Network Integration: Keeping Traffic in the Fast Lane

There’s not a lot of glory to be had in network integration, just grief from the lack of it.

When Inergy Automotive Systems began a series of network improvements, it measured success in terms of performance enhancements and cost reductions. But David Stephens, Inergy’s vice president of information systems and services, measures success just as much in the reduction of the complaints he hears at management meetings about slow or unresponsive applications.

Among other problems, network congestion was slowing access to the firm’s core SAP business applications at the end of the month, when the finance department was closing the books.

“As our CEO says, ‘The best thing I hear about our I.T. networks is nothing,'” Stephens points out. So, when chief executive Pierre Lecocq took a few minutes during a planning meeting earlier this year to note that he hadn’t heard any complaints lately, Stephens took that as high praise.

What changed? Part of the reason for the inefficiency of Inergy’s network was fragmentation. The Paris-based company, one of the world’s biggest manufacturers of plastic motor-vehicle fuel tanks with 2005 revenue of 1.3 billion euros (about $1.6 billion), was formed in 2000 as a joint venture of Plastic Omnium and Solvay SA. The new organization inherited systems and networks from both companies that had never been intended to work with each other-—a common dilemma not only for joint ventures but also for companies that grow through mergers and acquisitions. In addition, Inergy found itself with 40 sites in 13 countries, each with its own local networks and user directories and locally installed applications.

“I don’t want to say it was a mess, but it was very complex,” says Arun DeSouza, Inergy’s manager of systems engineering and information security.

When the joint venture was formed, Inergy limited itself to fairly superficial integration efforts, such as getting everyone on one e-mail system. But as management pushed the organization toward greater centralization, enabled by standardizing on systems such as SAP for financial management, Inergy needed to move to a more unified and efficient network.

DeSouza identified two major changes that needed to be made to improve network performance. First, he wanted to upgrade from the outdated version of the Microsoft network operating system, based on Windows NT, that Inergy was still running on. By upgrading to Windows 2003, Inergy would be able to unify its corporate network into a single, distributed database of users, passwords, privileges and network devices, stored in Active Directory, the more scalable network repository first introduced with Windows 2000.

One limitation of Windows NT is that it forces large, multi-site networks to be divided into many small network domains. Making them function like a single network means establishing a complex web of “trusts” between domains. Besides being less efficient, this arrangement requires more network administration—more people at more locations performing redundant work such as user account changes on separate domains.

With Active Directory, the entire network could be treated as one logical domain, with information entered into any one Active Directory server replicated to the servers at other locations. Inergy also outsourced management of its Microsoft Exchange e-mail servers to Verizon Business Services.

Once that was accomplished, the second phase of the network integration project would focus on improving the actual transmission of business data. As Inergy implemented more centralized computer systems, employees at remote offices around the world were being made increasingly dependent on the performance of the corporate wide-area network (WAN). In addition to annoying SAP users, data transmission delays on the network hampered the use of internal Web sites hosted on SharePoint, Microsoft’s Web portal server.

Rather than simply adding bandwidth to alleviate network congestion, DeSouza planned to take advantage of Quality of Service (QOS) technology that gives higher priority to the most important traffic and uses techniques such as data compression to squeeze higher performance out of the available bandwidth. In a traditional network based on Internet protocols, packets of data are routed across the network in essentially a first-come, first-served manner. QOS technology makes it possible to distinguish packets associated with a high-priority application, such as posting financial data to SAP, and let them cut in line ahead of less critical transmissions, such as traffic associated with casual Web browsing.

With a WAN optimization device at either end of the network link, verbose data formats such as Web pages encoded as long strings of text can also be compressed for transmission and de-compressed at their destination, lowering the amount of bandwidth they consume.

“These were all pieces of the puzzle, but the whole thing came together because of QOS,” DeSouza says. Specifically, he used WAN optimization devices from Packeteer to ensure that the most important applications would be given priority treatment. Packeteer is one of several vendors in the niche market for devices that can distinguish between Internet Protocol network packets associated with different applications, and apply QOS rules that determine how quickly they are routed to their destination.

For example, on his network SAP traffic is assigned a high priority (meaning that it will be transmitted across the wide-area network ahead of most other traffic that is using the network at the same time), along with guaranteed access to bandwidth. On the other hand, recreational Internet access such as music downloads is low-priority traffic and may not consume any more than 8% of the available bandwidth.

“It’s a global company, and people work late—if someone wants to listen to some music, I don’t mind,” DeSouza says. But at the same time, such non-essential traffic can’t be allowed to get in the way of the monthly financials, he says.

Such considerations are particularly important for global organizations that run centralized applications, which puts more stress on the network and elevates the effects of latency—the delay in transmissions caused by the distance a signal must travel and also by network bottlenecks. “Arun [DeSouza] was clearly the champion within Inergy of saying we’ve got to do a better job, we can’t just keep buying bigger circuits, we’ve got to build expressways for the high-priority traffic,” says Jeff Sahr, operations manager with the managed services division of CH2M Hill, a consulting firm based in Englewood, Colo., that helped with the QOS deployment and provides ongoing network management services to Inergy.