Not your grandfather's IBM

[webfusion]

The z10 mainframe from IBM was just introduced.

DOTD. $1-million.

http://www.physorg.com/news123230002.html

Sweeeeet.

[WildCat]

Feh. I'm not getting rid of my Eniac just because some newfangled machine comes along. :p

[Zep]

The new System z10, with a starting price at about $1 million, comes as IBM focuses on lowering the price tag for running its storied line of data-crunching workhorses.Oh, that's good of them! I should be able to afford a whole bunch of them now. :rolleyes:

[Paul C. Anagnostopoulos]

Feh. I'm not getting rid of my Eniac just because some newfangled machine comes along.
You have an Eniac?! Cool. I'm stickin' with my Z1, though.


Oh, that's good of them! I should be able to afford a whole bunch of them now.
They weren't made for you. You have to buy the equivalent 1,500 servers, which will only cost ... uh oh.

~~ Paul

[ddt]

Feh. I'm not getting rid of my Eniac just because some newfangled machine comes along.
You have an Eniac?! Cool. I'm stickin' with my Z1, though.
Z1? That relies on this un-robust modernity called electricity.

By the weekend, I should have this Analytical Machine up and running. Now only if Ada would also finish her programming...

[gnome]

Rubbish! I have everything I need right here!

http://www.bsutton.com/bill/Filk/diyAnnotated.txt

[Zep]

Rubbish! I have everything I need right here!

http://www.bsutton.com/bill/Filk/diyAnnotated.txtPPhhfffftt!!

Real programmers...

http://upload.wikimedia.org/wikipedia/commons/thumb/c/cc/Quipu.png/250px-Quipu.png

[ElMondoHummus]

Feh. I'm not getting rid of my Eniac just because some newfangled machine comes along. :p

You have an Eniac?! Cool. I'm stickin' with my Z1, though.


They weren't made for you. You have to buy the equivalent 1,500 servers, which will only cost ... uh oh.

~~ Paul

Z1? That relies on this un-robust modernity called electricity.

By the weekend, I should have this Analytical Machine up and running. Now only if Ada would also finish her programming...

Is this where I point out my Turing Machine's* currently clicking away, entering posts in the JREF forums as if it were me (Bom bom BOM :eye-poppi).

Ask me a question, see if you think I'm huuuuumaaaan :eek::boggled::D

*Yeah, I know Babbage's machine is older than Turings. I can't top it! Whaddaya want, an abacus joke? Gimme a break, will ya!

[Zep]

*Yeah, I know Babbage's machine is older than Turings. I can't top it! Whaddaya want, an abacus joke? Gimme a break, will ya!I already done it. Move on now. ;)

[GreNME]

Oh, that's good of them! I should be able to afford a whole bunch of them now.
They weren't made for you. You have to buy the equivalent 1,500 servers, which will only cost ... uh oh.

Yeah, my thought exactly.

Exactly what is this thing supposed to replace? I understand the functionality, but at that price it's completely impractical.

[Wudang]

Well the reduced energy footprint is great. Our z9s are creaking at the seams and we need more power. The racks of blades eat cooling and power. 50% performance improvement? Great.

[The Central Scrutinizer]

The z10 mainframe from IBM was just introduced.

DOTD. $1-million.

http://www.physorg.com/news123230002.html

Sweeeeet.

The article doesn't say how many vacuum tubes this baby has. I bet it's a lot!

[ElMondoHummus]

Yeah, my thought exactly.

Exactly what is this thing supposed to replace? I understand the functionality, but at that price it's completely impractical.

Virtual serving is the thing that comes immediately to mind. If you have an huge environment and an enterprise app that is best stretched across different servers for whatever reason, then you can virtualize one of these boxes into what, 4, 5, 10, whatever virtual servers and deliver that app's functions quite well. And virtualization's great for a variety of reasons: Easy recovery if one virtual environment gets corrupt (or "pwnd"), ridiculously easy application of patches and upgrades across your service, etc. And IBM's glory (i.e. announcement/advertising) page on this box specifically mentions virtualization technology; enterprise customers looking for powerful boxes to create virtual server farms in are a prime target to market this thing to.

I also imagine raw numbercrunchers looking to build supercomputing clusters would take a look at this box when shopping, too.

Anyway, those are two applications that immediately come to mind. Does that help?

[madurobob]

Yeah, my thought exactly.

Exactly what is this thing supposed to replace? I understand the functionality, but at that price it's completely impractical.

You might be surprised how many major datacenters there are that could benefit from this. Huge scalability, bullet-proof design, one box replaces several.

One major benefit is the ongoing costs. Most companies running large datacenters have no problem plunking down a lot of money on new equipment. Its the ongoing cost of running the equipment that bugs them. Think software. Companies may be able to consolidate multiple boxes into one Z10 and have more total computing power (MIPS/MSUS) but a lower annual SW charge.

So, spend some money up-front and reduce your annual bleed. Thats a big selling point.

[madurobob]

I also imagine raw numbercrunchers looking to build supercomputing clusters would take a look at this box when shopping, too.

Perhaps, but I don't think the Z10 will compete well with the BlueGene. I think they're up to 2,064 quad-core processors in a single BlueGene stack now.

[ElMondoHummus]

Perhaps, but I don't think the Z10 will compete well with the BlueGene. I think they're up to 2,064 quad-core processors in a single BlueGene stack now.

Oh, really? Wow, that's one hell of a system (http://en.wikipedia.org/wiki/Blue_Gene). Okay. Well, supercomp, number-crunch, simulate universe :D apps are sorta out of my direct experience. But virtualizing enterprise apps is something I see all the :rule10 time, and I can imagine a couple of our service admin teams drooling over what they could accomplish with this box.

