Lukas Beeler |
Were there even rackmount AS/400 models? There sure are rackmountable System i5′s, but i don’t know about the AS/400 models
Clusters/HA solutions in general add a vast amount of overhead – my general advice is to stay away from them as long as possible. If seen several cases where outages where causes by the HA system itself and user errors, sometimes even taking backups to recover.
Matt Stansberry |
Yeah, the discussion didn’t deal with any specific hardware models. They could have been referring to AS/400 generically, and there were no specifics on the Dell servers either. But the exercise was pretty interesting.
Open mouth insert foot and swallow hard
Jay Benton |
HA solution overhead entirely depends upon the package.
We run a solution that put overhead on our model 170′s that resemble a one or two programmers compiling programs.
They didn’t mention the iSeries (AS/400) stability and no viruses…. figures.
I guess it comes down to what you need the setup to do. For the most part many small businesses are better off with PC servers. When it gets to large scale computing your more focused on the software you want to run.
Frankly, IBM just keeps kicking themselves when it comes to iSeries. We moved off quite a few iSeries because of green screen tax. I guess IBM just thinks pissing off customers is a good idea. I hope they lose more business, might just wake them up. (as in, there are many applications where green screen (read:interactive) are needed and punishing that group because of come damn corporate hardon for the web even when its not applicable to many businesses is just idiocy)
Jerry Balon |
The IBM midrange has always been a good maxchine that keeps getting better. When you talked AS400 I asume you were talking about the iSeries and new blade servers also.
What I would like to see is a true comparison of cost/performance benchmarking between the two options. The thing with iSeries pricing is that the cost of the iSeries gear typically includes the software licensing cost for an enterprise class Database system (DB2/400), Webserver (Apache – okay that one is free on linux, but really free at Enterprise class levels with enterprise class support?), and Application server (WAS).
To really do this comparison, you would have to analyze the costs of the two options fully loaded with all of the associated software.
The iSeries is indeed incredibly fault tolerant and stable. But from a pure dollars and (common)cents perspective, I think the “60 Dell Servers” guy was close to the mark. Based on his “30 times the throughput” comment, you could buy 10 Dell servers and get 5 times the throughput. With the money left over you could buy Oracle’s database, the rest of the associated software, and still have enough money left over to hire an army of IT folks to do the cluster support and maintain uptime.
If IBM disagrees with my point, they should post price comparisons to prove me wrong. That said, I don’t think they will because they already know the answer.
I’m with Lukas – get your nomenclature correct.
What you are talking about is an i5 – a System i branded server. We need to get the word out that the server is modern, and the most powerful system on the planet. No one will believe an AS/400 can do any of those things – since it is so OLD.
Please use the correct names for the server (i5), the OS (i5/OS) and the brand (System i).
We have 10 Dell Servers and two AS/400s(Soon to be I5). The AS/400 places its own service calls(Only 1 in the last year). We have had two drives crash on a year old Dell Server, both within 2 weeks of each other. I’ll keep my domain on the Dells, but the applications that pay the bill on the AS/400.
I’ve never had an AS/400 or an i-series let me down. I can’t say that about DELL machines. In fact I’ve actually celebrated the retirement of a few DELL machines, they are the only hardware I’ve ever retired before they were obsolete.
I’ve never been disappointed in IBM support, I can’t say that about DELL support.
I have held several roles in the AS400/iSeries/i5 world. I have also been a datacenter manager with logistical responsibility for 1000+ intel servers. I can tell you in full confidence that I would rather have to worry about an i5 machine than any number of intel based servers.
I hope the numbers in this article aren’t wasted on most people 60 servers vs 1 i5. The environmental cost and administrative cost make this a lot closer race than most would imagine. Couple that with the stability of the i5 platform and the throughput numbers don’t stand out so far.
As an old AS/400 guy who is now into Java and Web apps, I have to say that each approach has its advantages. But “cost” is not limited to the hardware and software. 60 servers, about the number in our computer room, requires more than one 12 foot long cooling unit. Add on the cost of the various admins to keep everything going, and the cost picture becomes complicated. Also, a huge cluster for what? Just database? What about clusters for you Web servers? By the way, the “green screen tax” went away years ago, for those who don’t need interactive. Bottom line, a huge, dedicated database makes sense on a cluster. Diverse databases that go along with diverse applications, AS/400.
