![News News](http://a.fsdn.com/sd/topics/news_64.png)
![IBM IBM](http://a.fsdn.com/sd/topics/ibm_64.png)
US House Decommissions Its Last Mainframe 152
coondoggie writes "The US House of Representatives has taken its last mainframe offline, signaling the end of an era in Washington, DC computing. The last mainframe supposedly enjoyed 'quasi-celebrity status' within the House data center, having spent 12 years keeping the House's inventory control records and financial management data, among other tasks. But it was time for a change, with the House spending $30,000 a year to power the mainframe and another $700,000 each year for maintenance and support."
Change all IT geeks can believe in. (Score:4, Interesting)
From Big Iron to VMs and dedicated Unix machines.
I don't care what part of the political spectrum you fall under, that's change we can all get behind.
Unless your job was supporting old, proprietary big iron.
Re:Change all IT geeks can believe in. (Score:5, Funny)
I don't care what part of the political spectrum you fall under, that's change we can all get behind.
Congratulations! You may pick up your IT Peace Prize at the door!
Unless your job was supporting old, proprietary big iron.
Um, on second thought, never mind.
Re:Change all IT geeks can believe in. (Score:5, Funny)
Re: (Score:2)
From Big Iron to VMs and dedicated Unix machines.
I don't care what part of the political spectrum you fall under, that's change we can all get behind.
Going from VM technology to VM technology - plus ça change (plus c'est la même chose).
Re: (Score:1, Funny)
Re: (Score:2)
I don't know if this was supposed to be funny or serious.
Either way most people just shake their fist and scream, "GET OFF MY LAWN" when they want to sound ridiculously old.
Re: (Score:2)
Now for just the highlights
people IT IBM cloud OS virtual dynamically ON-DEMAND DISTRIBUTED distributed DISTRIBUTED OHHHHHH
dude you're some kind of perv...;-p
p.s. had to change some words to lowercase to avoid the yelling filter...
Re: (Score:3, Funny)
Re: (Score:1)
Someone updated computer hardware! Film at 11. (Score:5, Insightful)
Really? This is a story? They were running a server from 1997, and now they're running a server from 2009. Really guys?
Re: (Score:3, Funny)
Government 0.9.8
We would like your help in testing and improving the pre-release version, but we don't yet recommend its use in production environments.
Re: (Score:3, Insightful)
Re: (Score:2)
Re: (Score:2)
That's your problem... it's not the 4 year patches that need fixing the most... it's the 2 year patches that you keep forgetting about...
Re: (Score:2)
Re: (Score:2)
It wasn't just a mainframe, it was a quasi-celebrity mainframe. Whatever the fuck that means.
12 Year Old Mainframe = 20+ Other Servers (Score:5, Insightful)
The article notes that the House of Representatives took at least 5 years to replace the applications on its 12 year old mainframe. The costs (i.e. taxpayer funds) to perform this migration work were not disclosed, but it's a pretty safe assumption those costs dwarfed any others. Moreover, the article seems to suggest that it took at least 20 other servers to replace a single 12 year old mainframe, and that's even using virtualization on the new servers. One wonders how many (more) servers the House could have replaced with a single new mainframe.
But here's a more profound question: why is the House of Representatives running its own, separate data centers (primary and disaster)? Couldn't they at least consolidate with, oh I don't know, the Senate?!?! And, a related question: for all those 12 years, why didn't the House simply move its comparatively tiny mainframe workload to a bigger mainframe anywhere else in the federal government? (Yes, they can do that without also delegating any security control. Mainframes do that.) Quite simply, it sounds like the House was, and is, wasting a lot of taxpayer money. (Shocking, I know.)
Re:12 Year Old Mainframe = 20+ Other Servers (Score:5, Funny)
I'm pretty sure there's something in the constitution about separation of data centres...
Re: (Score:1, Funny)
No, it's separation of power supplies, so VMs were right out.
Re: (Score:2)
In theory... (Score:4, Insightful)
But here's a more profound question: why is the House of Representatives running its own, separate data centers (primary and disaster)? Couldn't they at least consolidate with, oh I don't know, the Senate?!?!
