|
by Robert Gast
Published 09/26/2006
On Saturday morning, September 24, 2005, as
Louisiana National Guardsmen distributed bottled water, ice, and food to people
whose lives were shattered by Hurricane Rita, the emergency IT staff of Unibill,
LLC was in the final stages of an unplanned failover of its business-critical applications to the company's dedicated backup site in
| http://ads.valuead.com/external_reference?bid=10011&eid=0" target="_blank" |
Dallas.
Unibill provides a hosted Customer Care and Billing package to
several cellular and landline telephone companies on its iSeries Model 810. As
one of its mission-critical processes, Unibill's AEGIS/Secure application polls
telephone switching equipment for "pay as you go" cellular telephone operators
in real time over a VPN link. It downloads call records in near-real time so
they can be rated. Then, it decrements the value of each call record from the
appropriate prepaid value bucket.
"We are in a business where you have to
be live all the time," says Mike Clark, Systems Manager of Unibill. "That's why
we located our main production data center in a building that more closely
resembles a fortress. Our building is literally made out of stone; we have a
280KVA diesel generator and a UPS that will keep us running for 200 minutes in
the event of commercial power loss and until the generator takes the load.
Furthermore, we have a fully mirrored high availability hot site backup in
Dallas. All of these things made us feel pretty secure."
Late on the
evening of Friday, September 23, as the wind picked up, Clark and a few members
of his IT team hunkered down in Unibill's data center. As Rita struck the gulf
coastline about 30 miles to the south at category-three intensity, the outlook
in the data center was still guardedly optimistic. Ten minutes later, Rita
thrashed furiously at Unibill's door. Because of severe tree damage and toppled
communication towers, telecommunications lines proved to be the Achilles heel
for Unibill. Unexpectedly, Unibill began to lose its Internet links to the
outside world as connectivity was lost to the ISP.
In retrospect, Clark
realizes that he should have done a role swap to the hot site (enable the backup
system to take control of applications) as soon as the storm threatened to hit
land somewhere along the upper Texas and southern Louisiana coastlines.
Unfortunately, as midnight approached on Friday, the VPN connection between his
iSeries box and the outside world was gone. According to Clark, "The lesson we
learned was to do the role swap sooner. We were confident of our building and
our emergency backup systems, but there are so many other things that can fail
over which we have no control."
As it happened,
Unibill had licensed Echo2 from premier high availability software
vendor iTera, Inc. in August 2005. As a
final test of the Echo2 installation process, Unibill had scheduled a
role swap test for October, but Rita was impatient and struck in September. When
the VPN links failed early Saturday morning, Unibill was forced to do an
emergency role swap (commonly called a "failover") and immediately called iTera.
Unibill had dispatched its offsite disaster recovery team to its Dallas backup
site, hoping the team would not be needed.
When Unibill staff arrived at
the backup site in the early hours of Saturday morning, iTera guided them
through the necessary system audits on the iSeries Model 520 backup machine.
Because Unibill was relatively new to Echo2, an iTera technician
stayed on the telephone throughout the whole process. "I was very pleased with
their commitment," says Clark. "They said they were
going to support us, and they were there whenever we needed them." Together,
Clark and iTera's support staff found that the backup machine had committed
every transaction up to the point where the communication lines had failed.
After the failover had been completed, all that remained was to have Unibill's
customers point their systems to the IP address for the backup 520 in Dallas.
With Echo2, replication latency is negligible because it
exploits the capabilities of the i5/OS remote journaling capabilities. Prior to
deploying Echo2, Unibill tried to use another HA product, but because
Unibill's AEGIS/Secure's application dynamically creates and deletes files so
quickly, replication latency times grew to an unacceptable point.
"I
don't want to sound melodramatic, but when I think about what the business
losses would have been without the hot site, I tremble," says Clark. "Our
customers require 24x7 uptime. Quite frankly, if I'm not able to ensure
continuous uptime, my customers will look elsewhere for their billing needs."
While waiting for electrical and telecommunication services to be
restored at the primary location in Lake Charles, Louisiana, Unibill continued
to run on the backup machine in Dallas for a little over four weeks. "Though our
customers were impacted as a result of the unplanned failover, they were
nonetheless able to continue with their mission-critical processes on the HA box
just like nothing had ever happened," says Clark. Once the production site
became available again in Lake Charles, Unibill was able to move operations back
via a planned role swap.
Today, Clark is well-versed in the operation of
Echo2. "I find it to be very straightforward and easy to use.
Echo2 runs on top of i5/OS and utilizes many of its functions to do
its job. iTera makes it easy to maintain the Echo2 software with an
automated PTF system. And I like the fact that I don't have to bring the entire
Echo2 system down to apply them."
In the end, the chaotic
situation that Unibill was thrown into would have been much worse without
Echo2. As a result of its experience with Hurricane Rita, Unibill has
made quite a few changes in how it runs its shop. According to Clark, as a new
hurricane season approaches, "I am confident that as far as Echo2 is
concerned, we have the right technology in place, and we are well-prepared for
whatever comes our way."
Robert Gast is a
regular contributor to MC Showcase and has reported on technology and
business since 1986. He is the managing partner of Chicago area-based Evant
Group and can be reached at bobgast@evantgroup.com.
iTera, Inc. is a
leading developer and integrator of high availability and continuous
availability solutions for System i. iTera's products have been implemented at
hundreds of organizations around the
world.
iTera, Inc.
5250 Wiley Post Way, Suite 500
Salt
Lake City, UT 84116
Tel: 800.957.4511 (USA and Canada), 801.799.0300 (other
countries)
Web: www.iterainc.com
Email: info@iterainc.com
|
