Time For Airlines To Get Past The Mainframe Era
from the up-to-date dept
There have been a number of high-profile snafus at airports this year, and while bad weather tends to be the most prominent cause, it's not the only culprit. Last month, passengers traveling on US Airways faced massive delays after self-service kiosks suffered a glitch, stemming from the company's ongoing attempt to merge its reservation system with that of America West Airlines, which it acquired two years ago. While most industries have invested heavily to upgrade their IT systems over the years, the major airlines still rely primarily on legacy mainframe systems that were developed in the 70s. For day to day operations, these systems get the job done, but they're not very flexible. Even when weather is a major factor in an airline's problems, the lack of a good IT setup can exacerbate the problem. In February, jetBlue suffered a major meltdown, in part because the company was not sufficiently prepared to track its assets in the event of a major crisis. Following the fiasco, the company claimed that it had upgraded many of its systems, although considering the difficulty and time it takes to upgrade enterprise systems, it's hard to imagine that much has already changed. Part of the reason the industry may be so behind in modernizing its IT is that for many years, most of the major carriers were hemorrhaging cash and teetering on insolvency -- hardly the ideal conditions for major investments. But now most of the airlines are profitable again, and with 2007 looking like another bad year for things like delays and lost luggage, it might be time for the industry to invest heavily in its own IT infrastructure.Thank you for reading this Techdirt post. With so many things competing for everyone’s attention these days, we really appreciate you giving us your time. We work hard every day to put quality content out there for our community.
Techdirt is one of the few remaining truly independent media outlets. We do not have a giant corporation behind us, and we rely heavily on our community to support us, in an age when advertisers are increasingly uninterested in sponsoring small, independent sites — especially a site like ours that is unwilling to pull punches in its reporting and analysis.
While other websites have resorted to paywalls, registration requirements, and increasingly annoying/intrusive advertising, we have always kept Techdirt open and available to anyone. But in order to continue doing so, we need your support. We offer a variety of ways for our readers to support us, from direct donations to special subscriptions and cool merchandise — and every little bit helps. Thank you.
–The Techdirt Team
Reader Comments
Subscribe: RSS
View by: Time | Thread
[ link to this | view in chronology ]
It knows where every plane is, where every bag is, and where every person is. It knows when a person is going to miss a connection and automatically rebooks that person on a later flight. It can predict how a delay will effect future flights and correct the problems
The only downside from the systems is a lack of a GUI, but with the efficiency that the employees use the CLI, I would hate to see it changed.
[ link to this | view in chronology ]
Not really any other option
[ link to this | view in chronology ]
Re: Not really any other option
Google isn't using a warehouse full of mainframes to do their work and they handle more data per minute than any airline will have to track in a week.
[ link to this | view in chronology ]
Re: Re: Not really any other option
Just what i want to trust my billion dollar business to.
[ link to this | view in chronology ]
See, the FAA has to approve everything in airline IT. No big deal; banking was the same way. Except FSLIC/FDIC regulators were a whole different breed than the FAA ones, and getting *anything* approved by the FAA was an exercise in futility. FAA regs are sufficiently badly written as to allow a great deal of latitude in individual readings, and if the guy assigned to your airline is so motivated (say, by you making work for him by asking him to approve a new system) he can choose to interpret things like crew schedules in the most inconvenient way possible. There's very, very heavy incentive to not upset the status quo.
[ link to this | view in chronology ]
Must be over 40
[ link to this | view in chronology ]
Nothing wrong with using a Mainframe
I've seen many cases where people (management) get a grand idea on replacing the mainframe system with either a turn-key product, or custom developed on "another" platform. Many have ended in disaster for various reasons, and they just default back to the mainframe. I've also seen people successfully migrate from a mainframe system too. If it ain't broke, don't fix it.
There are plenty of ways to modernize a mainframe system for today's age, you just have to do it properly! To bin the decision on "Mainframe == crusty old equipment" just isn't fair.
[ link to this | view in chronology ]
The real story
OK, you've seen a GUI at a kiosk - big deal. That is just a front-end system that still relies on the mainframe back-end. Travelocity is a HTML front end that has the Sabre mainframes on the back-end. These are in the same data center in Tulsa, OK. When you do a low-cost search on Travelocity, the mainframe's knees buckle.
And these are not MVS systems either. Sabre can run up to seven of the hottest IBM frames as a cluster, with the TPF operating system. 4000+ real-time transactions per second. A huge DASD farm is attached (disk drives, in Frame speak.)
The inertia of changing these systems is tremendous. Look at the history of Sabre on Wikipedia, it is very good and mostly accurate.
But, don't think you can replace seven frames with a lean OS built for transactions with Server 2003. Maybe a large Beowulf cluster could. Or perhaps Google could...
