storage platform that's going to give you the same level of scalability and agility as you get on the front end with your server virtualisation," says McBride.

"You get failover, high availability, the ability to do planned move of servers during business hours. To get a lot of those features in virtualisation you do need networked storage in the backend."

Buying patterns in small business

What's the right decision when choosing a first server?

Most micro businesses shanghai a desktop PC into the role of makeshift server. As a small business adds staff and IT overheads increase, the business owner is faced with a difficult decision. The right upgrade depends on how fast the business is growing.

The business owner with modest expectations and a budget to match might decide the most sensible course of action is to shell out for a single-socket system powered by Intel's Dual Core or Core i5 or i7 processor.
The performance gain might be modest but to the owner he or she now has a system that vendors call a server.

This lowers the barrier to entry for a fully fledged server and is a common intermediary step, says Gartner's Errol Rasit.

"We do consistently see that there are desktop branded CPUs going into towers and racks and customers deploying that. It's the smallest step up from a desktop acting as a server, and they come in at a similar price point as well," says Rasit.

Money is not the only restricting factor. There is a significant skills difference between maintaining a tower server and a rack-mountable server. For example, the former can plug straight into the wall socket while the latter can require a dedicated power supply.

However, if the business grows quickly and is forced to upgrade within the useful life of that single-socket server, the business owner has made the wrong decision.

A dual-socket system costs less than two single-socket servers yet can often deliver greater power and a wider range of features, including enterprise-class options that benefit businesses of all sizes.

The main volume for all vendors is in rack-mounted, two unit, two-socket servers. Sales of blade servers are also growing, which are easier to manage and more energy efficient.

Another attraction is the reduction in cabling. The back of many server cabinets is a mess of Ethernet, fibre channel and power cables. Upgrading to blade servers can untangle spaghetti set-ups.

Still, blade servers are not for every small business. "You have to find the appropriate place for it.
We don't try to move all customers to blade," says Dell's Jon McBride.

Power efficiency and processor speeds remain strong selling points, especially given the fact that 40 percent of servers in use by businesses are single core or first-generation dual-core servers, according to Intel.

The latest multi-core servers, such as HP's G7 server, offer a 20 to one increase in performance. Power management techniques now trace where each watt is spent, spinning up fans only if needed.
HP claims its G7 is 96 percent more efficient than the G4 it replaces, which was released in 2004.

"When Intel released the Westmere processor they talked about an eight month ROI. From our G4 servers to the G7 you have a two month ROI," says HP's Raymond Maisano.

Matching the right size server to a business requires understanding its plans for growth. If fast growth is on the cards, it is wiser to spend more on IT, earlier.