Compile Farms for Commercial Software?

unix-coder asks: "How can a small software company get to 'rent' accounts for short times on a wide range of machines running different commercial Unixes with different CPUs? SourceForge's compiler farms are great for open source projects (and besides, open source projects will get ported/tested/fixed on all the platforms that matter to the people that use them). But what about commercial projects where you want to port/build/test for a wide range of architectures and OSes (AIX, HP-UX, IRIX, ...) but don't want your several server rooms full of weird hardware of your own?"

Comments

If SourceForge wants to make money...

[Anonymous Coward]

They should sell this service for closed source types. They've obviously got the know how to do it.

Yes, making money! It can be done.

Re:If SourceForge wants to make money...

[dimator]

making.....? money..... ?? VA!???

are you visiting from some parallel universe?

Re:If SourceForge wants to make money...

[chefren]

1. Make money
2. ???
3. PROFIT!!!

dumbass

[hfastedge]

SourceForge's compiler farms are great for open source projects (and besides, open source projects will get ported/tested/fixed on all the platforms that matter to the people that use them).

This person can barely communicate in written sentances. He doesnt really say what about SF's compile farm is lacking.

Re:dumbass

[daviddennis]

The fact that the license doesn't let you use it for commercial software.

That seemed clear enough to me.

I'd probably just tell him to buy the novel hardware. I bought a very nice Indigo2 for $400-odd a few months back, and it's a cooler experience to have the real equipment someplace.

But that's only me.

D

Re:dumbass

[CFN]

"I'd probably just tell him to buy the novel hardware. I bought a very nice Indigo2 for $400-odd a few months back, and it's a cooler experience to have the real equipment someplace"

How is this a solution to the problem he wants solved?

He wants to test his software on many different platforms (different OSs, different CPS, multi-cpu machines), and more importantly, he wants to test it on the CURRENT platforms his (potential) customers use.

How does buying one, 10 year old machine, in anyway approximate what he needs done?
You know, maybe he can just use the Commodore 64 and Apple II I have lying in my basement.

Re:dumbass

[Anonymous Coward]

in written sentances

Quite.

Re:dumbass

[murat]

This person can barely communicate in written sentances. He doesnt really say what about SF's compile farm is lacking.

At least, his sentences are not misconstrued. Besides, he does not have tpyos in his sentAnces. Dumbass.

Re:dumbass

[elbles]

Regardless of the fact that an Indigo might be quite a few years old, most, if not all of them, are perfectly capable of running the latest versions of IRIX without problems, or being too slow to use. The point being a lot of UNIX hardware from 7-10 years ago is still similar enough to the technology of today that it can be used for a compile farm. You could get a Sun UltraSPARC workstation for less than $500 easy, same with an Alpha, etc . . . it's a pretty good solution.

Re:dumbass

[Anonymous Coward]

This person can barely communicate in written sentances. He doesnt really say what about SF's compile farm is lacking.

Yes I did! I said it's for open source projects (cf the title of this post). It also doesn't have particularly exotic platforms available.

That's "sentences" by the way. Dipshit.

Re:dumbass

[dohcvtec]

How does buying one, 10 year old machine, in anyway approximate what he needs done?
Well, you don't buy a 10 year old machine. Although my roughly 10 year old SPARCstation 10 that I bought for $15 does run the latest version of Solaris easily. You can snatch 2-3 year old, viable machines off of ebay for less than $1000. This is the beauty of the UNIX/RISC workstation/server market. Used machines depreciate even faster than plain-jane consumer PCs, because most companies don't want to buy used machines, and most consumers wouldn't know what to do with them. Who's left? Developers who just need machines for testing, and crazies like myself who dream of running an IBM zSeries in my basement.

Re:dumbass

[Anonymous Coward]

What the fuck are you talking about? I'll bet he can spell "sentence", at least.

Re:dumbass

[daviddennis]

As it happens, Irix 6.5 is the current version of Irix. There's not much you can do on a new Irix system that you can't do on an Indigo2, although of course the Indigo2 will be significantly slower.

And I did say "But that's just me", an indication that this was not a precise fit for his problem, but something worth considering. And I stand by my answer.

D

Compaq (HP) Test Drive

[mnordstr]

http://www.testdrive.compaq.com/ might do the trick. Never used it though...

Re:Compaq (HP) Test Drive

[Malcolm MacArthur]

IBM have a similar (but more formalised) system for developers.See details here. However, I have no idea how much it costs - the costs might be prohibitive.

Call vendors...ask for demo units.

[j-turkey]

You have a couple of options -- you could just call the vendors and tell them that you're a VC funded software startup and you want to evaluate thair systems. They'll send you a demo unit, and you'll be able to build on their systems (for a short time). Sure, its dishonest, but I leave the ethics up to you. You could also tell them the truth -- maybe some of these struggling Unix vendors would give you a free workstation (or a deep discount on one) if you're selling 3rd party software that will enhance their platform (though its a longshot).

An alternative would be to start looking for ISP's who will sell you shell accounts on a month-to-month basis. If any of your employees are alumni of colleges and universities, some may still have accounts on legacy Unix systems...these may help too.

The downside of using shell accounts on other folks' systems for building commercial software is that you run the chance of building on inconsistent environments. This can lead to a host of problems that you will not want to deal with.

The final alternative is to go out and buy the systems (I recommend this one). It may be cheaper than you think...although the affordable systems may not be so cutting edge, I'd recommend eBay. There are tons of old SGI, HP-UX, Digital Unix, Sun, and AIX stations out there for very cheap...probably as cheap as a year's worth of shell hosting (and you'll own the machines). You'll also have the security of knowing that you're building on a "clean" environment.

I hope this helps!

-Turkey

Re:Call vendors...ask for demo units.

[aridhol]

Another advantage of using old hardware is that you will test your code on a low-spec machine. You will know that it is usable on a low-spec machine, and will probably fight software bloat to make it run on your own test machines.

If marketing makes you add something, you can always make them run it on your test boxes, too. Show them why it's a bad idea.

Re:Call vendors...ask for demo units.

[billburroughs]

I definitely agree with this idea of getting demo gear. If you have a decent alliances guy this shouldn't be too tough to swing. Thing is, renting other folk's equipment can be problematic. You cannot just get rid of the equipment the moment the software compiles. There are too many support issues that require you to keep those exotic pieces of hardware around, or something equivalent you can get your hands on quickly, like in a few days.

Alternatively, try to get some used lower end gear that will still work and still be usable for a few more versions of the OS. Your alliances guy could probably find some other small non-competing software companies you could partner up with to share some of these resources.

Sourceforge

[ikeleib]

I'm sure that VA, who is always looking for some money, would be happy to strike up some sort of arrangement. They obviously have the resources already in place.

i386

[Captain Large Face]

I think it was on Slashdot that I read about a conversation between two users, one new to Unix operating systems and the other an old hand.

The new user was asking the older user why a lot of the packages had i386 in the title, warranting a reply from the experienced user of something along the lines of "because that is what they were compiled on...", leaving the new user to ponder about large farms of 386-processor machines.

Perhaps a compilation farm such as this could generate an even greater level of confusion!

Use machines at vendor's offices

[gadfium]

This is what I did about 10 years ago, when I was selling a program on many different versions of Unix.

I bought the most common 386 unix versions (SCO Xenix, SCO Unix, 386/ix), and had one risc machine (an AViiON, since DG was my main customer base). If a user wanted an RS6000 version, I just rang my local branch of IBM, arranged a time to pop in, and compiled my latest code on a machine there. It cost IBM nothing, and may have sold a few machines for them.

Why bother?

[Telastyn]

I mean depending on the software, just don't support it...

Say "Hey, we made it nice and portable as we can, and it *should* work on AIX,HP-UX, etc... but has not be verified." and tell your salespeople that if customers want to run on AIX, HP-UX, etc that the QA/dev team will need some time (and a box probably) to verify everything will work smoothly for them.

Re:Why bother?

[Malcolm MacArthur]

That's a nice idea, but remember we're talking BUSINESS here. If you can't deliver the goods, the customer goes elsewhere. They won't hang around waiting for you to port it - unless the application in question has a very small niche market.

Certainly, if I was selling commercial software, I'd try to avoid giving an answer like that. Best thing to do in that situation is to say "I'll have an answer for you by tommorow/next week/in half an hour" (depending on how desperate you are for the sale and how competitive your market is :). Then you go and verify your software runs on their platform, probably through an arrangement with the computer manufacturer. As software helps sell systems, most manufacturers should be willing to help.

Of course, it's all dependant on your relationship with your customers and the nature of the software you are selling. The more "mission-critical" an application is, the more the customer will want a firm answer. The more competitors you have in your market, the quicker your customers or potential customers will want an answer - or they go elsewhere. Niche markets - well, as long as they're not desparate for it, they will wait.

Apple does this for developers

[Anonymous Coward]

by giving out free time on a collection of just about every Mac it could be used on. Its great for testing compatability.

IN SOVIET RUSSIA

[Anonymous Coward]

Commercial software compiles YOU!

eBay SGI machines

[green pizza]

Search for "sgi" and you'll find even more silicon graphics machines. Most auctions don't have "workstation" in the title.

Re:eBay SGI machines

[j-turkey]

Search for "sgi" and you'll find even more silicon graphics machines. Most auctions don't have "workstation" in the title.

There were far too many listed -- I wanted to narrow down the search by removing the newer, more expensive machines (it sounds like this guy doesn't have much of a budget for the newer SGI's anyway).

-Turkey

Compaq Test Drive is non-confidential

[Myrcurial]

And therefore unsuited to closed source porting.

HP offers a closed source developer platform similar to IBM's Solution Partnership Centres at the Developer and Solution Partner Program.

I agree that SourceForge could/should provide this service, but the big guns offer it for cheap/free to have a larger family of ISVs servicing their platform. Maybe the question should be re-phrased.

Buy a Used Zoo

[4of12]

If you really want to test compile on all those exotic UNIX platforms, that's great.

You get extra credit for thinking about cross platform portability, which is closely related to thinking ahead. You'll reap long term rewards in reduced maintenance costs from having less fragile code.

So why not go ahead and buy some of those machines used?

Through `em all together in a room with reliable power and an Ethernet switch and you'll be ready to go. Given that you can get free *NIX flavors on x86 that are good performers, the old boxes are typically quite cheap.