AmigaOne based WebServers

[DaveP]

Depends on how serious you are!

There is absolutely no reason why you can have a LAMP, LATP or any other configuration of server on an AmigaONE running Linux.

Given that you can load it up with 2Gb of RAM and have multiple 100Mb ethernet cards or even GigE the only concern you would have would be would the processor be able to keep up? The answer is yes.

I have a webserver that takes several million pageviews per day running on a 300Mhz PPC RS/6000 ( 604 chip ) with only 512Mb RAM and it holds up lovely it uses four 100Mb ethernet cards to connect into the network and runs a cluster of four apache servers. This sits behind a server that uses dns redirection to round-robin around the ip addresses. AnyNet or something.

The web, unless you have complex server side includes or complex query, is a mostly data oriented operation ( which is why mainframes and mid-range servers are designed around data operation and data speed rather than raw processor power ) so make sure you have at least two drives, both the fastest you can buy and set up one to mirror the other.

Then run a minimal linux installation either behind someone elses firewall or running its own firewall and you will be able to manage very well.

The AmigaONE has been through extensive testing so don't worry too much on that point.

I think "sixgirls" does machine hosting  so there is a good possibility.

Don't worry overmuch about the BSD vs Linux bigotry uh sorry "debate" because both do very well thank you very much.

Simplest way to test it is to have at least two other boxes connected up on the same network to it running many concurrent curl downloads to see what percentage of cycles and memory goes to apache, mysql etc. This helps you identify bottlenecks in your code - you may need to lighten your queries to stop bottlenecking on mySQL or you might want to decrease the amount of processing code you have in your PHP ( or EJBs or whatever you use ).

AmigaONE boards are fine for rack-mounting too.

Ignore the FUD, get one and get playing!  :-D

[DaveP]

"I'd never buy an AmigaOne or a Pegasos to use as a webserver ... their not tested...this 'april' fix implimented in hardware apperantly is only done to the pegasos... I dunno about the AmigaOne... that could cause issues if it hasnt been corrected( is it?) ... "

Do you know what the April chip fixes? Can you explain precisely how without the April chip somehow magically it is going to have an effect on data transfer between an ethernet card and the rest of the system? If not, dont bother.

"I wouldnt call it FUD to say a non-tested system with a solderd on g3 600mhz..and questionable chipset wouldnt be the best choice for a reliable web server."

Oh come on, what you think the solder is going to come loose? Questionable chipset? Questionable how? Sheesh. Non tested by whom??

Finally I am running in 32 bit mode on that lil old 604 and it is slower than the g3@600Mhz according to benchmark tests already done.

If you aren't going to be precise then it frankly smacks of FUD. Like alluding to the fact that the Amstrad 512 didnt ship with a fan so the hard drive might melt.

Ive still got visions now of solder coming loose or melting under the strain.
 :-D

[DaveP]

"compareing an RS/6000 to an AmigaONE is like comparing a ferarri to a yugo..."

I think that is overselling the RS/6000 significantly. I would say like comparing a Maserati Biturbo ( expensive, huge depreciation and fun for all the family performance ) with an Alfa Romeo Spider 2.0.

[DaveP]

April is fixing an IDE issue last I read... therefore it could effect disk access... that most certinly effect a server of any kind..

And yet the demonstration of it happening on an A1 has not been shown at least in public by anything other than a zealot and in fact not even by a zealot. ;-)

Would you call Tom a zealout? Or MAIs testing engineers zealouts? Hmmm?

No its this whole A1 is an unknown factor, untested, no April chip fud bandwagon that is going along. It would be nice to see some details of what exactly it fixes and some EVIDENCE that it impacts the A1 board for once but that is OT.

I compared the A1 with IBM 43P hardware for a very good reason. Firstly the A1@600Mhz benchmarks at 30% increase in speed worst case over the 43P  ( Yes it would be nice to see some public tests results I'll see what I can arrange ) and secondly it is demonstration that PowerPC chips are lower rates than the current 2GHz Intel based systems can cope with a reasonable load.

What you get in the A1 package, XE or SE is superior to what I have in my 43p ( rated 332 Mhz not 300Mhz just looked on the front cover ).

Plus when you look in the 43p there are surface mounted ( soldered! ) chips everywhere and that has done quite a few years of service now. Must have been 5 years since i bought it.

