Bring It on Home
Plug and Play
The plug-and-play aspects of shared services can also mean great resiliency in a disaster recovery situation, as HFP's Lowenthal found out on September 11, when the company's 7 World Trade Center offices were destroyed. ""We didn't have to worry about losing an e-mail server,"" he offers as an example. ""I had people come up to me and say 'I can't believe the message I was in the middle of writing when we ran out of the building was waiting for me when I came back and started up the company's e-mail program again!' You can't do that with a back up tape,"" he says. The e-mail system, running Microsoft's Exchange Server, was at Hartford's regional facility, located in Boston, on September 11. It also houses HFP's AS/400 containing the firm's mission-critical data.
HFP also enjoyed the advantages of resources availability that come with being a satellite within a larger organization. ""On September 12, I had 90 brand new machines Compaq PCs waiting for me in our temporary location in Hauppauge, NY, on Long Island,"" he says. ""And just having that shared services environment meant very easily being able to pop in switches and routers based on standards. It was a very plug-and-play situation.""
Of course, the picture might not have been quite as rosy if the services were being generated at the site that was destroyed. In the past, according to Chris Heeley, CTO Royal & SunAlliance (R&SA, Charlotte, NC, $899 million in written premium), ""disaster recovery was about paying a company like SunGard (Wayne, PA) or Comdisco (Rosemont, IL) to allow you to come to their space in the event of a disaster that prevented you from processing."" What has changed, he adds, ""is the realization that if you go to a more central model, the latency in coming back for your critical applications may not be fast enough.""
Regional Hubs
In the wake of September 11, that realization is being pondered as R&SA re-evaluates its data center approach. Having grown through acquisition, the carrier finds itself with 107 data centers worldwide and 13 in the Americas (Canada, United States and Latin America), according to Heeley. ""We have a plan in place to do a global data center consolidation that we've been working on for about six months,"" he says. ""The immediate strategic goal is to reduce the scattered centers to four regional hubs,"" with an outsourcer managing data center services for the Americas.
What led the carrier to the consolidation strategy was, in the first place, efficiency. ""We think there's redundant hardware and software in our model,"" Heeley says. ""For example, we might have a data center in one region having excess capacity and another in a different region that could use that capacity."" Another driver is to achieve a shared services model on a global scale. ""We think there are economies of scale higher up than just the hardware and software, but they are prerequisites to doing real global sharing of applications,"" Heeley adds.
Having outsourced its US data center operations in 1996 (to Dallas-based Affiliated Computer Services, Inc.), R&SA saw the possibilities of achieving economies of scale by centralizing and outsourcing for the entire Canada-US-Latin America region. ""By the end of next year we'll have our rationalization data center strategy in place, so for the next 12 months we're going to figure out how to do it and actually talk about vendors potentially supporting us on a regional basis,"" he says.
While distinguishing the hybrid model of regional data center management from a purely centralized approach, Heeley says, ""the bottom line is that for 99 percent of the time the most cost-effective model is still a centralized model. You want to do that wherever possible, or get as close as you can, and on the largest scale: regional if not global,"" he asserts. But in choosing a more centralized approach, he adds, ""I think we have to be prepared for the possibility of an outage that could have a major impact. With this question there's no perfect answer: Either way you have a management issue. There are many benefits to centralizing. The problem is, it puts all of your assets close together.""
----------------------------------------------
BEST OF BOTH WORDS
SunAmerica Insurance (a subsidiary of AIG, New York, $268 billion in assets) sought to enhance its capabilities to introduce products faster and reduce the use of expensive resources that the process required, while at the same time optimizing return on investment.
The solution was to replace a mainframe-centric policy administration system with a multi-tiered component-based Web-centric browser environment. ""The mainframe program just wasn't flexible enough,"" says Bill Pan, Century City, CA-based SunAmerica's CTO. ""By the time we would complete a new product we were late in capturing market conditions.""
But in crafting an improvement on the sluggish centralized model, Pan wanted to avoid the pitfalls of the opposite approach. ""One of the issues we had was that, in a distributed computing environment, we have so many servers to manage,"" Pan says. ""This has become very expensive, so we looked for the opportunity to consolidate the servers.""
SunAmerica's approach in terms of hardware is to deploy a 32-way Unisys (Blue Bell, PA) ES/7000s running on Microsoft (Redmond, WA) Windows 2000 data center version. ""The 32-way boxes can replace at least eight eight-way boxes,"" Pan says. ""Having only one physical box to worry about will reduce our total cost of ownership.""
From an applications point of view, SunAmerica has adopted a Web services concept involving writing reusable horizontal components that can be applied to vertical applications. ""By building horizontal infrastructure we're able to achieve 'write once, use many' for all the other particular applications,"" Pan says. ""We're taking a 'cafeteria-style' approach to building those applications, meaning every time we want to deploy new products, we're basically saying you take the shopping cart, go to the cafeteria, pick up all the features and combine them to build a new product.""
Pan is also satisfied that reducing the number of units to manage will not increase exposure in the case of disaster. ""Every box has multiple power supplies, built-in fail over and redundancy for almost every single component,"" he says. In terms of performance level, he adds, ""we're trying to achieve 'five nines'-99.999-availability. We believe that with multi-way servers we can build a high-availability, fault-tolerant environment with high scalability, but with reduced total cost of ownership,"" according to Pan.
Anthony O'Donnell has covered technology in the insurance industry since 2000, when he joined the editorial staff of Insurance & Technology. As an editor and reporter for I&T and the InformationWeek Financial Services of TechWeb he has written on all areas of information ... View Full Bio
Tweet This