Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Nothing wrong with using old hardware until it dies (provided you have appropriate redundancy, of course).


I am still running an HP K380 server with the HP-UX 10.20 operating system. It has 250mhz risc CPUs and takes 220V power.

This server is likely on the far side of "[nothing] wrong," and we had it turned off and ready to e-waste.

However, the effort to rewrite the code that runs on this machine had some surprising setbacks, and I had to wheel it back in. It will likely be here after I'm gone.


The way you push to migrate (even moving to a QEMU-hppa binary emulation environment) is the following equation:

cost(maintain) is the cost of keeping the system running as it stands right now, with the risk of 100% "it's gone" loss.

cost(recover) is the cost of rebuilding what you have from wreckage, including any monetary loss from SLAs.

cost(port) is the cost of operating plus the cost of ensuring compatibility at some level (or just "rebuilding it for functional equality"), with the inherent risk of Second System Syndrome.

when cost(port) + cost(maintain) < cost(recover) by some factor (say, cost(port)^2 ) then it's time to port the system in-place after documenting its behavior as closely as you can. If you have source, that's great -- it's a UNIX system, you should be able to dredge most of it over and might need to patch in parts here and there.


Donate it to a computer museum instead. It’s an interesting machine, from a dead lineage.

Or sell it on eBay. HP/UX boxes tend to be expensive, for reasons I can’t really understand


> HP/UX boxes tend to be expensive, for reasons I can’t really understand

Probably:

> However, the effort to rewrite the code that runs on this machine had some surprising setbacks, and I had to wheel it back in.


You can still emulate them.


That is part of a solution but that emulation still has to run on something. Getting that emulation hardware certified for the application could be prohibitive. Are there drop-in replacements for the hardware that perform equally to the HP boxes?


If there are legal requirements for certification, then you are out of luck, but you can probably convince HPE to do a pinky promise their new boxes will perform as well as the old ones.


I have, of all things, a logic analyzer that runs HP/UX 10.20 (HP 16702B). Of course, it's not on unless I'm actively using it, so I don't have to be concerned about the power consumption, unlike a K-box.


At some point, power efficiency may become an issue.

Let's say you server uses 100W, continuous and runs 24/7, that's 876 kWh per year, at $0.20/kWh, that's $175 at the end of the year. Something like a Raspberry Pi is maybe $100 with shipping and accessories, and uses less than a tenth of that power, the hardware will pay for itself in less than a year.

Note that the "old hardware" is an enterprise class, 10 year old file server. Hard to replace with a Raspberry Pi, but some old servers could be. Virtualization is another option, a common one for businesses.


I agree with you in general but at some point the less efficient older hardware costs more to do run than something newer and more efficient. It doesn’t have to fail to be obsolete. This is a simple finance and accounting exercise.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: