![](/i/fill.gif) |
![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
>> If you want to use it just for yourself, sure. If you want it to
>> provide a service to other people... well, no way to know when those
>> service requests will arrive, eh?
>
> Sure. Then it's doing something.
Perhaps you're missing my point. I get charged for running the server
24/7, when it only gets maybe 3 hits per month. That's really expensive, no?
>>> If you want to serve static web pages, use S3, not EC2.
>>
>> S3 only stores data. There's no way of accessing it from the outside
>> world again. (As far as I can tell.)
>
> It serves any file there via HTTP, with a URL like
> http://s3.amazon.com/mybucket/myfile.zip
> or some such.
Interesting. It appears from the documentation that the contents of a
bucket cannot be hierachical...
>> I'd actually like to run custom CGI. But paying 72$/month for a server
>> that gets maybe 4 hits per month seems... excessive.
>
> There is that. What CGI is this?
I have a blog, but I'm not very happy with it. I'd like to write my own
blog engine and run that. (In Haskell, naturally.)
>> Three hours does not concern me. I have an animation running that's
>> likely to take about a week to render.
>
> That's a pretty good use of it.
But would it be cheaper to buy a new Intel Core i7 and run it on that?
(Once I've bought it, I can use it endlessly for nothing.)
>> Looks like you'd need to script it to run POV-Ray, run VirtualDub,
>> somehow copy the data to S3, and then automatically shut down ASAP
>> before you go bankrupt!
>
> Not sure what the virtualdub part involves
Transforming 6,000 frames into a single AVI so I can download it.
> but you can log in remotely, you know.
What, and be charged for 7 hours of computer time when the instance has
finished rendering and is just sitting there waiting for me to tell it
to transcode the video? :-P
> Posting the stuff up to S3 isn't hard either. I can give you a
> script that pushes a directory up to S3 if you want, too.
I guess the fun part is if an instance "fails" during a render.
Apparently it can do that. No refunds...
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Orchid XP v8 wrote:
>>> If you want to use it just for yourself, sure. If you want it to
>>> provide a service to other people... well, no way to know when those
>>> service requests will arrive, eh?
>>
>> Sure. Then it's doing something.
>
> Perhaps you're missing my point. I get charged for running the server
> 24/7, when it only gets maybe 3 hits per month. That's really expensive,
> no?
I understand your point. :-)
> Interesting. It appears from the documentation that the contents of a
> bucket cannot be hierachical...
URLs aren't either. Put a slash in the name of the file you store in the bucket.
It's a bit more complicated, really.
> I have a blog, but I'm not very happy with it. I'd like to write my own
> blog engine and run that. (In Haskell, naturally.)
So you're talking about the form that takes comments?
> But would it be cheaper to buy a new Intel Core i7 and run it on that?
> (Once I've bought it, I can use it endlessly for nothing.)
Certainly renting is more expensive in the long run than buying. But that's
why it's the *elastic* compute cloud, you see.
>> Not sure what the virtualdub part involves
> Transforming 6,000 frames into a single AVI so I can download it.
Ah, something you can script, then?
>> but you can log in remotely, you know.
>
> What, and be charged for 7 hours of computer time when the instance has
> finished rendering and is just sitting there waiting for me to tell it
> to transcode the video? :-P
You realize that 7 hours of compute time translates to less than a dollar,
right? Even if you let it go that long without checking, and didn't bother
to set it up to send you an email when it finished?
>> Posting the stuff up to S3 isn't hard either. I can give you a script
>> that pushes a directory up to S3 if you want, too.
>
> I guess the fun part is if an instance "fails" during a render.
> Apparently it can do that. No refunds...
I've been running one for a company for something like 2 years. The only
time it "failed" is when they told me a couple months in advance that they'd
be cutting over to a new way of doing IP addressing, and please migrate
anything over to a new machine. I.e., I've never had the machine crash.
I think it's far more CYA - "don't run your database on a machine that might
crash and never make backups then blame us" - than any expectation that any
given machine will fail at any particular time.
--
Darren New, San Diego CA, USA (PST)
Why is there a chainsaw in DOOM?
There aren't any trees on Mars.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Darren New wrote:
>> Interesting. It appears from the documentation that the contents of a
>> bucket cannot be hierachical...
>
> URLs aren't either. Put a slash in the name of the file you store in the
> bucket.
>
> It's a bit more complicated, really.
You're allowed to have special characters in object names?
>> I have a blog, but I'm not very happy with it. I'd like to write my
>> own blog engine and run that. (In Haskell, naturally.)
>
> So you're talking about the form that takes comments?
Yes. But also the interface that allows me, the blog owner, to add new
blog entries without logging into the virtual server over SSH and
manually editing half a dozen HTML pages. (And, ideally, convert my
custom markup into valid HTML.)
> Certainly renting is more expensive in the long run than buying. But
> that's why it's the *elastic* compute cloud, you see.
Uh... no, I don't really see.
Unless you mean that ranting 20 quad-core virtual machines for a day is
cheaper than *buying* a renderfarm?
>>> Not sure what the virtualdub part involves
>> Transforming 6,000 frames into a single AVI so I can download it.
>
> Ah, something you can script, then?
Indeed.
>> What, and be charged for 7 hours of computer time when the instance
>> has finished rendering and is just sitting there waiting for me to
>> tell it to transcode the video? :-P
>
> You realize that 7 hours of compute time translates to less than a
> dollar, right? Even if you let it go that long without checking, and
> didn't bother to set it up to send you an email when it finished?
Right. Because sending an email is way easier than scripting VirtualDub.
