About Network Testing Labs
Contact Network Testing Labs
Down on the Farm
migrated from centrally-managed mainframes to distributed
workgroup-managed servers. Then you reclaimed the servers from the
workgroups to create centrally-managed server farms. How are other
companies coping with these disruptive, cyclic paradigm shifts?
By Barry Nance
computing promised to make massive, monolithic data centers obsolete.
As companies distributed workgroup servers and their management across
the organization, however, they discovered a pent-up need for both
computing power and storage capacity far beyond what the companies had
envisioned as a role for small- to mid-range servers. As a result of
the dramatic increase in the number of servers and the corresponding
management workload, many enterprises are in the process of reclaiming
(or have already reclaimed) server management as a centrally-
administered function. Some have even moved the servers themselves to a
central location, but most use network-based tools to remotely
administer the dozens, hundreds or sometimes thousands of servers.
We’ve found that companies with centrally-managed server farms tend their fields of file, database, Web and application processors using very individual, carefully-crafted approaches that combine a diverse set of commercially-available tools with a smattering of custom-programmed workflow automation. The software tools are different and the user interfaces are graphical, but the server administrator’s management chores are essentially the same ones a mainframe administrator performs. The key difference is the greater number of machines the server administrator has to keep running.
|We also found many
companies are generally reluctant to talk about their server farm
management techniques and tools because they’re afraid that describing
the server environment and management procedures will divulge important
internal company structures and competitive business strategies. On the
one hand, this attitude is a sign of the vital role IT has come to play
in enterprises today. On the other, however, it shows a measure of
shortsightedness on the part of IT professionals who ought to be
willing to improve the industry by share their knowledge with others.
Fortunately, we located a few forward-thinking IT professionals who
understand the need to let others know of their server management
The profiles that follow reveal and explore four enterprises’ experiences with multiple server management. They show that the right mix of tools and procedures is different for different organizations, but having the right tools and the right expertise in using those tools is universally important.
James Governor, an analyst with Illuminata, Inc., said it best. “As IT tries to keep pace with the exponentially-increasing need for computing power, companies that stay on top of the management of that computing power will undoubtedly fare better than those that don’t.”
|Organization: Origin Technology in
Business, Inc. (subsidiary of the Philips Electronics company), Dallas,
Netherlands and Singapore
Assignment: Provide global outsourcing services to enterprises.
Solution: Many different kinds of UNIX and Windows servers managed primarily by Tivoli Systems’ Enterprise suite of tools
Origin manages 22 huge server farms not only on behalf of its parent company, Philips, but for hundreds of outsourcing clients with hundreds of thousands of users in 31 countries. The company’s data centers house nearly 60 mainframes and about 3,000 mid-range servers. These servers are a mixture of about 1,800 IBM AIX, Sun Microsystems Solaris, Hewlett-Packard HP-UX, DEC UNIX and IBM OS/400 machines along with about 1,200 Windows Server and NetWare machines. More than 400 of the computers are Lotus Notes servers, many are file, print and database servers, while the remainder are application-specific processors for ERP and other vertical market systems.
Origin employs just over 60 highly-skilled people, located at seven global sites, to manage the servers.
|Keith Pelphrey and
Mark Eimer of Origin’s Enterprise Management Systems (EMS) branch
explain how they’ve achieved this remarkable level of efficiency and
productivity. “We work hard to make sure the computing environment is
consistent and up-to-date, and we adhere to stringent standards - that
we created for ourselves - for how we run our servers.” Eimer ruefully
describes EMS as challenging, and says his people are always very busy.
Pelphrey indicates that businesses having trouble maintaining a
consistent, standard environment are good candidates for becoming
Pelphrey also says that, while Origin certainly doesn’t throw away a new customers’ existing procedures and tools as it absorbs the maintenance of that customer’s technologies, it does prefer to use Tivoli Systems’ Enterprise suite to manage multiple servers. To a lesser extent, Origin uses Computer Associates’ UniCenter TNG and Hewlett-Packard OpenView. Origin centrally controls the server management software from its mainframe computers, using Origin-written programs. IBM NetView and NetScout Manager Plus help keep the server’s network connections healthy, and for backup/restore Origin uses Tivoli’s Storage Manager (formerly ADSM), Legato, Seagate ArcServe and Hewlett-Packard OmniBack.
Both Pelphrey and Eimer were emphatic in their advice to other server farm managers. “People underestimate the complexity of EMS products. To make an EMS tool do what you want, you need to invest significantly in the training of the people who will live with the tools.”
|Organization: Bell Atlantic Yellow
Pages, Inc.(subsidiary of the Bell Atlantice phone company), Boston
Assignment: Publish yellow pages directories in print and on the Web
Solution: Compaq servers running Windows Server with IBM, Hewlett-Packard and Sun Microsystems servers running UNIX, managed by Compaq Insite Manager, Microsoft SMS, Tivoli Expert Advisor, Adkins Resource Hyena Wise Solutions “Wise” installer and Seagate Software ArcServe 6.6
Bell Atlantic’s yellow pages subsidiary manages just under 200 servers, scattered from Maine to West Virginia in 50 divisional offices and 6 data centers. Most of the machines are Compaq Proliant servers running Windows Server, but about 20 run IBM AIX, Sun Microsystems Solaris and Hewlett-Packard HP-UX flavors of UNIX. There’s also an MVS mainframe in the mix. The servers support about 5,000 end users, all in a single Windows domain regulated by one Primary Domain Controller. Microsoft’s RAS runs on 2 servers to provide 150 ports of dial-in connectivity.
The publishing company’s mainstay applications are Microsoft Office Word for Windows and Excel productivity tools, Microsoft Back Office database (SQL Server) and connectivity (SNA Server) tools, Lotus Notes for e-mail, PeopleSoft, Quark Express for page layout and a custom-written system for processing yellow pages advertising orders.
Five people manage the Bell Atlantic subsidiary’s servers.
|John Farrer says his
organization has carefully and thoroughly evaluated dozens of software
tools to come up with the best combination for his team to use. The
group’s criteria for server management tools include sophisticated
functionality, support for the company’s predominant platforms, ease of
use and an intuitive interface.
Farrer says he relies on Microsoft SMS for systems management and software distribution. Tivoli Systems’ Expert Advisor automates the helpdesk, while Adkins Resource’s Hyena maintains user IDs, home directories and standard resource sharing configurations for users. Wise Solutions’ Wise installation tool helps Farrer build new correctly-configured Windows Server machines, and Seagate’s ArcServe does the backup/restore data protection chores.
He’s a firm believer in staying ahead of hardware failures. For example, each Intel-based server has two RAID disk controllers capable of failing over to the other, redundant network adapters and redundant power supplies.
Farrer’s advice to other server farm managers: “Spend the extra money upfront to get fault-resistant, redundant hardware. Buy the best equipment possible. You’ll save money over the long term because the cost of server downtime, such as an unavailable production system, is high. You’ll also get longer lifecycles with better hardware.”
|Organization: The Forzani Group Ltd.
Assignment: Support the largest retailer of sporting goods in Canada.
Solution: Data General DG-UX servers and Windows-based clients, managed by Hewlett-Packard OpenView, 3Com Transcend and Intel LanDesk
The Forzani Group Ltd. is Canada’s largest sporting goods retailer, with about 300 stores nationwide. It’s growing rapidly and expects to soon have hundreds of geographically-dispersed Data General DG-UX servers that it will manage from its main office. About 800 users will access these machines, a few of which are NetWare file servers. The company’s principal applications are a sophisticated point of sale system, a sales data tracking and trend analysis system and a business-to-business EDI-based interface system.
Three Forzani Group employees will manage the servers, but this number might change because the company is changing its organization by merging its helpdesk and network management responsibilities.
|IT Director Michael
Flood, who’s orchestrating the server growth plan, describes it as an
offensive strategy to deal with his rapidly expanding server farm. He
says his goal is to smoothly integrate appropriate new technologies as
his group rolls out the servers and management tools. Canada’s
geography is one of his challenges, and Flood’s also identified time
zones, multiple native languages, different operating procedures in
stores and server upgrade paths as obstacles he’ll need to overcome. He
also wants each store to be as fault-tolerant as possible, insulated
from server and network failures.
He’s selected Hewlett-Packard OpenView, 3Com Transcend and Intel LANDesk as his primary tools. He looked briefly at Tivoli Systems’ software, but found it pricey and unable to handle NetWare to UNIX migrations. Legato software carries out the backup/restore work.
Flood’s encounters with software vendors’ salespeople have given him a healthy skepticism regarding their claims, and he passes on this advice: “Don’t believe vendors when they tell you they can automate all management tasks for all servers.”
Administrative Services, Inc. (subsidiary of Policy Management Systems
Assignment: Administer life insurance products on behalf of major insurance companies
Solution: IBM AIX servers with NetWare and Windows on Compaq, Dell, NEC and clone machines, Novell NWAdmin, Microsoft User Manager for Domains, Syncsort Backup Express and Novell ZENWorks
Financial Administrative Services is a third-party administrator that subcontracts to life insurance companies wanting to outsource the customer data maintenance and customer service functions associated with a particular kind of insurance, such as Variable Annuity Life. The company has only a few dozen servers, accessed by about 700 users. However, the diversity of application environments the company has to support as it assumes the identities and workloads of several different insurance companies puts a strain on the server management people. On behalf of those insurance companies, Financial Administrative Services runs NetWare, Windows Server, WinFrame, AIX and a little Red Hat Linux in its garden of servers. The Linux machines act as FTP servers, while the Windows Server, WinFrame and AIX servers store customer insurance data and process customer transactions.
Six people manage the servers.
|Scott Wiggin says his
biggest problems have been finding tape backup software (Backup
Express) that would work with all the company’s operating platforms and
then getting the vendor (Syncsort) to fix bugs in Backup Express. He’s
not concerned that it takes six people to manage the servers because
the people spend much of their time responding to ad hoc requests to
move data from one operating environment to another, a task he hasn’t
found a way to completely automate.
Wiggin’s group uses NWAdmin to administer a NetWare NDS tree and the Windows User Manager for Domains program to administer the company’s two domains. He’s looked at Novell’s NDS for Windows product and concluded it isn’t capable enough for his group’s purposes. Novell’s ZENWorks distributes computer programs and data files.
Like Origin’s Pelphrey and Eimer, he suggests to other server farm managers who have heterogeneous operating platforms to support that cross training is a sine qua non for keeping servers up and running.
|The More Things Change, the More They Stay
Data center administrators and server farm managers perform similar tasks, such as updating user records, setting permissions, dealing with hardware failures, diagnosing connectivity problems, monitoring processor and storage capacities and of course running batch programs according to a preset schedule.
However, the person responsible for the servers has to worry about keeping a greater number of computers running, propagating changes across more machines and dealing with different operating systems. Furthermore, a data center administrator typically uses a set of documented corporate standard procedures for his or her updates, while the server farm manager’s change procedures are often less formal.
The data center’s MVS is more sophisticated and mature than the operating systems found on small- to mid-range servers (although, ironically, Windows Server consumes more memory than MVS). For instance, MVS easily detects and discards runaway applications, and it gracefully handles hardware failures and disk space exhaustion.
A server farm manager’s job description may look like his data center counterpart’s, but don’t be misled. The server farm manager has the tougher job.
About Network Testing Labs
Contact Network Testing Labs