On the Bleeding Edge of MIMIX Adoption with APL Logistics

Most IBM i shops, by and large, are conservative entities that adopt new releases of software only after extensive beta testing by others. But occasionally one encounters an early adopter that’s willing to be on the bleeding edge of development–in exchange for certain benefits, of course. You can count the Arizona company APL Logistics in the latter category when it comes to its MIMIX high availability software from Vision Solutions.

APL Logistics is one of two subsidiaries of Singapore-based Neptune Orient Lines, the other being the ocean liner company American President Lines (APL). From its headquarters in Scottsdale, the $1.6 billion logistics firm manages the flow of goods across 60 countries, including 200 logistics facilities covering more than 20 million square feet. Together, the firms help make NOL one of the biggest container shipping and logistics companies in the world.

IBM has been APL Logistics’ trusted supplier of servers and storage for years, and today the company’s stable of IBM gear includes Power Systems servers running IBM i and AIX, System x machines, and System z mainframes. All of the hardcore logistics work of moving goods around the world is handled by Manhattan Associates‘ battle-tested PKMS warehouse management software running on IBM i servers. The AIX environment, meanwhile, powers SAP and EDI workloads, while the mainframe runs CRM and the firm’s liner business.

Uptime Funk

Because of the global nature of its business, unexpected downtime is not tolerated.

“Uptime is very important to us,” says APL Logistics’ IBM i Platform Manager Dave Fehrenkrug. “We own warehouses around the globe. In the middle of the night in the U.S., it’s the middle of the day in the Middle East.”

The downtime goal for APL’s production server–a six-core Power 750 machine–is just four hours per month, which is long enough to execute small system saves and apply PTFs. But Fehrenkrug has whittled that down to just two hours and 25 minutes. A large part of that uptime efficiency is the result of its use of MIMIX Availability and MIMIX Global software.

“When we put MIMIX in place, one of our primary goals was to offload the nightly backups,” Fehrenkrug tells IT Jungle. “We replicate our production environment to the HA box, as well as to the DR box. But all of our backups happen on the HA box, so we don’t have to have downtime on our production box.”

APL Logistics relies on MIMIX to replicate transactions from its production Power 750 server to its HA box, which is an identical machine that sits just 12 feet away from the production server in Hewlett-Packard‘s data center in Plano, Texas. At the same time, MIMIX is also replicating those transactions from the production box to APL Logistics’ disaster recovery (DR) box, which is located thousands of miles away in an HP data center located in Littleton, Massachusetts.

This “broadcast” style of replication, where all transactions are simultaneously sent from point A to points B and C, provides an extra level of protection against downtime for APL Logistics over the “daisy chain” style replication, where transactions are sent from point A to point B, and then from point B to point C.

With the daisy chain method, the company is potentially exposed to losing transactions if a disaster were to take down the production box while the HA server was offline for nightly backups. With the broadcast style of replication, the DR box is ready to take over production workloads even if the HA box is offline due to backups.

Testing, Testing, ABC

With the broadcast replication configuration involving three machines, APL Logistics’ HA environment may be more complicated than some. But aside from some library mapping functions, put in place to ensure the proper functionality of the Advanced Job Scheduler in the event of a failover, Fehrenkrug strives to maintain as plain-vanilla a setup as possible. There aren’t even any LPARs to worry about; those were swapped out years ago for standard libraries.

Fehrenkrug has tried to boil as much complexity out of the actual failover process as well. He’s bundled everything needed to successfully complete a failover into an application group. That allows him to complete a failover in less than seven minutes and 30 seconds, on average.

“All my environments are built under this application group,” he says. “The whole intent was to make this as automated as possible to reduce my HP operations staff from getting their fingers in the middle of it. All you have to do is switch the application group, and it’s a done deal.”

Fehrenkrug relies on the Vision Solutions Portal to monitor and maintain his MIMIX environment, including running daily audit reports to ensure that everything is properly in sync and he’s always ready to failover at a moment’s notice. Every year, APL Logistics conducts a full DR test to ensure that everything works as expected. And every year, MIMIX passes the test with flying colors, Fehrenkrug says.

“The IBM i was always 100 per spot on,” he says. “That’s was primarily due to MIMIX doing its job and the audits we run on a daily basis. The information was telling us we were in sync. But one of our goals was to make sure that what we were seeing and being told was truly gospel on the other end of the DR box.”

“I know it’s working great,” he continues. “We keep in sync with sub-second transaction speed. You make a change on production, and you can’t get to the DR box fast enough to beat the transaction. A lot of that depends on your bandwidth, obviously. But MIMIX does a good job getting our workload transferred up to the DR box.”

The Bleeding Edge

APL Logistics has been a beta tester for the last several releases of MIMIX, including versions 7.0, 7.1, and 8.0, which Vision made generally available last fall. As a former implementer and technician for Lakeview Technology, where he was employee number 32, Fehrenkrug is perhaps better positioned than others to experiment a bit with his HA setup. But that’s only because it provides his company with certain advantages.

“I work with the Vision development teams to test the new stuff as it comes out,” Fehrenkrug says. “I’ve only had one small glitch, but the guys are fast about getting on it. It’s been a good thing. Some people don’t like to be on the bleeding edge, but in this case, I like being there because of the feature and functionality that’s being supplied. It helps me to move forward.”

Fehrenkrug cites the new auditing capabilities that Vision has added in recent releases as being the top new MIMIX feature he’s tested. “Building on the audits and making sure I’m switch ready is probably number one. Because we rely on them to make sure that, at a moment’s notice, I’m square between point A and point B.”

The new object replication capabilities are another area where Fehrenkrug has found benefit. “Instead of every group having its own set of send and receive functions, I can now share that amongst multiple data groups,” he says. “So it cuts down on my memory requirements and CPU utilization required to replicate object changes and IFS changes. That’s been a major plus also.”

Not every IBM i shop is cut out for being a beta tester. But for more sophisticated shops that have a very firm grasp on the technology and are willing to accept a little more risk in exchange for getting their hands on product upgrades quicker than regular customers, it can pay for itself in the end.

RELATED STORIES

Vision Gets Serious About Hosted HA with MIMIX 8

Vision Merges HA Products Into MIMIX Availability 7