Storing large project files separately in its 20 offices worked fine for environmental engineering firm GEI Consultants until employees with different specialties had to start collaborating on jobs.
Once those electronic documents had to be worked on at more than one site, both transmitting and keeping track of them became a challenge, according to CIO Ray Barry. Server consolidation looked like the answer, and the company started planning for it. But there was so much latency across GEI's network connections that transfers of CAD (computer-aided design) and GIS (geographic information systems) files, which were 800M bytes and larger, took several hours.
"People quickly realize the effect of latency when you're passing a very large file," Barry said. "That issue was really a roadblock for us trying to look at centralizing storage of my company's files." GEI had to take the project off the drawing board.
GEI, with 340 employees, is not alone among medium-size enterprises in wanting to centralize its servers or make the links between its remote sites run faster, according to industry analysts. Most medium-size companies have at least three or four facilities, and just having a file server, print server and Active Directory server in each location can create an expensive headache for a small IT staff, said Rob Whiteley of Forrester Research.
"It's a quick hit for cost savings and operational efficiency to centralize all that," Whiteley said.
But before it could do so, GEI needed a WAN (wide-area network) optimization system to reduce delays. Server centralization is one of several factors, including backups and Windows file sharing, that are leading medium-size companies to seek WAN optimization, said Michael Brandenburg of Current Analysis.
"Eventually, every enterprise or [small and medium-size business] is going to get to that point," Brandenburg said.
GEI is based in Woburn, Massachusetts, and provides consulting services to developers, governments and other clients concerning geological, water and environmental issues. Its offices throughout the U.S. range from five to about 50 employees, and its engineers have a wide range of specialties, Barrie said.
Until recently, each of those offices tended to work on its own projects without much help from others, Barry said. But as GEI's projects grow more complex, employees with different skills rely on each other more.
"Our collaborative work between our offices is only going to increase," Barry said. That often means importing CAD and GIS files from across the country, making changes, and sending them back. Also, IT staff needs to give employees from different offices the rights to get into the remote files they need. Putting all that information in one data center, with a standard system of naming and organizing files, would help employees find the correct version of each file to work with, he said. It would also make it easier to properly keep or discard versions of files, which helps with record-keeping laws and court cases, he said.
Even as it plans for server consolidation, GEI is making changes in its network that make WAN optimization more critical. The company is rolling out IP (Internet Protocol) telephony across its offices, which adds new traffic to its data network. And as part of that move and to increase security, it is switching from individual Internet connections for each office to an MPLS (multiprotocol label switching) managed service. That means all offices share one connection to the carrier network, a potential bottleneck, so GEI needs to make more efficient use of its bandwidth, Barry said.
While investigating server consolidation, Barry was introduced to Exinda Networks, an Australian vendor of WAN optimization and application acceleration gear. GEI set out to solve its latency problem and now has Exinda appliances in nine of its sites that already have MPLS connectivity. Ultimately, the company will probably install twice that many, Barry said. The devices have slashed transfer times by 30 percent to 40 percent, he said.
The market is full of WAN optimization products, but many are designed for large enterprises. GEI chose Exinda partly because it cost less than alternatives such as Packeteer or Riverbed, Barry said. He's not alone, Forrester's Whiteley said. The first wave of a network technology is usually geared toward large enterprises, and WAN optimization is now in its second phase, so vendors such as Exinda are going after smaller customers that are more sensitive to price and want a product that's sized right for their business, he said.
Another highlight of Exinda's product is that it can deliver detailed information about how GEI's network is being used, Barry said. This became more important with MPLS, because with the new network, all the offices share a common entry point to the Internet, he said. It showed Barry's team what kinds of traffic were going through that link so they could block non-work applications, such as Internet radio, that got in the way. The visibility is also providing baseline information about network load so the IT department can make a case for more bandwidth if it's needed as a result of data-center consolidation, he said.
But before centralizing its servers and storage, which Barry estimated at 4T bytes of data, GEI needs to find an effective document management system it can afford, he said. The company will probably consolidate its data center in early 2009, Barrie said.