I don't know. I kind of like the current situation: Two different significantly powerful political entities (House and Congress) to have their own separate data that the other entity has no control over. I could certainly see potential benefits from that in the times of major political upheavals.
Re:In theory... (Score:4, Informative)
Re: (Score:1)
Re: (Score:1)
Lets see, Congress is the opposite of Progress.. Yeah, that's it!!!
Re: (Score:2)
You can have them as far as I'm concerned. As long as you promis to take them to another planet.
Re: (Score:1)
Re: (Score:2)
When I was in the Government, replacing local computing resources with centralized resources was always pushed as a cost-saving move, and always cost more money. Always.
Remember, there is no competition in the Government cloud. (Having one big mainframe somewhere was the 20th Century version of cloud computing.)
Re: (Score:2)
When I was in the Government, replacing local computing resources with centralized resources was always pushed as a cost-saving move, and always cost more money.
Similarly in the business world. The basic problem is that a mainframe requires a professional staff to support it, and that always turns into a department that is its own separate power center. The computer department (under various names) invariably puts strict controls on how their equipment is used, and if they don't approve of what you're try
Re: (Score:2)
Sometimes it's beneficial to have a lot of smaller servers vs a big consolidated one. It's the same principle that's behind RAID: you independently upgrade, repair, replace the smaller servers and it's easier to add capacity. Also mainframe expertise is a dying art and so will get more expensive, so in the long run maintenance costs will probably be lower and access to skilled administrators assured.
Re: (Score:2)
Re: (Score:2)
And right there we have the answer to the earlier comment about "... there's something in the constitution about separation of data centres ...".
In management circles, worship of IBM (and His Son Microsoft) now qualifies as an ancient religion that is sorely challenged by the flock of Protestant religions based on various unix (and its bastard offspring linux) computer systems. The US Constitution is clear in forbidding Congress to make any
Re: There's more to life than direct costs.Moreove (Score:4, Insightful)
Moreover, the article seems to suggest that it took at least 20 other servers to replace a single 12 year old mainframe, and that's even using virtualization on the new servers. One wonders how many (more) servers the House could have replaced with a single new mainframe.
Talking about costs makes good news. If you want to pick it apart with speculation, go right ahead. You might be right, but without real numbers and real analysis we'll never know. I think the real reason they switched is this:
How many people have you known throughout your IT career that have mainframe experience? I've known exactly two. One of which was my next-door neighbor while growing up who worked as a programmer for Unisys (now retired). The other was a young kid who was hired by company who still had mainframes, and was trained by the old guys who knew how to operate and maintain them. The other several dozen people I've known throughout my career have no real world experience with maintaining them. I knew one guy who had to learn AS400 in tech school (this was only maybe 10 years ago), though never applied any of it and now works with Windows and Unix.
So the point is, how well do you think a piece of technology is going to work if you can't find anyone who knows how to work with it and maintain it? Sure, salaries are cheap in comparison to migrations.. but what are the costs of not being able to do what you want to do because you can't find enough people familiar enough with the technology to accomplish what you want?
Technology always has been, and always will be about the people. Someday all our modern technology, operating systems and the like will die not because it's not "good enough", or is "too expensive" but because the people of that era will have moved on to New Technology Z, and hardly anyone understands Old Technology Y.
Re: (Score:2)
While it's apparently a dieing concept, that's what on the job training used to be about. If you have even one well skilled mainframe person and someone else from the *nix world, you can have two skilled mainframe people in short order.
Re: (Score:1)
If you have even one well skilled mainframe person and someone else from the *nix world, you can have two skilled mainframe people in short order.
I've always wondered how mainframe people reproduce.
Linux on the mainframe (Score:2)
It's worth remembering, though, that they could have chosen to run up to 60 Linux LPARs on a mainframe. That way they'd get the five nines reliability of mainframe hardware, the lower power requirements and reduced maintenance pain from having everything on a single box, and they'd be able to maintain it all using familiar Linux tools.