[ link to this | view in chronology ]
Re: The real story
Any company engaged in a lot of mergers, aquisitions and sell offs (such as almost the entire airline industry) will encounter major problems in their IT. The problems facing the Airline sector at the moment are due largely to the sheer number of technologies and applications in use, and the corresponding number of conduits, portals and methodologies required just to get them to talk to one another - all this dramatically increases the possibility of error
The other key problem I often experience is that companies expect to just merge and be able to cut costs immediately (usually by shedding staff). This becomes a problem if a company does not consolidate its IT infrastructure, then fires the guys who knew how x and y legacy systems it just inherited really work
So I think basically my point is that its not just a platform issue but the way IT tends to fall through the crack in general in mergers, often due largely to the way boards tend to think of an IT merger as just "pluging that there machine thingumy from our new company into our own interweb thingy"
[ link to this | view in chronology ]
Airlines require a central data repository that requires data locking and central management.
It is possible to do that with a cluster or network solution, but those solutions do not have the history that the mainframe ones have.
Yes, a GUI will make usage easier.
I've used PC's, clusters, and mainframes.
Each have their own strengths and weaknesses.
I've seen some real-time solutions that run great on a mainframe.
And I've seen some huge databases run great on a Unix cluster.
I think what matters is the System Architecture. Choosing the hardware is part of the process after deciding what the most important requires are and how they can best be addressed.
[ link to this | view in chronology ]
It's not the platform its the code
The world was smaller and simpler when these systems first went online. Over the years I imagine that the airlines have chosen to not start over with a solution that was architected from the ground floor to be a comprehensive solution. Instead they most likely just kept cobbling and patching on top of their original code base.
When you take this approach changes can take a tremendously long time to engineer, not because the type of code itself is "inflexible" rather because the code that was written has so many dead ends and workarounds that any changes requires numerous and extraneous changes to patch programs that run in the background.
And before any starts with the "you must be over 40" comments, I'm actually 31.
[ link to this | view in chronology ]
Please base arguments on fact not fiction
Almost all stories of computer related problems have occurred on new up-to-date systems not on legacy systems. Below are two of the best.
Navy ship Yorktown when their computer system tried to divide by 0
http://www.windowscrash.com/modules.php?name=News&file=article&sid=1
Wall Street Melt Down
http://www.msnbc.msn.com/id/17424307/site/newsweek/
Try to even find a headline of a legacy system having a compute error that crashed a company. It's almost impossible.
Even the article quotes the company as saying it systems were "upgraded", but instead of putting the blame on the new system, Joe implies that it must have been the "old" system that went bad, because it works better for the story.
Look at the gaming industry that relies heavily on mainframes (as400's). To my knowledge there has never been a computer related problem that has shut down a casino that has made a headline and this is in a business that runs 24x7x365.
I imagine that the airlines kiosks where being run by a new up-to-date computer system that with a GUI that then updated a mainframe (if a mainframe was involved at all since it hasn't been proven).
[ link to this | view in chronology ]
Wow
The reality is that technology is moving forward and so should any IT department that doesn't want to be on the news. The airlines will suffer a catastrophic issue resulting from aging technology and that will prompt them to change. It has very little to do with how well the system works and everything to do with the amount of money it would cost to upgrade.
[ link to this | view in chronology ]
Its about cost
[ link to this | view in chronology ]
[ link to this | view in chronology ]
The real story continued
So bad example.
And for what it is worth, TPF is an awesomely powerful OS.
Can you name me one other OS where you can run multiple versions of the OS in parallel, upgrade application versions without bringing the system down (open transactions finish in the old version, new transactions open in the new version), and unplug the whole data center, and then resume in mid transaction with no data loss the minute the power is turned on again?
Issues such as the merger problems you refer to are as "that guy" says, due to coding issues. business rules and fare structures are so different from one airline to the next (in fact they are considered closely guarded secrets) that merging the two is a monumental undertaking akin to changing the tires on a car while it is still moving.
I feel certain that there would be even more interruptions that would take longer to fix if this were being ported over to a web application at the same times as the operations merging was taking place.
[ link to this | view in chronology ]
[ link to this | view in chronology ]
Re:
When a query is finished, that customer is "out the door" so to speak, the whole life of that transaction is finished.
Airlines not only have to handle queries, but also keep up with flights, tickets, bags, payments, seats, crews, etc.
This is like saying "I wrote an excel macro I must be a progammer".
[ link to this | view in chronology ]
Re: How many mainframes does Google have?
[ link to this | view in chronology ]
Re: Re: How many mainframes does Google have?
Yes, the major GDS systems (that is the generic names of Sabre, Worldspan, etc.) do more than reservations. They also load balance the planes, track cargo reservations, and track maintenance. If a GDS goes down, planes don't fly. All of the major airlines use one of the big GDS systems.
[ link to this | view in chronology ]
The Future Is Almost Here!!!
[ link to this | view in chronology ]
The capital investment can be High (mainframe) or low(Linux)
Maintenance can be easy( 1 mainframe) or hard (2000 Linux servers)
A server can be reliable (mainframe) or not (beige box)
You can do any number of these comparisons which is "best" for the situation.
