The secrets to reining in server costs

By April Jacobs
Network World, 11/19/01

Jim Reese, chief operations engineer at Google.com, could be called the server doctor. He has about 10,000 servers to manage, and he also happens to be a physician, but it isn't the contents of a doctor's bag that keeps things up and running with only a dozen IT staff.

Reese's prescription for network executives struggling to reduce their budgets is a combination of good planning and a suite of automated server management tools. Others agree that users should automate as many management processes as they can via customized tools or off-the-shelf packages.

Observers say that blindly buying servers to keep up with demand or to handle new applications is no longer economically prudent; rather, users will have to determine how to handle more with less. This idea is bolstered by the fact that market research firm IDC predicts server spending in the U.S. will sink from $25.7 billion in 2000 to $21.8 billion in 2002.

Google uses a variety of homegrown tools built with Python and Perl scripts to automate server management and synchronization. Python and Perl are highly portable scripting languages. Instead of using a hands-on approach to checking server health and updating software, the IT staff lets the scripting tools do the work.

Google's management tools include some that allow for software synchronization. If Reese needs to install new code on servers en masse, the tools handle it for him. Another tool set confirms that each server not only accepted the code, but also that it's up and running.

"We have statistics on just about every facet of every server including the load it's carrying, how many queries per second it's receiving and the amount of network bandwidth it's using," he says. With that kind of extensive monitoring, he says, network executives can pinpoint if there is an isolated problem or an ongoing problem with their servers.

Catching problems early can help trim maintenance budgets. And while Google uses homegrown tools, others are turning to off-the-shelf products, such as Microsoft Application Center, TurboLinux's EnFusion and PowerCockpit cluster and server management products, and products from server vendors such as Dell's OpenManage software.

OpenManage helps your network managers anticipate problems before they occur, track IT assets and reduce management costs. The software features tools that provide network managers with event filtering and problem alerts, centralized inventory and asset information, and remote server management. The package also monitors storage arrays including volume, disk, hot spare and RAID-level controls.

For those in the Windows world, Microsoft Application Center monitors hardware conditions.

"We use Microsoft Application Center to make the management of our server farm a lot easier," says Bob Janacek, co-founder and executive vice president at CertifiedMail, a secure messaging company. He says Application Center simplifies the process of making changes to his server clusters.

"We update a file and it's on seven machines. Something that would take one hour per machine - that could take an entire day - now it only takes one hour, and it does it right every time," Janacek says.

Application Center can also let network managers set rules that provide for an automatic reboot if a failure occurs. "You don't want to have to baby-sit your platforms," he says.

Get the most out of servers Here are some server management tips from Vernon Turner, IDC analyst:
1.	Reconsider server consoli-dation as a serious invest-ment and create a contain-ment program to reduce overall site complexity.	4.	Never keep server technol-ogy more than three years. Negotiate technology upgrade plans through leases or aggressive trade-in plans to avoid mainten-ance costs.
2.	Standardize on two vendor platforms and let both know they are constantly compet-ing for site dominance. This will dramatically reduce the number of variables and single points of failures.	5.	Invest heavily in automation software to manage the hardware (from storage management and capacity performance.)
3.	Keep a third vendor in the ‘lab’ so it’s clear you are always looking out for better technology. Usually that third vendor technology will be on loan, at no expense.

There are also other hardware advances taking place in servers - all aimed at helping users manage systems more efficiently. IBM's Project eLiza is aimed at automating system and network management. The idea is to centralize and automate a lot of the management tasks your network managers now spend time performing manually.

As part of its eLiza initiative, IBM recently introduced its e-business Management Services program, which essentially gives users a software-based dashboard view of their critical business systems, including applications and hardware. The company also plans to bring down some of the resource provisioning technology it offers on its high-end servers to its low-end eServer line.

IDC analyst John Humphreys predicts users will also see other major server vendors come out with offerings featuring resource provisioning on low-end servers. Others, such as Dell, Compaq, Sun and Hewlett-Packard, are all working to bring more intelligence to their servers.

In the Linux arena, there are offerings from TurboLinux, including its EnFusion 7.0 clustering software that lets users cluster Linux, Unix and Windows servers so that they function as a virtual single server. The company also recently announced PowerCockpit, software that lets network managers capture an image of a server and its associated operating system and applications. They can then use that image to reassign tasks to servers when needed.