We have both: Dells and one AS400 which we exchanged for a new i5. I am encharged of about 20 plus Dells and I am constantly dealing with crashed drives and rebuiliding arrays and the regular OS in addition to application maintenance. But we’ve had only two service calls in the last 5 years for the 400. Alhtough we never had a complete server failure due to HW in the as400 and Dells. To accomplish this we have a backup dell for each critical server which doubles maintenance and cost, but we only have 1 as400. However, we only run one financial application on the 400 as opposed to the plethora of enterprise applications required for a company.
There are some people who argue in favour of a PC-solution because of the green screens. This, in my opinion, is a non-argument and I will try to explain why.
The interface to the operating system is for maintenance personnel only, they need some interface to communicate with the OS and more than often they will prefer a (granted bare) fast way of doing so to a GUI in which they have to click their finger-tips blue to get to where they want to be. If however they prefer a GUI: IBM has friendly enough to supply ample means by also delivering the command-centre (or which ever name they attached to that).
Arguing that the server is not usable for companies simply because it doesn’t have a GUI is a bit moot. The server needs to do what it is supposed to do and that is offer a platform on which to run the applications needed (and offer peripheral services like a printer server and a RDBMS). These applications can be written to have a GUI, and IBM offers the companies all possibilities to write these GUIs.
By the same token Linux as an OS has no GUI either, it offers several GUI interfaces with Gnome, KDE, fluxbox etc. on the X-framework and no distribution offers the GUI to the applications out of the box either, and for that matter, neither does Microsoft Windows. All of the platforms need to be configured (read: applications need to be written) to offer the end-user an GUI and an application to work with. Getting rid of the AS400 just because it doesn’t have a native GUI is the same as getting rid of a house because it doesn’t have a certain colour wallpaper. No house will have your colour wallpaper in it unless you put it there. Okay, this analogy doesn’t hold water for very long, but you will get the gest I hope.
Back to the subject of comparison. We need to compare the machines on their merits, and all of them at that (or to keep it comprehensible: as much of these merit as is deemed relevant for the discussion).
So to compare the 60 Dell machines with a single AS400 and saying this cluster outperforms the one AS400 is simply not enough. You have to compare more parameters as some of the reactions also suggest. The GUI (or the lack thereof) should not be part of the equation, or it must be as part of the comparison on the development time for the different platforms, but then the comparison to a cluster of 60 machines makes no real sense.
In this case, one has to compare the total cost of ownership including the cost of electricity (the cluster uses electricity as does the A/C, etc.) and the throughput given a certain environment (scientifically, industrial, financial, etc.) and then one can say something useful about the different configurations.
John deCoville |
I remember in the early to mid 80′s, PC’s including the “Clones” being produced by companies like Dell were little pigs of balky sets of “tack-ons.” The System 3x world, although not anywhere as nice the the system i, was quite integrated. New device? No problem, just plug it in and away you go! My early AS/400′s in ’88, 89, 90 were truly integrated.
Fast Forward to now. Microsoft is winning the Sales Game with their expensive IDE integrating Visual Studio together with everything else. Middleware producers for the System-i cannot compete with the money being spent, result: a system that feels balky, everything added in the last ten years feel like afterthoughts: Tack-ons.
Wat do we need: Get IBM to get its act together so appearances (80% of reality for many folks) is in line with the system-i’s wonderful features.
I liked the rebuttal on the 60 Dells versus one System-i.
OK…lets say.. I had to choose between 60 servers to maintain or 1 server to maintain. Well I really do not see why anyone in there right mine would pick the 60 servers. This is totally an overhead cost issue no matter what the brand name of the servers are. Plus it’s at least 60 points of failure…. no brainier….. 60 servers equates to at least 2 technical people to support. No one mentioned SAN attached to that group of 60 servers, that’s another technical person. 1 iSeries server, 1 part time technical person probably on contract when needed. IBM Wins