If you really piss your pants about surface mount chips go and get an XE.

But Ive seen chips work their way out of sockets before now! Quick front page news! Better not buy socketed chips anyone ;-)

[DaveP]

@mips_proc

"I want evidence 'either way' ... I want to see a good review... I "

Yep. Me too, and Id rather see well documented /objective/ evidence either way before I see any more "AmigaONE doesn't have the April chip " implying that the AmigaONE is somehow fatally flawed.

Sure there may be a flaw, but how it impacts system stability or performance or manifests itself whilst running an operating system is totally undocumented - at least in the public and private domains I have access to. I see a lot of hype but not a lot of meat on the bone.

I want to see a technical review of Pegasos and AmigaONE SE and XE variants on a trusted third party site done by trusted engineering test staff but the only problem is are anyone other than zealots going to buy these boards? Probably not. :-(

"but for a webserver speed isnt everything...bieng able to handle a wide load of many users/etc matters more...but "
Definately.

Enjoy your weekend!

[DaveP]

"Well.. It's good to see that there are experienced Hw professionals on making that system too. But I allways thought you at Hyperion were only supposed to do game ports, not to be specialized on designing and testing hardware? "

Ah oh fountain of fud, what fun you are.

To suggest that MAI do not have the expertise in hardware design and testing to come out with TeronCX, TeronPX, Artica* and many other hardware products not only denies their current portfolio of delivery but it also denies the credentials of the staff that work there. As if it was ever in doubt!

Secondly to imply, as you do here, that Hyperion are somehow just a bunch of game porters is further to turn the FUD screw and is doing them a subtle discourtesy but a discourtesy nonetheless.


I wonder, what are your engineering credentials ( assuming of course you have any ) that makes you think that testing hardware cannot be done by people that actually use that hardware for the purpose that was intended by it?  I don't want a link to something you just read on the web I want to know what your experience and credentials are that you think people should really listen to your opinion without substantiation?

Im sure I would rightly get a whole bunch of screaming loonies on here having a go at me if I said "but I thought you at BPlan were only supposed to do flawed accelerator cards, not to be specialised in designing and testing complete motherboards?".

If your motives werent so transparent every time you posted Joanna I would be suspicous of them.

[DaveP]

"The whole idea behind OpenBSD is that security is built right into the system and the code is audited for bugs, so there are still exploits that will work even with a fully patched and hardened system on (otherOS) but not on OpenBSD because of this."

So because OpenBSD is audited and has security built in ( as do all Unixes to one degree or another ) it means that all other systems can be exploited?

Or are you saying that OpenBSD cannot be broken into regardless of patch level ( not credible ) or are you saying that Linux does not undergo security reviews ( wrong ) or are you saying that no other OS has security built in ( wrong ) or are you saying that OS with ESMs are not as secure ( wrong )?

I've lost count of the number of security holes that have had to be patched with Apache that are security holes regardless of host OS ( including BSD ).

Any system is proportionally more vulnerable the more back level its security PATCHES (typo) are.

Nah, I think that OpenBSD, fearful of Linux popularity has decided to build up a wee bit of mythology around itself to help it survive. Encrypting swap aside ( as if you could really get to it ).

Plus, what idiot runs an out of the box non "hard" Linux distro?

From http://www.developer.com/open/article.php/990711

AIX
10 vulnerabilities[6 remote, 3 local, 1 both]
Debian GNU/Linux
13 vulnerabilities[1 remote, 12 local] + 1 Linux kernel vulnerability[1 local]
FreeBSD
24 vulnerabilities[12 remote, 9 local, 3 both]
HP-UX
25 vulnerabilities[12 remote, 12 local, 1 both]
Mandrake Linux
17 vulnerabilities[5 remote, 12 local] + 12 Linux kernel vulnerabilities[5 remote, 7 local]
OpenBSD
13 vulnerabilities[7 remote, 5 local, 1 both]
Red Hat Linux
28 vulnerabilities[5 remote, 22 local, 1 unknown] + 12 Linux kernel vulnerabilities[6 remote, 6 local]
Solaris
38 vulnerabilities[14 remote, 22 local, 2 both]

[DaveP]

>that makes you think that testing hardware cannot
>be done by people that actually use that hardware
>for the purpose that was intended by it?

