4 ms·

Interesting: Can you provide some examples of Mainframe capabilities not available to laptops or x86 machines in general?

reaperducer 8y ago
Error-checking at the bit level? https://en.wikipedia.org/wiki/Bi-quinary_coded_decimal https://en.wikipedia.org/wiki/Bi-quinary_coded_decimal It's esoteric, but something that popped into my head.
merb 8y ago
* extremly reliable * have each component twice to have high availability inside a system (not confusing with internet-era high availability where you have multiple machines) * high i/o * backward compatibility until it hurts
- Aloha 8y ago
  This is the answer worth pointing at.
Spooky23 8y ago
Primarily, they run the software built in the 1960s-1990s. They are highly secure/trustworthy from an isolation POV. It's a much more baked platform that what you see in x86.
- walterbell 8y ago
  Did they escape Spectre/Meltdown?
dec0dedab0de 8y ago
I think the biggest benefit is business logic written in the 70s, that the company has been relying on since.
bitwize 8y ago
IBM won't sue you. In this day and age, x86 hardware is good enough to run many IBM mainframe apps through the Hercules emulator. But if you do this, you run the risk of running afoul of IBM's patents and being threatened by their lawyers. IBM has made a promise not to enforce patents against open source software, but is only willing to uphold that promise as long as you don't cut into their bread and butter which is mainframe support.
rbanffy 8y ago
RAID-like memory: tolerates one - or more, IIRC - failed memory modules without stopping the process using that memory. Failure prediction and recovery: if a core presents correctable errors, it's workload is moved to a different one. The process never needs to know what happened. In the case of mainframes that have spare cores, an inactive spare is brought online automatically. Bunches of smaller computers to offload everything away from the CPUs - the main processor rarely bothers to do any IO - it's all done from separated computers with their own memory, IO channels and so on. Even powering up the machine is, sometimes, accomplished by a secondary computer that sets everything up before the actual machine is powered up. There are a lot more nice things mainframes do that x86 boxes simply can't.
larrik 8y ago
Good question. I used to sell AS/400's, which aren't quite mainframes, but fit the typical picture in your head (green screen, dot-matrix printers, high price tags and all that). 1) Stability. That stuff changes VERY rarely, and when it does it tends to be very backwards compatible. Our software had binaries compiled in the 80's still fully operational. 2) Reliability. You don't reboot the server. You don't wipe it and reinstall it. You don't even upgrade it or replace it that often. You can do things like add or remove drives from your RAID array while the machine is still at full utilization. Heck, IBM AS/400's will detect the error, order the part, and get an IBM guy there in the morning to install it. You just have to let him in the door. In fact, you can hot-swap things like the RAID controller card itself, which is WAY not something you can do on x86. They last forever, too. It wasn't uncommon for us to have machines in daily use for 10+ years without a single bit of maintenance or an IT guy even looking in its direction. This is for a business critical machine, mind you. 3) Single vendor Ever buy a Dell or HP server and have an issue, and you spend 2 weeks trying to get HP/Dell vs Microsoft vs some other company to admit fault and fix the damn issue? Then you find it online yourself because that's nuts? Well, IBM is responsible for hardware and software on these machines (besides what software you put on it) so if it isn't working, you call them and it's their damn problem to get it sorted. Kind of like Apple, except they'll send as many people as necessary for as long as necessary to sort that out for you. 4) Entrenchment It works, why bother changing it? It doesn't even slow down as it ages, either, since you aren't upgrading a GUI OS like Windows or whatever.
- starsinspace 8y ago
  Now this sounds fascinating about the stability and reliability, but it leaves me wondering how such decent systems can from the same company which makes the horror that is IBM Notes...
  - dingaling 8y ago
    IBM was more like a federation than a single coherent entity. System/38 and AS/400 evolved in Rochester MN, mainframes in White Plains NY, and PC software all over the place. None of the divisions had much to do with each other.

reaperducer 8y ago

Error-checking at the bit level? https://en.wikipedia.org/wiki/Bi-quinary_coded_decimal https://en.wikipedia.org/wiki/Bi-quinary_coded_decimal It's esoteric, but something that popped into my head.

merb 8y ago

* extremly reliable * have each component twice to have high availability inside a system (not confusing with internet-era high availability where you have multiple machines) * high i/o * backward compatibility until it hurts

Aloha 8y ago
This is the answer worth pointing at.

Spooky23 8y ago

Primarily, they run the software built in the 1960s-1990s. They are highly secure/trustworthy from an isolation POV. It's a much more baked platform that what you see in x86.

walterbell 8y ago
Did they escape Spectre/Meltdown?

dec0dedab0de 8y ago

I think the biggest benefit is business logic written in the 70s, that the company has been relying on since.

bitwize 8y ago

IBM won't sue you. In this day and age, x86 hardware is good enough to run many IBM mainframe apps through the Hercules emulator. But if you do this, you run the risk of running afoul of IBM's patents and being threatened by their lawyers. IBM has made a promise not to enforce patents against open source software, but is only willing to uphold that promise as long as you don't cut into their bread and butter which is mainframe support.

rbanffy 8y ago

RAID-like memory: tolerates one - or more, IIRC - failed memory modules without stopping the process using that memory. Failure prediction and recovery: if a core presents correctable errors, it's workload is moved to a different one. The process never needs to know what happened. In the case of mainframes that have spare cores, an inactive spare is brought online automatically. Bunches of smaller computers to offload everything away from the CPUs - the main processor rarely bothers to do any IO - it's all done from separated computers with their own memory, IO channels and so on. Even powering up the machine is, sometimes, accomplished by a secondary computer that sets everything up before the actual machine is powered up. There are a lot more nice things mainframes do that x86 boxes simply can't.

larrik 8y ago

Good question. I used to sell AS/400's, which aren't quite mainframes, but fit the typical picture in your head (green screen, dot-matrix printers, high price tags and all that). 1) Stability. That stuff changes VERY rarely, and when it does it tends to be very backwards compatible. Our software had binaries compiled in the 80's still fully operational. 2) Reliability. You don't reboot the server. You don't wipe it and reinstall it. You don't even upgrade it or replace it that often. You can do things like add or remove drives from your RAID array while the machine is still at full utilization. Heck, IBM AS/400's will detect the error, order the part, and get an IBM guy there in the morning to install it. You just have to let him in the door. In fact, you can hot-swap things like the RAID controller card itself, which is WAY not something you can do on x86. They last forever, too. It wasn't uncommon for us to have machines in daily use for 10+ years without a single bit of maintenance or an IT guy even looking in its direction. This is for a business critical machine, mind you. 3) Single vendor Ever buy a Dell or HP server and have an issue, and you spend 2 weeks trying to get HP/Dell vs Microsoft vs some other company to admit fault and fix the damn issue? Then you find it online yourself because that's nuts? Well, IBM is responsible for hardware and software on these machines (besides what software you put on it) so if it isn't working, you call them and it's their damn problem to get it sorted. Kind of like Apple, except they'll send as many people as necessary for as long as necessary to sort that out for you. 4) Entrenchment It works, why bother changing it? It doesn't even slow down as it ages, either, since you aren't upgrading a GUI OS like Windows or whatever.

starsinspace 8y ago
Now this sounds fascinating about the stability and reliability, but it leaves me wondering how such decent systems can from the same company which makes the horror that is IBM Notes...
- dingaling 8y ago
  IBM was more like a federation than a single coherent entity. System/38 and AS/400 evolved in Rochester MN, mainframes in White Plains NY, and PC software all over the place. None of the divisions had much to do with each other.