Integration can suffer when networks bog down, agrees Rob Adams, CIO of Bay and Bay, a family-owned freight mover. Bay and Bay already uses Salesforce and plans to port more of its applications off-premise, he says. To pass month-to-date metrics about Bay and Bay's customers between Salesforce and on-premise systems, the company bought Informatica, a data-integration tool that Salesforce recommends.
Informatica pulls the data from two systems running at Bay and Bay--a transportation-management application that routes trucks and an IBM DB2 database for analytics--and pushes it to Salesforce.
Network downtime is a rare problem, Adams says, but still one that CIOs should address by having redundant connections on-premise and also by specifying minimum uptime--and the consequences of not receiving that uptime--in service-level agreements. For a small or midsize company without many redundant systems, he says, an "Internet connection is really going to be your weakest link."
Still, some industries may find only marginal use for cloud computing because integration can't be done as transparently as it can on-premise, says Colburn, the Finra CIO. Financial services companies must deal with so many compliance issues that a clear and complete view of their vendors' operations is imperative, he says. For example, like other securities exchanges and associations, Finra must adhere to the strict recordkeeping regulations in the Securities Exchange Act, which require keeping many documents, including e-mail messages, for five years. For the first three years, they must be "easily accessible." If a cloud provider won't specify exactly where Finra's e-mail is at any given time, Colburn says, he can't be sure that it is easily accessible. So for now, Finra's e-mail will stay in the regulator's own data centers, he says.
Dan Greller, co-CIO at Legg Mason, a global asset-management firm with $685 billion under management, hasn't put any systems in the cloud. Security isn't good enough yet and integration challenges give him pause, he says. "Commodity" applications, such as help-desk management, would be best suited for early cloud experiments, he says, in part because they don't need to be tightly integrated with Legg Mason's core business systems.
Bad integration can degrade overall application performance, which a financial services firm can't afford, he says. "Going to a more distributed computing model always brings up concerns about performance, and cloud is extreme distributed computing," he says. For example, if an analytics application running on-premise draws data from different transaction systems in the cloud, business analysts have to know how the integration occurs and with what latency so they don't make decisions or run analyses on wrong or old data. So while Greller doesn't dismiss cloud computing, he wants to see how standards evolve before making a serious commitment. If he knows that any cloud vendors he might work with will adhere to solid standards, then integration will be easier and more reliable, he says.
Sign up for Computerworld eNewsletters.