There's also a 31 bit version of Linux for System z, so they might not even have had to replace the hardware.
I hope someone did some solid cost/benefit analysis.
[Opinions mine
Re: (Score:1)
That still requires in-house mainframe skills - someone still has to manage the hardware.
Re: (Score:2)
My understanding is that typically IBM manages the hardware. That's what the service contracts are for. When a drive fails, an engineer turns up and replaces it for you. You don't even have to call, the mainframe detects the failing drive and calls IBM for you.
Re: (Score:1, Insightful)
Did you only read what you wanted to from the article? The 20 servers replaced 150, not this one server. Also, the reason they did not stick with mainframe technology was also mentioned. It was because they had limited skills in mainframe technology. Thus keeping mainframes, they'd have to hire people trained and experiences in mainframes (a dying breed). Please read the whole article and not just pick and choose the words you like.
Re: (Score:2)
Moreover, the article seems to suggest that it took at least 20 other servers to replace a single 12 year old mainframe, and that's even using virtualization on the new servers.
Replaced is such a misunderstood word. The new servers would be way faster than the old clunker. The only replacing is the physical displacement.
That leads one to question. If the old machine was measured for performance and an equivalent performing set of servers were specified, what would be the calculated outcome?
One wonders how m
Re: (Score:2)
That was by design. Then the sysadmin died under suspicious circumstances right before testifying about it.
Re: (Score:2)
That's the first I've heard of anything like that, and a quick Google search for "white house sysadmin email" didn't turn up anything interesting (except for a whistleblower study from 2000 about then-Vice President Gore using the White House email system for unethical campaign work). Got any references for this allegation?
Re: (Score:2)
$700K/yr not out of line (Score:5, Insightful)
$700K/yr for software support and hardware maintenance isn't really out of line for a high-capacity system with 99.999% uptime.
Maybe they don't need that level of reliability, but if they do five-9s, they will probably find that whatever system or group of systems replaces it will have similar support costs.
Re: (Score:3, Insightful)
Sadly though, many of the management types just read "$700K/yr" and think "I like $350K/yr better, let's do that". Ignoring what they're actually getting for that money.
It would be no problem to cut the cost by half for the mainframe... if you're willing to go down to average Windows server service levels.
$700K/yr gets you how many people with how many workstations, hardware, software, facilities, managers, support people, etc.?
Re: (Score:3, Funny)
The worst thing is, going from $700k a year to $350k a year doesn't just halve your uptime, it takes it from 99.999% to 99.99%. Or from 52 minutes to eight hours and 45 minutes. The first time they can't access their records for an entire work day, maybe they'll realize what they were paying for before.
Re: (Score:2)
Those numbers are typically measurements of actually achieved uptime over a specific period of time. So those aren't just some random marketing numbers but real percentages relating to real time. Perhaps it's different in other classes of server hardware though.
With mainframes, the people making the purchasing decissions generally know what they're talking about, so lying in marketing isn't a smart idea.
Re: (Score:1)
Re:$700K/yr not out of line (Score:4, Funny)
Re:$700K/yr not out of line (Score:5, Informative)
Unfortunately the article doesn't mention what exactly makes the $700k. I'm not into mainframes, maybe someone else has details on what makes mainframe maintenance so costly?
It's the 99.999% uptime that is a typical requirement of mainframe apps. That means things like remote monitoring by the vendor via a direct link to the system so that the diagnostic subsystem can tell the vendor that parts are failing before they fail and then the vendor will usually have a 4 hour or less requirement to get new parts on site, the logistics of which are a lot more expensive than they appear at first glance (gotta have local hardware depots with enough spare parts to cover all contingencies, including multiple simultaneous failures at multiple sites at different side of the city, etc). Then there is the cost of the human expertise - mainframe customers expect 1st-line support to be one level away from engineering - absolutely no scripted phone support weenies. The on-site hardware techs are also a couple of orders above the typical vendor hardware tech who is frequently a jack-of-all-trades and master of none - the mainframe guys are dedicated to mainframe support and are typically on a first-name basis with the engineers who designed the hardware.