For those of you with less experience, the mainframe is a perfectly fine File Server, Database Server and Application Server
[ link to this | view in chronology ]
mainframe legacy nightmare
[ link to this | view in chronology ]
Ignorance and Competitive FUD
[ link to this | view in chronology ]
Re: Ignorance and Competitive FUD
JetBlue not only runs Microsoft software, but they also run beta code to lower costs. JetBlue also admits this is a risk they are willing to run to be cheaper.
Cutover? Consider US Airways - they had their own tiny reservation system until 1999. They decided that it wasn't worth implementing a four digit year, and cut over to Sabre instead. So, check your November/December 1999 news - can you find any problems with US Airways?
The issue is continuity. How long is IBM going to make a compatible mainframe? How many years have they made one? In contrast, try to load Server 2003 on a 80286.
These companies have been doing volume transactions for years. Last I knew, Sabre had a 99.95% uptime. They have to have a long-term, clustered reliable platform.
I'm not say it is perfect. DST is done by hand, and in the fall the system is shut down for an hour to prevent time duplication. They cluster up to seven frames, but they don't all the time. The count varies with load. That way they can do hardware changes.
Some airlines still run standalone. Last I knew, Southwest was still running VMS on VAX hardware.
But, the big res companies are in trouble. Sabre outsourced the data center to EDS for "cost saving", but still had to go private.
[ link to this | view in chronology ]
outdated
[ link to this | view in chronology ]
Re: outdated
Clusters are considered "custom" installs.
[ link to this | view in chronology ]
to using mainframes is there isn't really any z/OS for
dummies books out there that so many so called IT
pro's rely on these days to get by.
[ link to this | view in chronology ]
[ link to this | view in chronology ]
stick with mainframes
[ link to this | view in chronology ]
The airlines are still to reliant on mainframes?
[ link to this | view in chronology ]
[ link to this | view in chronology ]
What should happen with luggage?
Less than 1% of luggage actually remains lost for over 60 days, so even though it is not in your possession there is an excellent chance that it will be found. Your luggage will not be considered officially lost (unsalvageable) until 90 days are up, at which time the airliner must declare the piece(s) lost in order to comply with FAA regulations. Unfortunately this process may drag a little, but at least people will be looking for it for a full 90 days.
Maybe a better system would prevent that lost period.
[ link to this | view in chronology ]
Re: What should happen with luggage?
Also, the rate of lost bags is lower than the error rate of most programmers (defects per lines of code).
[ link to this | view in chronology ]
Sabre is moving to C++, MySQL, and UNIX
The article gives an indication of the benefits and drawbacks of the two platforms -- mainframes have the reliability built-in at a high cost, commodity cluster solutions require software to manage the inherent unreliability.
[ link to this | view in chronology ]
Misconceptions abound.
Mainframes have evolved just like PC and UNIX systems, but too many people have their preconceptions of what mainframe computing is like.
Mainframes have UNIX(for 12+ years), JAVA(10+ years), TCPIP(12+ years), can run Linux(7+ years), are web enabled... can be used as a server and as a client.
The issues are not the technology but the money. Management doesn't want to spend money to modernize, unless they are forced too... Period. and who can blame them... What business value will be gained by spending $100M to migrate them off mainframes.
Also please note that many of the airlines are outsourced to EDS, a major advocate of the MMA. Even they aren't going to spend that kind of money without some financial benefit.
It would also be interesting to see what people's comments are on the Denver Airport's baggage handling system. a UNIX cluster 'state of the art' system. It damaged baggages, wrongly routed baggage... and after two months it was declared FUBAR. Now Denver doesn't have a 'state of the art' system. Is this what we want? Is this what you are suggesting.
To J:
You DO realize that mainframes are clustered computing... and have been for 30+ years....
NJE mainframe computing is the equivalent of homogenous grid computing...30+ years...
sysplex mainframe computing is the equivalent of heterogenous grid computing(about 20 years)
parallel sysplex computing(10+ years) has no equivalent in the midrange world, yet(HP is working on it).
Also, unlike the midrange world... alot of the mainframe software(applications, databases) can take advantage of this clustering out of the box, and is not a custom job.
to ehrichweiss
Almost all mainframe business clients do keep their equipment up to date. IBM won't support old machines or old OS systems. Any business that doesn't keep their production on support hardware/software, is generally one that goes out of business. The technology you cite is from the 70s/80s, ask yourself if your Trash 80 could do anything nowadays-so let's compare apples to apples, shall we... A z9 mainframe server, can generate up to 96 GB(that's byte(BIG B)) of i/o throughput. ... This is just one machine... imagine that clustered?
You may also want to keep yourself up to date. Virtual switch technology(where the mainframe has advantages(other platforms do have them, too)) can BYPASS those gigabit pipes. Data transfers at memory speed...Try that with traditional client/server.
[ link to this | view in chronology ]