Hyperion already was wrong the last time they said
that there are no bugs in Articia. So why should
somebody trust them this time?

Why is this relevant to the question at hand? Do you think that testing cannot be performed by actually using the hardware for the purpose intended?

I also see this said a lot of times "Hyperion already was wrong the last time they said that there are no bugs in Articia. "

Can you or someone else point to a link where Hyperion actually said precisely that?

[DaveP]

@JoannaK

"And seeing your posts"
Yes I post. Congratulations.

" and personal attacks,"
Oh, now what do you think is a personal attack?

" it's quite obvious you are even less neural than I am.. So blame me here.. "

Neural? Care to sit an IQ test?

I notice you continue to dodge the question, what are your credentials?

"Well.. At least I'm honest with my views"
Im honest with my views too.

" and don't attack others anomously on ann.lu."
I don't either. But then you claim that and there is no way of proving it. I guess we will have to say innocent until proven guilty.

" And that's a lot more than can be said among few aroud here. "
Quite. But can you for once instead of just throwing FUD cogs into threads actually substantiate your points with evidence? That is all I ask of you - alternatively you can answer the question and tell us why we should all give your opinions such credit that they don't require substantiation.

[DaveP]

zacman wrote:
>Do you think that testing cannot be performed by
>actually using the hardware for the purpose
>intended?

In this case yes. Because those issues happen
randomly under "normal" use but can be
reproduced under very specific conditions.

If they happen randomly under "normal" use then they are detected under normal use. It is the job of a tester to find the specific conditions that can reproduce it 10 times out of 10 and raise the defect.

The problem is initial detection.

So yes, testing can and should be performed using the board under "normal" conditions. Normal being defined as a set of scenarios ( or use cases ) for the given target market.

If an error *only* occurred under specific conditions in a scenario where the board would not normally get used it is still a bug but is it one worth fixing now or patching later if anyone finds it? Thats the job of the delivery team.

>Can you or someone else point to a link where
>Hyperion actually said precisely that?

Sure.

"Posted by Ben Hermans/Hyperion- (134.58.222.78)
on 30-Sep-2002 19:20

For the record, there is no Articia bug."

He then continues with some FUD against bplan
which I won't quote here.

Thank you for actually doing what was asked, although it would be nice to have the link to that ;-). The only thing I have found that matches the phrase is a repost on a Polish website which I can't translate! ( www.amiga.pl/comments.jsp?nid=1595  )

The other thing I would like to see is some documentation of exactly what the bug is that April fixes and documented proof that it is actually in Articia and is somehow critical to operation?

http://www.pegasos-uk.com/english/support_pegafaq.html

..is the only vaguely useful document on the matter but even that does not contain a lot of meat.

I think it would be ludicrous to claim that any product
is bug free ( but then Ben Hermans is not technical ).

I realise we are straying from the main point here by a long way but it is about time a bit of put up or shut up happened.

Finally there are more than one way to patch around a bug, more than one way to find a bug and more than one way to fix a bug. Remember even the push
on chips we used to have to put on A1200 mobos to allow us to use Zorro II busboards? Or the pin to pin wire connections for the earlier versions?

To claim that somehow the AmigaONE is flawed because it does not have "April" is disingenuous.

[DaveP]

Just for this once I'll humor you. But would you mind telling me why I'm the only one who has to show credentials here?  

Because, as I have already said, you throw assertions into threads without substantiation. Is it so mad as to require "humouring" to ask for some evidence?

I've been around Amiga since early days. Was one of OS1.4 -Beta testers

Now that is very interesting, anywhere you can back this up? I thought I was the only one alive who actually had disk images of 1.4. ;-)

..
Since Commodores demise

Whats this, a paid beta tester for 1.4? Did you actually work FOR C=?

moved to hw engineering and other patforms. Lately being designing and testing embedded PPC devices at Internet Technologies.. Some publicly displayable reference boards are on company website at www.iti.fi

Cool, Ill go check it out. Now how exactly do you TEST those PPC devices?

And you? Yet another Physiotheraupist reborn as a Os-analysists or self learned lawyer ?

Well at least you don't resort to personal insults eh?

No I'm a "Software Engineer" who has worked on such projects as OS/2 2.1, OS/400, DCE, the JTS implementation shipped with WebSphere, JVMs for Itanium ( pre-production working a week at a time at Intel sharing sparse time with other teams ) and PPC etc. Software, software, software.  