So in summary - extremely rapid response plus top-flight human talent equals big bucks.
The article did say that the mainframe was old and thus support costs were even higher which is common - as hardware is obsoleted it becomes more and more expensive to stock replacement parts (and engineering staff). So maybe they could shave a hundred grand or two off that price if they were using a mainframe that had not been end-of-lifed a while ago.
had to because ibm dropped pre-64bit compatibility (Score:1, Informative)
The article says the retired mainframe is from 1997 so it's almost certainly a G3 or G4-based model.
If it was a G4 then replacing it with a mainframe that wasn't end-of-lifed would have entailed all the complications of migrating mainframe apps from 31-bit to 64-bit that are described in this article http://www.computerweekly.com/Articles/2003/12/15/199268/mainframe-users-face-upgrade-dilemma.htm [computerweekly.com]. If it was a G3 then they would have faced even more obstacles...
Re:had to because ibm dropped pre-64bit compatibil (Score:2)
The article says the retired mainframe is from 1997 so it's almost certainly a G3 or G4-based model.
If it was a G4 then replacing it with a mainframe that wasn't end-of-lifed would have entailed all the complications of migrating mainframe apps from 31-bit to 64-bit that are described in this article http://www.computerweekly.com/Articles/2003/12/15/199268/mainframe-users-face-upgrade-dilemma.htm [computerweekly.com].
The article speaks of a "31-bit compatibility mode" in z/OS, but that's a mode that allows it to run on the 31-bit-address/32-bit-data pre-z/Architecture machines. Dropping that would be the equivalent of Windows 8 requiring x86-64 but still running 32-bit apps, or {pick your distribution} dropping its i386 version but still supporting 32-bit apps, or OS X 10.n, for some value of n > 6, running only on x86-64 Macs but still running 32-bit apps, or....
The article also quotes somebody as saying "Migratin
Re: (Score:1)
Re: (Score:2)
$700K/yr for software support and hardware maintenance isn't really out of line for a high-capacity system with 99.999% uptime.
Maybe they don't need that level of reliability, but if they do five-9s, they will probably find that whatever system or group of systems replaces it will have similar support costs.
They probably don't need 99.999% uptime, considering the House schedules itself for downtime 2 months out of 12. :)
Re: (Score:2)
Re: (Score:2)
It only needs to operate when Congress is in session.
Five twos of up time should be sufficient.
Re: (Score:2)
Yeah... that's not really how it works.
Re: (Score:2)
Really? Perhaps you could explain it to me while I fix your irony detector. :)
Replacement? (Score:4, Insightful)
Whereby the underlying question needs to: "Hopefully it's not replaced by a Microsoft 'solution'".
Re: (Score:2)
The congress critters have learned a lot from the "terrible mistake" of email backups.
Re: (Score:2)
As long as the software itself isn't proprietary, I have absolutely no problem with the government using a Microsoft solution.
Use the right tool for the job. In the hands of a competent admin, Microsoft's server offerings aren't half-bad. Microsoft's licensing fees are only going to be a drop in the bucket for a datacenter that evidently has no problem forking over $700k per annum on maintenance. (In fact, for a system of this size, Microsoft might even bid lower than Red Hat or Novell on a project of th
Support costs (Score:4, Interesting)
Wow, that's some support bill - or was the House doing the usual Government thing and buying its 'Government Grade' punch tapes at $500 for a pack of 5!? I expect there was also one very highly paid guy who was the only living person left who knew the correct sequence of toggle key entries to start the IPL!
A corporate for which I once did some consulting was transitioning their code from an ancient mainframe to a group of PC-based servers. For some bizarre reason, the Company decided to make their in-house hardware engineer/support guy redundant BEFORE they had finished the change-over! Suffice to say, his consulting daily (or part daily) rate to come back in and kick the old system back to life as needed during its last few months was *very* high as he was the only one who knew how to sweet-talk some of the bespoke hardware. I heard of times where a 10-minute site visit was netting him a 1/2 day fee of something mad like 700UKP
No doubt the House mainframe's replacement is the $900 Dual Xeon unit previously used as a front-end processor for the mainframe's 32-port serial mux!
