Moving Off Mainframe

Sabre Inc. is the biggest provider of air booking in the US. Their real-time computing technology initially used in 1960 was designed by IBM, and the company adopted the use of the IBM System/360 mainframe computers in the early 1970s. Since the turn of the century, the company has been vocal about swapping out its mainframe systems for more “modern” and less costly systems. This was undoubtedly fuelled by IT vendors actively marketing and selling their new products as modern replacements for legacy technology. Here is a short timeline of Sabre’s mainframe replacement announcements:

In 2001, Sabre announced that they were in the process of shedding its mainframe “legacy”. At that time, Sabre was making the move off the mainframe with the help of Compaq Corp. and Electronic Data Systems (EDS). We’re not sure of the details of that, but clearly, goals were not achieved—no announcements of a successful project conclusion from Compaq, EDS, or Sabre. What we can conclude is that either they stopped the project in the planning stage, or they burned through serious capital expenses before giving up. Hopefully, it was the former.

More recently, in 2018, Sabre announced that it would be done with its mainframe systems by 2023. Well, it’s 2023, and their mainframe systems are still online. They say that the project will “will reduce operating expenses by more than $100 million annually”—if they can pull it off, it could be worth the effort. If not, I wonder how much CapEx is going into the project?

By 2021, they were saying that, while they were having issues with running multi-cloud environments, they were going to have all data moved off their mainframe systems and into Google Cloud by 2023. Again, it’s 2023, are they done yet? What they’re saying is that they plan to “spend about $2 billion on Google’s cloud platform over… [a ten-year period].” Trying to keep track, they’re planning to spend $200 million annually on the Google Cloud platform; add that to how many years of who-knows-how-much CapEx spent to get there, and one might begin to wonder if the project is making sense.

Last year, they announced that they were heading to the Google Cloud. However, at the end of this piece, they said—“the Google Cloud solution is built on top of Sabre’s existing data warehouses, but in the long term, the plan is to consolidate data onto one system, but for ‘some time’ Sabre is likely to be operating both worlds with the new world layered on top of the existing world…”  What does that mean? They seem to be in a situation where they are running the “existing world” mainframe systems AND their big new and expensive Google Cloud solution. Ouch. This solution means they’re paying for two platforms—two solutions—to do what their mainframe systems were doing all along. Does this count as a success?

Will they ever get their new systems to work without running the “existing world” systems in parallel? Possibly. But if they must have the performance and throughput of their existing systems, in their production, staging, test, and development environments. It is extremely unlikely that the operating costs for their new systems will be able to compete with the costs they have been paying for their “existing world” systems. 

As many client-server CIOs in the past have learned (and what has long been known by IT execs with expertise on both platforms), the mainframe architecture was designed from the ground up to handle the most intensive transaction processing environments. Cloud systems, while extremely flexible, are often based on commodity server architecture technology. You can match mainframe performance and throughput numbers with commodity servers, but you can only do it using brute force – by piling up servers ad nauseam until you reach your goal. In the end, you lose the cost game. And that’s an expensive way to go on more than one level, considering the reason for the replacement / “mainframe modernization” project in the first place, was to save on operating costs. Oops. It’s not like this hasn’t happened before—there are many stories floating around—stories about organizations footing the bill, and CIOs failing to succeed—they’re embarrassing, and people tend to stay pretty closed-mouth about them.  To be fair, though, Sabre will probably succeed eventually, because it is now no longer a cost reduction project (how can it possibly be?), but rather a political engagement. They’ll make it work if it breaks the bank—which it probably already is.

In addition to the planning, development and management of the Planet Mainframe blog, Keith is a marketing copywriting consultant where he provides messaging for corporate and partner products and solutions.