Hey, if it reduces downtime and keeps people from calling my team to bitch, I'm okay with it. :D

[madurobob]

Oh, really? Wow, that's one hell of a system (http://en.wikipedia.org/wiki/Blue_Gene). Okay. Well, supercomp, number-crunch, simulate universe :D apps are sorta out of my direct experience.

Yah. They've developed a nine-core chip that is in test phase for BlueGene. 8 processing cores plus one "master" core to coordinate. I think LNLL (big BlueGene customer) is building a time machine, or maybe an entire alternate universe on the computer.

[Zep]

I can see that it's a great way to get a helluva bang for your buck, especially if it is a virtualised setup. The problem with having your own personal universe on one machine is exactly that - it's all on one machine. Turn off the cooling water (or freon or liquid sodium or whatever you use), or not control the power adequately, or poke any of a myriad of other soft spots we could identify, and your personal universe gets shunted. If virtualised, everyone else's does too. At least spreading the load out over multiple machines tends to minimise universe collapse so completely.

Apart from that, it's a real sci-fi design, is it not! And IBM always knew that managers were always impressed with flashing lights!

[Gord_in_Toronto]

PPhhfffftt!!

Real programmers...

http://upload.wikimedia.org/wikipedia/commons/thumb/c/cc/Quipu.png/250px-Quipu.png

You smartass Incan programers managed to encode your work with a code so good that centuries later we have not been able to interpret it. :mad:

[GreNME]

You might be surprised how many major datacenters there are that could benefit from this. Huge scalability, bullet-proof design, one box replaces several.

One major benefit is the ongoing costs. Most companies running large datacenters have no problem plunking down a lot of money on new equipment. Its the ongoing cost of running the equipment that bugs them. Think software. Companies may be able to consolidate multiple boxes into one Z10 and have more total computing power (MIPS/MSUS) but a lower annual SW charge.

So, spend some money up-front and reduce your annual bleed. Thats a big selling point.

I understand the concept, but that price is way above comparable setups. I admit, that seems like a virtualization dream setup, but that price keeps tripping me up.

[madurobob]

I understand the concept, but that price is way above comparable setups. I admit, that seems like a virtualization dream setup, but that price keeps tripping me up.

Well, yeah, but keep in mind that high-end servers are priced like high-end... anything else. There is often quite a gap between list price and street price.

[Zep]

You smartass Incan programers managed to encode your work with a code so good that centuries later we have not been able to interpret it. :mad:Like the Next box, look what good it did us! :rolleyes:

[GodMark2]

You smartass Incan programers managed to encode your work with a code so good that centuries later we have not been able to interpret it. :mad:

You call it obfuscated code: I call it job security.

[GreNME]

Well, yeah, but keep in mind that high-end servers are priced like high-end... anything else. There is often quite a gap between list price and street price.

In a company of sufficient size to make use of this, the bottom line would clearly come into play. Starting from zero and building up, I can see the possibility (though still not certainty) of ROI within five years on a purchase like this. For a datacenter that's already established with infrastructure, though? Not seeing it besides the "wow" factor. The best I can see for them is with leasing, where they can either be replaced when better tech comes in at a better price or having the option to buy outright at a huge discount a few years down the line.

That said, it is impressive. :)

[Zep]

You call it obfuscated code: I call it job security.Look, ma! No comments!

[Wudang]

For a datacenter that's already established with infrastructure, though? Not seeing it besides the "wow" factor. The best I can see for them is with leasing, where they can either be replaced when better tech comes in at a better price or having the option to buy outright at a huge discount a few years down the line.

That said, it is impressive. :)

It's only a million. I don't know the financial model used for z-series at the bank I work for but I know that over 1 year a couple of DBA's saved nearly 600k expense avoidance in terms of kit we didn't have to buy by db and app tuning. For big companies the need for IT contiunally grows and capacity requirements grow not just with new apps and business opportunities like data mining customer profiles but compliance with SOX, Basel 2, etc.
Or absorbing shed loads of work from the US because the Patriot Act means it can't be processed on US soil any more.

[GreNME]

It's only a million. I don't know the financial model used for z-series at the bank I work for but I know that over 1 year a couple of DBA's saved nearly 600k expense avoidance in terms of kit we didn't have to buy by db and app tuning. For big companies the need for IT contiunally grows and capacity requirements grow not just with new apps and business opportunities like data mining customer profiles but compliance with SOX, Basel 2, etc.
Or absorbing shed loads of work from the US because the Patriot Act means it can't be processed on US soil any more.

I'm not doubting you, but for the type of equipment this is supposed to be replacing it seems to have an unnecessary premium on it, especially for replacing older-but-not-too-old models. If this thing were replacing a 15-year-old farm, I can see how it would even out with ROI. Maybe that's who this is marketed toward. Heck, just a few years ago a company I did some contract work with before unloaded a bunch of old hardware-- in the process someone I know who worked there at the time managed to snag an old AS/400, though they unloaded the bigger stuff to other places.

Maybe I'm just looking at this too much like end user systems, and that's skewing my opinion. I've dealt mostly with user systems and smaller servers so much over the last few years it's just habit.

[Wudang]

Maybe I'm just looking at this too much like end user systems, and that's skewing my opinion. I've dealt mostly with user systems and smaller servers so much over the last few years it's just habit.

I think that's part of it and possibly you're not aware of the real issues confronting many big data centers that got too many blades in too quickly without appreciating just how their cooling and power needs were going to skyrocket.
Like people comparing to BlueGene when they're entirely different kinds of boxes for different markets.