A right-sizing revolution in on-demand network capacity is well underway, thanks to innovations in bandwidth usage monitoring, prioritization and throttling technologies.
Carriers like Broadwing Communications Inc. are pitching "Liquid Bandwidth" services that close the disconnect between the specific needs of a specific service or customer and the processes used to size network capacity for the job at hand.
Data center capacity appears long overdue for a similar right-sizing revolution as many network service providers (NSPs) enter the managed hosting business. Fortunately, a range of suppliers of data center design and management services are lining up to help the NSPs do the job right.
By some accounts, data center capacity management remains a work in progress, even in large enterprises where information technology (IT) departments have wrestled for decades with sizing server, storage and other capacities to support constantly evolving business processes and applications. These challenges are compounded by the emergence of far-reaching Internet, extranet and intranet projects. Nearly a third of IT spending on such projects now goes toward Internet infrastructure, according to Aberdeen Group Inc.
"It took a long time to get systems for right-sizing mainframe systems, but these [Internet infrastructures] are infinitely more open systems," says Tom Buiocchi, vice president of marketing for Peakstone Corp., whose eAssurance e-business capacity management system is designed to quantify and dynamically allocate application infrastructure capacity.
"All service providers should have a capacity planning system," Buiocchi says. "They're building a factory for their customers and somebody is going to suffer if factory capacity is guessed at, rather than truly known."
Such suffering is widespread, according to a survey of some 60 chief information officers released by Peakstone in August. Among respondents, 79 percent concede their data center capacity is "wrong-sized," and three-fourths of those admit they have engaged in expensive over-provisioning of servers and other infrastructure to hedge against uncertainty. Nearly half described a "less than fully predictable" IT support system, thanks to "a disconnect between IT infrastructure and business needs and objectives."
NSPs determined to provide web collocation services and higher order IT infrastructure and business application outsourcing, face the same dangers -- at a time when they are racing to launch managed services before their competitors beat them to it.
"Collocation companies can supply space and bandwidth, but they lack a complete set of operations tools and processes," says Sateesh Andra, cofounder and vice-president of strategy and business development of Euclid Inc., a startup provider of Internet operations lifecycle management services for service providers and their enterprise customers.
Andra says any service provider, enterprises or integrators determined to get to market fast with minimal capital expense should expect to see an initial $1 million price tag on testing and monitoring system development, network operations center (NOC) deployment and recruitment of staff.
As an alternative to pricey in-house initiatives, NSPs and other outsourcers can turn to their own outsourcers of Internet infrastructure management. Avasta Inc., Euclid, Loudcloud Inc., Totality Corp. and others offer turnkey planning, testing and management services for Internet-based infrastructure.
In some cases, hosting of the data center infrastructure itself is available. These vendors claim an Internet expertise lead over traditional mainframe outsourcers like Electronic Data Systems Corp. and IBM Corp.
Alternatively, more specialized Internet management service providers (MSPs), such as Keynote Systems Inc., Nuclio Inc., SiteLite Inc. and siteROCK Corp. offer discrete pieces of the capacity management puzzle -- in some cases to traditional outsourcers.
Dollars Per Square Foot
In IT parlance, a data center may describe a space full of servers, databases, storage devices, switches, load-balancers and the like. A true "production environment" involves much more.
New enterprise initiatives beget new Internet applications that must be stressed-tested and debugged for scale before deployment. The applications must be staged on servers, linked to distribution and access infrastructure and integrated with other applications.
After all that, systems must be applied for monitoring and managing application availability, performance and security.
Ideally, there is a unified administrative interface for real-time and historical trouble tickets and reporting on all the infrastructure components associated with each service.
"Most service providers face three risks in deploying operations management: time to market, money to build and acquisition of expert sales and support personnel," says Euclid's Andra. "Our model is to give the service provider a suite of operations services, a network to deliver those services, tools to host them, a portal through which to select and manage them and tools to make them available and supportable to end customers. Euclid provides the services transparently underneath."
For hosting companies and enterprises, the costs of Internet operations management must be balanced against revenue in a dollars-per-square-foot equation. If basic server collocation services tally $40 per month, Euclid says it can raise that number to $190 with its Trinity Testing Service and to $400 with the testing service and Trinity Managed Monitoring combined.
"Build and they will come has not worked with data centers," Andra says. "As an alternative, with managed Internet operations services, a 100,000-square-foot data center can produce the same revenue as a million-square-foot data center."
As for end-user benefits, these services promise more predictable Internet operations costs, he says, especially as enterprises going forward "also need to monitor Internet operations extended to partners and customers through sell-side and buy-side extranets."
Through Euclid's Trinity Portal, the host can define, size, price and chose to build a management service. Trinity services comprise an a la carte menu of testing services (including performance and scalability testing, vulnerability assessment and availability analysis); build services (including design, integration and deployment of infrastructure), and monitoring services (including availability, transaction performance, intrusion detection and Internet operations management). Results of deployed operations services circle back to the portal in the form of performance assurance alerts and reports.
Employing standard monitoring and management tools, such as ASPRE Inc. Hewlett Packard Co.'s OpenView, Mercury Interactive Corp.'s Topaz and Rational Software Corp.'s ClearCase, ClearQuest and Rose, along with its own "build analysis" tools, Euclid claims to monitor data center and network infrastructure end to end, from databases, applications servers and web servers to routers and access devices.
Pro-Act
This comprehensive view is mandatory, says Peakstone's Buiocchi. "You don't know how healthy your car is by checking the gas gauge, and it's the same with looking only at your web server processor usage," he says.
Yet Peakstone's survey found that a lack of integrated tools forces IT administrators to gin up their own kluge of performance metrics. "As important as this is, there has been no science to it," he says.
If the goal is to make that capacity as liquid as bandwidth, precise measures of data center capacity fill only half the equation, he adds.
"In case of congestion, would you like to assure quality of service for a buying customer by stealing capacity from a browsing customer?" Buiocchi says. He notes the larger opportunity may lie, not with customer-facing commerce, but with "intranet and extranet applications-users of a corporation's engineering portal get priority capacity over an employee checking his 401(k). Our two largest customers are Fortune 50 companies that don't sell online. They want us for enterprise resource planning and other corporate portal uses."
Indeed, says Andra, by automatically and proactively allocating capacity to different classes of users on the same platform, systems like Peakstone's eAssurance "could sit on top of our monitoring system to switch traffic on a needs basis."
While IT budgets are shrinking, service level requirements are not, so enterprise customers "will get increasingly educated and demanding in what they request," Buiocchi says. "Here's an opportunity for hard-hit service providers to bring bottomline impact to clients by lowering risk on right-sizing of infrastructure to fit real business needs."
| The Links |
Aberdeen Group Inc. www.aberdeen.com ASPRE Inc. www.aspre.com Avasta Inc. www.avasta.com Broadwing Communications Inc. www.broadwing.com Electronic Data Systems Corp. www.eds.com Euclid Inc. www.euclid.com Hewlett Packard Co. www.hp.com IBM Corp. www.ibm.com Keynote Systems Inc. www.keynote.com Loudcloud Inc. www.loudcloud.com Mercury Interactive Corp. www.merc-int.com Nuclio Inc. www.nuclio.com Peakstone Corp. www.peakstone.com Rational Software Corp. www.rational.com SiteLite Inc. www.sitelite.com siteROCK Corp. www.siterock.com Totality Corp. www.totality.com |
