I used my debit card today, tapped the NFC area on the payment terminal and waited for my latte. An unremarkable situation, but there is so much at stake in the transmission of this data, and very powerful machines working at incredible speeds making it all possible.

The average latte-enjoyer is not typically going to consider the hops a payment request will make as it travels from tap to bank. It was my personal delight to discover that the most crucial transaction processing technology, capable of handling up to 1 trillion web transactions daily, is mainframes.

The first iteration was created for the US Navy Bureau of Ships in 1937, with commercialization taking place in 1951 under the Eckert-Mauchly Computer Corporation, and the first modern IBM System/360 reaching the market in 1965. In 1970 we saw magnetic ferrite cores replaced by silicon memory chips, but by 1991 as other technologies accelerated, the end of the mainframe era was predicted to occur only a few short years later.

We are now two decades beyond the expiration date given to mainframe computers, yet the financial industry is still largely held up by these servers, as well as government agencies, healthcare organizations, and other institutions with mission-critical systems requiring high reliability.

Entities like the NYSE or NASDAQ stock exchange must have no more than 5.26 minutes of downtime per year to avoid the impact of negative economic consequences, equating to what we call five nines of availability. This level of reliability is something only mainframe can provide.

Why Mainframe 

It is understood that mainframe computers are still in use today, but how are they different from other transaction-focused machines? Mainframes are relied upon for exemplary performance in three primary areas: reliability, availability, and serviceability (RAS).  

Reliability and Availability 

Mainframes are built tough! A recent 4.8 magnitude earthquake in New York state shook a corporate campus housing over 200 mainframes, none of which were affected by the groundbreaking event. Mainframe boxes are built to handle shock and vibrations, and may even include seismic isolation systems around the mainframes to protect them further. In fact, IBM implements standard testing that simulates the rough shaking a magnitude seven earthquake would produce.

Within a mainframe computer, built-in redundancy is another major benefit that can be found at various hardware levels, including processors, memory, power supplies, and I/O paths. As an example, the IBM z15 features up to 190 configurable cores across all processors. This contrasts with externally hosted storage and processing solutions, such as a distributed cloud environment like Amazon’s Aurora, wherein the redundancy occurs across multiple locations instead of natively within a single dedicated physical device.  

To prevent downtime, mainframes employ error-detection and correction like error-checking code (ECC) memory to detect and correct data corruption. Mainframe has been adept for many years in virtualization for efficient hardware utilization and process isolation and now incorporates comprehensive system management tools to continuously monitor performance for failure prediction. These monitoring tools can initiate preventive measures as well, like automatically switching workloads to backup systems for extra reliability assurance. 

Even the software running on these computers is specifically engineered for high reliability. Able to support things like transaction rollback, checkpoint restart, and complex job scheduling, mainframe systems maintain operation in adverse conditions that would otherwise knock competing options down.  


It may seem that mainframes would be more complex to maintain than smaller systems, but there are actually some great ease-of-use features supported.  

Mainframe computers are modular by design, so repairs and upgrades are less complicated and time consuming. Individual components (processors, disk drives, memory cards) can be swapped out with no impact to the overall system. This is huge when we remember the downtime limitations of certain industries being about five minutes maximum per year; repair time counts as downtime when it impacts the ability to complete required tasks. 

Modern systems are also equipped with thorough logging that can be used to support advanced diagnostic tools, and remote management access is available for updates and troubleshooting at will. A mainframe box hosted on-site also means access to physical repairs or updates is owned wholly by the organization employing the machine, a solid benefit to those requiring tightly controlled access to their business-critical systems.

Something to consider regarding serviceability with mainframe computers, however, is the specialized nature of the technology. Though they are not so far removed from other transaction-focused servers, there is a known mainframe practitioner shortage. This is a result of many organizations assuming mainframes would be fully sunset by the mid-1990s, effectively causing a reduction in funding for new hires and training programs. Though some organizations are now working to address this gap, businesses that employ mainframes for mission-critical processes would be wise to enlist the support of the greater mainframe security community to help ensure all bases are covered and defense-in-depth is being achieved.

What’s Next  

The Artificial Intelligence revolution has been underway for several years now and is advancing rapidly — and the continued expansion and maturation of cloud-based business solutions for storage and processing show no indication of slowing. These technologies among others could be viewed as threats to mainframe, but the reality is no solution exists able to replace mainframe computers that is as private, as powerful, or as trustworthy. It is more likely we will see hybridization of cloud and mainframe environments, with native AI enhancements in place for use-cases like advanced fraud detection.

Modern mainframes incorporate crypto, network, and compression cards all with their own processors and memory. Integrated encryption for data at rest and in motion is standard for mainframes, a feature other servers simply do not have. Mainframes are not frozen in time either, with new innovations for native AI inferencing built into the processor cores themselves and further potential to host virtual private cloud environments or proprietary LLMs, there is a healthy future in store for mainframes. 

Beginning with a basic purchase, your payment transaction data traveled from the vendor eventually to a mainframe, and from there made it back to your bank records. Perhaps the next time you tap for a latte, you’ll consider the backbone infrastructure making your debit card worth more than the thin rectangle of plastic it consists of, meaningless without the machines transmitting and processing some of your most sensitive data across time and space.