Cross-ocean clouds gain despite millisecond delays

Japan's AIG Edison says cloud computing benefits outweigh latency issues that come with using's U.S. data center

1 2 Page 2
Page 2 of 2

AIG Edison has found that latency from the U.S.-Japan link varies according to connection speed and amount of data. On average, it takes 132 milliseconds to send and receive 32KB, according to Appirio. In contrast, it takes about 52 milliseconds for a similar amount of information to travel via a Japan-based host site.

AIG Edison's entire client environment, which includes virtual desktops for all of its salespeople, experiences a maximum lag of 300 to 400 milliseconds, or about one-third of a second, according to Appirio.

That latency rate comes after optimization and tuning. Among the things AIG Edison did was transmit batch loads of data into the Salesforce data center, where the majority of data was located. Also, instead of doing four sequential queries, the system was optimized to do four queries at the same time.

"If the application is written in such a way that it is minimizing the number of round trips to the database," latency "can become a minor issue," said Andy Poulter, CTO at Runaware Inc., which provides a number of SaaS- and cloud-based offerings, including online software testing. It has data centers in Sweden and Miami, and it serves a worldwide customer base.

Some AIG Edison data, along with customer history checks, still has to be pulled off a mainframe in Japan.

For AIG Edison, the decision to work with a cloud provider in another country raised security concerns that also had to be addressed.

"Just having the data residing outside of Japan was hard for some people to get over," said Jason Park, general manager of Appirio's Japan operations.

But executives' concerns were mitigated by explanations of the physical and logical security Salesforce had in place. They found that Salesforce "was probably better both from a reliability and uptime perspective and provided more robust security as well," Park said.

AIG Edison was acquired last month by Prudential Financial Inc., and Futami decided to leave his position. But decisions like the one he made to move to the cloud are becoming increasingly common: U.S. revenue from public cloud computing is expected to increase 24% this year alone to $17.6 billion, according to IDC.

Phil Garland, a partner in the PricewaterhouseCoopers advisory practice, said that whether latency is an problem or not depends on the user's expectations, level of tolerance, and what works for the business.

"It really depends about what performance levels are acceptable to you, said Garland. "There is no question that the farther one is away from a data center, there will be an impact on performance," he said.

"But there are ways that many providers work around that by balancing between actual performance and perceived performance," said Garland. "A clever client design can alleviate many of those issues that are presented by high-latency environments -- at least up to a point."

Garland said there's no rule of thumb on user acceptance of latency. It depends on the function of the data and how critical it is. But the topic has grown in importance as companies move toward data center consolidation at the same time as global customer bases are expanding. "It's a very common discussion point right now," he said.

Patrick Thibodeau covers SaaS and enterprise applications, outsourcing, government IT policies, data centers and IT workforce issues for Computerworld. Follow Patrick on Twitter at @DCgov, or subscribe to Patrick's RSS feed . His e-mail address is

Copyright © 2011 IDG Communications, Inc.

1 2 Page 2
Page 2 of 2
Shop Tech Products at Amazon