Re: (Score:1)
Re: (Score:2)
For some bizarre reason, the Company decided to make their in-house hardware engineer/support guy redundant BEFORE they had finished the change-over! Suffice to say, his consulting daily (or part daily) rate to come back (...) was *very* high (...). I heard of times where a 10-minute site visit was netting him a 1/2 day fee of something mad like 700UKP
In other words, he sold himself cheap. There was recently a big case in media now about the media consulting bill after a big scandal that they were called in to handle, the leading senior advicer - and he really is senior though - was 3500 NOK/hour or about 390 GBP/hour, that's his standard rate. Noone disputed the prices, they were just arguing over who would be paying the bill. If I got laid off and you wanted to hire me back in, I think me "WTF you got to be kidding me" rate would be even higher than th
Re: (Score:2)
Re: (Score:2)
I believe it was replaced with 20 servers.
Re: (Score:2)
No doubt the House mainframe's replacement is the $900 Dual Xeon unit previously used as a front-end processor for the mainframe's 32-port serial mux!
Actually, I'll bet it's some sort of Sun/IBM blade cluster running redundant VMs. Gives you many of the advantages of a traditional mainframe, but offers much more flexibility in terms of hardware. One of the coolest things about server virtualization is that your applications and operating systems become completely hardware-agnostic.
It also eliminates the need for any huge "transition" when the system eventually becomes obsolete. Individual applications and blades can be retired/upgraded as is necessary
Full circle ? (Score:2, Insightful)
From what I've seen lately, virtualization is kicking in even on desktop field and normal PC's are being replaced with more power efficient thin clients.
I know thin clients aren't same as simple terminals were with mainframes since they connect to the vm-servers using gigabit ethernet instead serial cable, and instead serving unix shell, they now provide entire desktop experience to end-user.
But what's interesting for me is to see if the thin
Re: (Score:3, Insightful)
IMO what drove the cycle to favoring desktop PC's away from dumb clients was being able to have applications with around as much power, only with a vastly richer user experience. I.e. there was a cycle at all because it replicated mainframe's positive and replaced mainframe's negative. I'm not sure what the impetus could be for another full cycle. Security management considerations and the ability to serve up that rich user experience might induce another half cycle. But it would take something that only de
Re: (Score:1)
Re: (Score:2)
Flashy graphics might look sexy, but that wasn't what sponsored the move away from a mainframe-centric organization. The mainframe programmers were not able to provide data in a form the users wanted it in a timely manner. In most organizations the users would ask for a 10,000 page report to be printed out in 30 minutes, look at the first page and say "That isn't really want I wanted" and start the whole process over agin.
PC software vendors, including Microsoft, would woo users with cute demonstrtions of
Re: (Score:2)
We've proven that the people are by and large too stupid to have complete control over their "PC" in a networked environment. So, why not have a managed thin client instead of a desktop? Anywhere that the unit is captive to an enterprise / office / school, F the "PC" model. It's done. Waste of power to run. Waste of resources to build each unit. Waste of money to manage and troubleshoot the 'workstations'. I laughed the first time I heard a "PC" called a workstation. The fad is coming to an end.
Of course, d
Re: (Score:2)
Mainframes aren't going away. Not if your business is processing credit cards because virtualization is not PCI-Compliant. (We can debate the rules and how stupid they are, but as it stands right now those are the rules). Neither is any "cloud" services. When we looked, none of them could give us in writing that their services were Level-1 audited and certified compliant. Now, if you're a Level 2 - 4 merchant, not a big deal. But if you're a level 1 merchant and processing enough transaction, the extr
Units (Score:5, Insightful)
Uh... what? No wonder they had to pull this thing offline, that's 1.68 - 2.52 GW per week!
It's been online for 12 years, so by the time it was shut off it must have been using at least 1.57TW.
Re: (Score:1)
Re: (Score:1)
That paragraph isn't quoted, so it may have been the author of the article that mangled the units (really, if he had the faculty to notice the problem, he should have gone back for clarification, so he is mostly responsible anyway).
Re: (Score:2)
And the article talks about its size: 8 cubic feet. Even if this number doesn't include storage, it is surprisingly small.
Considering the usual government lag (Score:3)
Keep the Mainframe (Score:5, Funny)
Decommission the representatives. Then put the mainframe in charge. I'm sure it is much more efficient at processing bribes, though it probably lacks sex scandal capabilities.
Re: (Score:1)
I wouldn't be so sure. Any port in a storm, IYKWIMAITYD. ;-)
Re: (Score:2)
Well mainframes take up a lot of physical space. It wouldn't be unheard of if one had a wide stance...
Re: (Score:2)
it probably lacks sex scandal capabilities
They could interface this [welookdoyou.com] to it. (Warning, NSFW).
Re: (Score:2)
Decommission the representatives. Then put the mainframe in charge. I'm sure it is much more efficient at processing bribes, though it probably lacks sex scandal capabilities.
While Rule 34 exists, don't be so sure.
"Congressional mainframe found downloading drawings of it raping schoolgirls."
Re: (Score:2)
Are you familar with the movie "Terminator"?
Re: (Score:2)
though it probably lacks sex scandal capabilities.
A daughter card provides that additional functionality.
Is it really more efficient? (Score:2)
Werent there some studies shown that a mainframe was actually more energy efficient than a cluster? I wonder if they did any actual scientific studies of efficiency or if the people in charge here just made some assumptions, and went along with the hype.
Re: (Score:2)
Werent there some studies shown that a mainframe was actually more energy efficient than a cluster?
Studies have shown that marketers are 75% more likely than the average person to use the word "myth."
Government 2.0 (Score:1)
So what was it? (Score:2)
What kind of box was it? My guess goes for a VAX...
Re: (Score:3, Insightful)
No, it was an IBM (RTFA).
As for price, well, some people have money, earn money using the tool that a computer is, and consider *correct* performance worth their money.
Seen from a business view, it is *good* to know that if your system breaks down there *will* be a really competent guy (or two) on site in less than a hour. I've seen it happen. At 2300 hours on a Friday evening. It is *good* to know that if something really bizarre happens, and the front-line guys really don't know what to do, mobile phones
Re: (Score:2)
Redhat does sell support contracts. I hear they make a good deal of money from it too.
Re: (Score:2)
Re: (Score:2)
Bet IBM is mad (Score:2)
I bet IBM is mad they didn't spend more time on their BlackBerry integration piece to PROFS.
Final score
IBM: 0
Obama's BlackBerry: 1
It won't be the last (Score:1)
Like the moon-landing missions, it's just the last remaining one.
The age of the mainframe will rise again!
OK, seriously, I don't know if the mainframe will come back to the House of Representatives, but mainframes are particularly good at many 21st-century tasks, and it would be just as arrogant to presume they won't ever come back as to insist, as I jokingly did above, that they will.
real news is (Score:2)
They replaced it with an atom powered net-top -- twice the power
Re: (Score:2, Interesting)
The Cloud = Mainframe V2.0?
'scuse me while I try and calm down my hysterical laughter.
(what U smokin'?)
Re: (Score:2)
Re: (Score:1)
The Cloud = Mainframe V2.0?
'scuse me while I try and calm down my hysterical laughter.
(what U smokin'?)
The concept is almost the same. Just switch away custom-made hardware for commodity hardware running clusterized applications and you have a sort of "virtual mainframe" called "cloud" just because it isn't in your datacenter. The final costs can be just as high, it all depends on how good at negotiating agreements the company is.
Re: (Score:2)
when their dedicated cluster of Walmart E-machines dies on the next Windows update reboot cycle... haven't they heard of Bloated Capacitors!!!!
"Bloated Capacitors? Didn't they play just before the Stones at Altamont?" is about the level of reaction I'd expect from congresscritters.
Re: (Score:2)
I'm sure the US government will just as efficient running our health care system :(
*sarcasm*You racist!*sarcasm*