The price of (near) perfection

When NASA officials at opposite ends of the United States try to e-mail

a document in time for a meeting with agency Administrator Daniel Goldin,

there is no time to waste — and no leeway for outages.

If the e-mail server is down, workers resort to less risky technology,

such as fax machines, said Milton Halem, assistant director for Information

Sciences and chief information officer at NASA's Goddard Space Flight Center,

Greenbelt, Md.

Halem found himself in such a situation once while trying to draft a

memorandum of understanding with his counterpart at Ames Research Center,

Moffett Field, Calif. Ames expected a visit from Goldin within hours, so

they had to fax the document back and forth. Halem wants to avoid such time

squeezes in the future by researching what it would take to provide near-perfect

reliability on the e-mail server.

"There's been a general complaint with respect to responsiveness, accessibility

and long-term capability," Halem said. He wants the e-mail systems covered

under his site's seat management program to handle collaborative, time-

critical efforts, such as creating official documents and working on NASA

missions. Those tasks are done now using primarily Unix and high-speed networks.

Halem asked ACS Government Solutions Group (formerly The Intelli-source

Group Inc.) to show what it would cost and what infrastructure is needed

to give Goddard's e-mail servers "99.999 percent" reliability. ACS is responsible

for managing and refreshing Goddard's desktop environment and operating

its servers under the multibillion-dollar governmentwide Outsourcing Desk-top

Initiative for NASA (ODIN) contract.

Many NASA employees work from home in the evenings, telecommute during

the day and collaborate on deadline-oriented work with people in other

areas of the country, Halem said. "There is a discomfort level in not being

able to read your generic mail as well as not see the documents you want

to."

NASA's ODIN program man-ager Mark Hager-ty estimates that about half

the e-mail servers at Goddard are managed under ODIN. He's excited about

the effort to improve their performance. "I'd love to see us get standardized,"

he said, rather than have a mix of Qualcomm Inc.'s Eudora, Microsoft Corp.'s

Outlook and others. "I'd like to see us all be on the same page."

ACS recently conducted a survey of Goddard e-mail users served by the

nine servers under ODIN. According to the survey, about 58 percent of users

said the current Post Office Protocol (POP) mail services were adequate,

while 25 percent said they were minimally adequate.

Halem indicates that there's good reason for this low satisfaction rate.

A number of viruses and other system compromises have made the system unavailable

for long periods of time, he said, often extending to several hours.

"EBay, Amazon.com and Yahoo can't afford for their systems to go down.

I want to know what we need to do here to get that level of service." He

believes the situation could be remedied by an uninterruptible power supply

and backup mirror sites on the networks.

Halem expects answers from ACS soon. Depending on the results, he may

try to raise the issue as an agencywide problem, he said. "I want my people

to say they're happy with the e-mail."

ACS has faced two problems: how to improve the current system to provide

faster throughput and better availability, and how to meet the more stringent

requirement of 99.999 percent reliability, said Chris Austell, ACS's enterprise

services manager for ODIN at Goddard.

To achieve 99.999 percent reliability, he said, "each building block

needs to be bullet-resistant," which gets expensive very quickly. The goal

could be achieved with a redundant network path into and out of the system,

a redundant data store and the high-reliability processors that connect

the two, he said.

"You don't have to look hard at the problem to see why it's an issue,"

Austell said, noting that information systems management has moved e-mail

away from centralized guarded systems to decentralized, unguarded systems.

People who use POP mail, he said, don't pay attention to performance management

because POP is easy to plug in and "it just works."