You can work out the chronology by the release dates I would guess. Recently I have been working on several things at once, none of which I can talk about until release mores the pity. I happen now to have a side business that I own through invested capital selling solutions as well as work under contract.

All of these you will note are pretty large scale development and testing operations. I happen to know from bitter experience that you only have problems at product integration to the extent of having to go back to the drawing board if

a) you are being managed by a bunch of retards at "architecture" stage.

b) you are a retard or engineering collectively has been retarded and did not follow the architectural plans for the system.

So this is why when I picked up your assertion ( fud cog ) that the worst bit of producing OS4.0 was likely to come at the integration level as being unlikely.

This is also why ( JVM use of Itanium being an example ) I happen to know that the second best method of picking out bugs with hardware is to run software on it.

This is why I have questioned your assertions each time and asked for substantiation because it is so much at odds with my own experience.

[DaveP]

"a) you are being managed by a bunch of retards at "architecture" stage"

which may or may not be the case ;-) But what I meant to add here was that given the syntax and semantics ( the implicit contract ) of the OS exposed to application programmers are well known and well documented ( if nothing but through simulation by running OS3.9 ) that the only big concern would be whether or not Petunia works properly in this context.

[DaveP]

@JoannaK

www.iti.fi looks like it needs a bit of a web page update! Any customer testimonials for download? I'd like to understand your target market a bit more.

You seem to be a collection of different people probably just doing stuff that you enjoy but that is reading too much into the site ( embedded technoilogy, internet consulting ( education ), web stuff and linux support ) .

Interesting but it doesn't exactly say to me large scale pro development operation.

Don't want to sound offensive here because it is actually quite interesting - but in the context of what is being discussed and claimed by you on these forums in the assertions that I have underlined with a response recently and further back in the past - you will forgive me if I continue to ask you to back up assertions rather than take your opinion as gospel.

Good luck with it, Id like to get onto doing something that I enjoy rather than being a wage slave.

[DaveP]

@T-bone

An AmigaOne would be a crummy choice for a webserver, you can get a doorstop pentium practically free for this purpose as they litter the earth, why spend all that on an exotic piece of Amiga hardware just to run a mundane webserver that any old $0.02 discarded PC can do?

Well if you wanted to support the platform against the odds I guess ;-) I have to agree with you. The only reason my 43p is sitting there on the T1  is because the OS it runs is probably my favourite and secondly because I have a lot of SCSI devices attached to it.

If I were buying from scratch for it I would probably not buy an old mini or micro ATX PC because of the power problems trying to run so many networking cards and devices micro-ATX presents - it would probably be a second hand netfinity or something for a couple of hundred quid.

I would not personally consider the A1 SE or XE boards as a logical step up from where i am and not a logical first buy for a web server.

So... agreed.

( actually therein is a good reason to consider an ATX based board over a Micro-ATX based board - I wouldn't even be able to get all my network cards into a microATX ).

[DaveP]

I agree with most of what you say but this:

" If you plan to use a dynamic site with a database behind it, then forget about it. They're just not up to the job."

Au contraire, they are quite up to the job. The issue then is less data transfer ( esp with databases that don't always go to persistent storage to retrieve results  :-D ) and more CPU power.

The issue is less maintenance ( backup sites can always be brought online and switched over to take the other offline ) or uninterruptable power supplies. The issue is  cost vs value.

If the value you want is low volumes ( >1 million complex dynamic page hits per day ) and you aren't doing business critical applications ( such as a banking site or shopping site where a missed transaction equates to money ) then an ATX based board is quite adequate to the task.

Data mirroring can happen without costly RAID by using an ethernet connection and an exit to drive a mirror transaction against a remote version of the database. Maybe one, two hours of programming!

These days a hell of a lot of overkill is ordered by greedy web hosting techies that don't think hard enough about fitness for purpose and try to scare money out of procurement managers ;-)

Frankly I would rather pick up a second hand netfinity for 500 quid but...

I don't know what Wayne uses but it is a good example of a popular site which could easily be hosted on an ATX.

Something like www.tesco.com should be hosted on special hardware with data and processor redundancy built into the design. But then if you are going to do that use something that has ACID properties to run it on ( like WebSphere, MQSeries and DB2 ).