Itâ€™s a good time to unlock the padlock on your IT budget–but donâ€™t let recovery euphoria get the best of your better judgment.
This year we are finally starting to see some good news in the area of IT spending. After a few years of declining budgets, 2004 is starting to look like a banner year.
In December, IDC raised its estimated 2003 sales of PCs, servers and notebooks to 152.6 million units, an 11.4 percent increase over the previous year and an 8.8 percent increase over the year 2000’s record numbers. 2004 is looking even better. AMR Research’s quarterly IT survey of firms with 1000 or more employees found IT investments growing by 4.3 percent in Q4 2003 and indications that the trend would continue into 2004. A October survey of over 600 IT buyers conducted by Gartner, Inc. and SoundView Technology Group, Inc. discovered that capital spending budgets would grow by 1.6 percent in 2004.
But just because budgets are increasing, that doesn’t mean that we will be seeing a repeat of the late nineties. A good portion of spending will need to go to replace aging hardware bought during the Y2K scare era. Purchases will be closely scrutinized to determine business need, rather than being driven by new technology. So, while the Gartner/SoundView survey mentioned above says that budgets will rise, Gartner’s Technology Demand Index shows that companies are still being cautious in their spending and are spending less than their budgeted amounts for IT.
ROI worksheets are one key way to justify technology expenditures. But often those numbers seem like they were just pulled out of the air. What is the real financial value to a company of getting a file to a users screen a few milliseconds faster? But there is another set of tools, capacity planning software, that can help IT managers economically plan their expenditures.
Capacity planning involves two basic disciplines. The first is accurately determining current resource utilization. The second is predicting what resources the organization will need at some point in the future.
In the first place, tracking current usage helps to avoid costly overprovisioning. While many companies once used overprovisioning to be ensure that they wouldn’t run short in the future, that is no longer necessary. Vendors are much more flexible in offering hardware in smaller increments. For example, IBM ships some high-end servers with extra processors. If the customer doesn’t use them, it doesn’t pay for them. But, if it does need the extra processing power in a hurry, it can just execute a command to activate the idle processors. Blade servers are another way to quickly scale capacity as the need arises. This even applies to computing environment itself. American Power Conversion Corp. (West Kingston, R.I.) has a infrastructure product line – including UPSes, generators, rack power and air conditioning – which scales in increments from a 2KW IT room up to a multi-megawatt data center.
“This eliminates the need for companies to oversize their equipment since they can add capacity later as needed ,” says Michael Proffit, senior product manager for APC’s enterprise systems and services group. “Working this way reduces the overall capital expense.”
There are also technology upgrades which, while attractive, aren’t necessarily called for given actual usage patterns. For example, vendors are pushing for companies to move gigabit Ethernet out to the desktops. Although doing this does give plenty of room for future growth, Garner analyst Serverine Real points out that few people actually use even a tenth that capacity. She says that if companies wait till the end of 2005 before deploying it, their costs will be considerably lower.
There is one other area where capacity planning is extremely useful: reallocating resources. Storage and computing virtualization technologies help to reallocate resources as needed, but these technologies are far from ubiquitous. It is more common to see a server assigned to a particular application, storage allocated to certain users and so on. In this type of setup, some of these will be underutilized while others are operating at capacity. In such cases, companies can attain a huge cost and time savings by reassigning resources rather than purchasing more.
Most management software vendors incorporate capacity planning modules into their product suites. IBM has the Tivoli Storage Resource Manager which provides reporting, planning and provisioning functions across heterogenous environments containing a mix of direct attached storage, network attached storage and storage area networks. Computer Associates’ Unicenter NeuMICS Resource Management Capacity Planner Option uses utilization trending to predict when additional resources will be needed. It also as allows managers to build statistical models based on these historical trends so that they can see how changes in number of personnel, or network upgrades, will affect service levels.
There are also smaller firms which specialize in the capacity planning arena. Clear Lake, Iowa-based TeamQuest Corporation has a product called TeamQuest Model which takes information from its own performance monitoring software and uses it to generate resource trending information. Managers can also use the product to model capacity and performance issues based on growth trends or the impact of a single event. It works with Windows and Unix (HP-UX, AIX, Solaris, IRIX). A version is also available which takes its data from Hewlett-Packard’s OpenView Performance.
Another firm offering capacity planning products is The Information Systems Manager, Inc. of Bethlehem, Pennsylvania. It has modeling suites for IBM’s OS/390, Windows and Unix/Linux. These suites incorporate the company’s PerfMan Analyst reporting and analysis tools, and add capacity planning and modeling functions.
A Lower-cost approach
Modeling tools give an IT manager a good way to predict future needs and to restructure workloads throughout the network. But there are drawbacks. As with any piece of software, there is the cost of the capacity planning package, its drag on system resources and the time spent installing, learning and operating the software. The return is great enough to make it worthwhile for large enterprises, but smaller entities need simpler, cheaper tools. One approach is to use network monitoring and management software to generate the necessary historical information and then take this information to adjust current traffic loads or plan future purchases.
That is the route taken by Sheridan Press, one of the seven companies composing The Sheridan Group of printing companies.
“We had issues where certain elements were becoming overutilized or saturated,” says Benjamin Natal, Sheridan’s network engineer. “We needed to implement a capacity planning process covering our critical systems–key servers, network segments, routers, switches, T-1 links – that would let us be proactive in managing our resources.”
Located in Hanover, Pa., Sheridan primarily uses Windows on its desktops and servers. But, being in the graphics business, about a quarter of its workstations consist of Macs for the production departments. Six Unix servers are also part of the network since a major printing application runs more efficiently on Solaris than on Windows.
Sheridan was already using Ipswitch, Inc.’s (Lexington, Mass.) WhatsUp Gold monitoring software, but needed something that would provide the trending and reports needed for planning. A team was formed with members from the companies that make up The Sheridan Group to evaluate the options. Natal reports that they compared eight different products, putting the results into a spreadsheet. In the end, they selected Denika performance trender from Sanford, Maine-based Somix Technologies, Inc. which costs $1,090, including a year of support.
“The other products were extremely expensive,” says Natal. “Denika was very cost effective in comparison to the others, yet it could do what the others could do.”
Besides the price, other features affected the decision. One was the way that Denika tightly integrates with the WhatsUp Gold they already had installed. Natal says they also liked the ability create custom scripts and applications using Perl and Visual Basic scripting. In addition, it was easier to set up.
“The beauty of Denika is that it is agentless,” he explains. “You can have it up and running in a couple hours, not days.”
One problem he has been able to resolve using it concerns a critical server that has a 169GB volume of direct attached storage. The data in that volume is highly dynamic, and the server kept running out of space. As the volume filled up, it would shut down the server. Rather than having to add more storage capacity to solve this problem , Sheridan set up Denika to monitor thresholds on that volume to alert the IT staff to clean it up before the processes grind to a halt.
In addition to keeping an eye on the hard drives, Natal also uses it to identify and resolve network bottlenecks. But these are extras. Sheridan bought Denika to use for capacity planning, and uses it for that purpose. By keeping watch on utilization trends, they can see where use is growing and make plans to obtain the needed resources before it starts causing problems for the users.
“If we see high volume on certain machines or certain network segments, we can immediately plan to upgrade or replace the systems,” says Natal. “It’s been very beneficial to have that kind of capacity planning.”
Dick Dimock is a technical writer from North Hollywood, California.