Oh, wait - no it isn't. :-P
>> I guess the fun part is if an instance "fails" during a render.
>> Apparently it can do that. No refunds...
>
> I've been running one for a company for something like 2 years.
Running what? An EC2 instance?
Perhaps you can answer a small question for me then: It seems simple
enough how a random dude like me can sign up and start using AWS. But
what if you're trying to use it for commercial purposes? How does giving
multiple administrators control over the system work?
> I've never had the machine crash.
>
> I think it's far more CYA - "don't run your database on a machine that
> might crash and never make backups then blame us" - than any expectation
> that any given machine will fail at any particular time.
Mmm, OK.
Of course, you get charged money for performing backups. ;-)
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Orchid XP v8 wrote:
> You're allowed to have special characters in object names?
Define "special". Sure, I think you can have most any unicode character you
want in a name. There might be a handful you can't, but slash isn't one of
them. Indeed, slash is in *most* of the names, because the tools that deal
with heirarchical names assumes that.
> But also the interface that allows me, the blog owner, to add new
> blog entries without logging into the virtual server over SSH and
> manually editing half a dozen HTML pages. (And, ideally, convert my
> custom markup into valid HTML.)
That's what offline processing is for. :-) Make the changes locally, run
your script that turns them into valid HTML, and post any changes up to S3.
Run the CGI script on your own machine at home and do the same. (Doesn't the
URL for posts let you specify something to go as a mailto without human
interaction? It's been so long since I had to worry about not actually
having a server running.....)
>> Certainly renting is more expensive in the long run than buying. But
>> that's why it's the *elastic* compute cloud, you see.
>
> Uh... no, I don't really see.
It's elastic because you can rent more machines when you need to go faster
and fewer when you need to go slower. It's just problematic for you that you
can't rent less than one machine at a time. :-)
> Unless you mean that ranting 20 quad-core virtual machines for a day is
> cheaper than *buying* a renderfarm?
Exactly.
> Right. Because sending an email is way easier than scripting VirtualDub.
> Oh, wait - no it isn't. :-P
Sure it is, if you have things configured properly.
mail -s "your render is done" me### [at] example com </dev/null
Admittedly, getting the mail to come out at your ISP without the firewall
etc getting in the way is tricky sometimes. Spam has made most ISPs really
lock down their servers.
Or create a page you can look at on the web server running on the EC2
instance to see how far along you are. :)
>> I've been running one for a company for something like 2 years.
> Running what? An EC2 instance?
Yes.
> Perhaps you can answer a small question for me then: It seems simple
> enough how a random dude like me can sign up and start using AWS. But
> what if you're trying to use it for commercial purposes? How does giving
> multiple administrators control over the system work?
The same as if its your own machine. You create multiple logins, give each
a different SSH key, and let people log in that way.
When you start up the machine, you give it the public key for root to log in
with. From there on, it's completely your machine. If you get the one
running Windows, I imagine you can get it to join your domain, for example.
> Of course, you get charged money for performing backups. ;-)
No, but you get charged money for holding on to backups. Data from EC2 to S3
is free, last I looked.
--
Darren New, San Diego CA, USA (PST)
Why is there a chainsaw in DOOM?
There aren't any trees on Mars.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
>> You're allowed to have special characters in object names?
>
> Define "special". Sure, I think you can have most any unicode character
> you want in a name. There might be a handful you can't, but slash isn't
> one of them. Indeed, slash is in *most* of the names, because the tools
> that deal with heirarchical names assumes that.
Looking at the protocol-level implementation, I'm guessing you can't
include newline characters in an object key. ;-)
>> But also the interface that allows me, the blog owner, to add new blog
>> entries without logging into the virtual server over SSH and manually
>> editing half a dozen HTML pages. (And, ideally, convert my custom
>> markup into valid HTML.)
>
> That's what offline processing is for. :-) Make the changes locally,
> run your script that turns them into valid HTML, and post any changes up
> to S3. Run the CGI script on your own machine at home and do the same.
> (Doesn't the URL for posts let you specify something to go as a mailto
> without human interaction? It's been so long since I had to worry about
> not actually having a server running.....)
Yeah, you can do it with offline processing. But I found that when I did
it that way, I almost *never* posted any blog entries. It was just too
much effort to add a source file, run the processing tool, run a
seperate tool that zips up the changed files, log in to my host, upload
the zip file, unzip it over the WWWroot folder...
Currently I have a real blog interface. I can update my blog from any
PC, anywhere on the Internet, without any special tools. The only
problem is that it's 100% non-compliant as far as standards go. Oh, and
buggy as hell! (It keeps "eating" my formatting instructions... and the
CSS eats bullets.)
>> Unless you mean that ranting 20 quad-core virtual machines for a day
>> is cheaper than *buying* a renderfarm?
>
> Exactly.
Hmm. I still don't think I can afford that, so it's kinda moot.
>> Right. Because sending an email is way easier than scripting
>> VirtualDub. Oh, wait - no it isn't. :-P
>
> Sure it is, if you have things configured properly.
>
> Admittedly, getting the mail to come out at your ISP without the
> firewall etc getting in the way is tricky sometimes. Spam has made most
> ISPs really lock down their servers.
Er, yeah. This is the problem part. It is almost impossible to get any
SMTP server on earth to accept inbound SMTP messages. I've tried.
>> Perhaps you can answer a small question for me then: It seems simple
>> enough how a random dude like me can sign up and start using AWS. But
>> what if you're trying to use it for commercial purposes? How does
>> giving multiple administrators control over the system work?
>
> The same as if its your own machine. You create multiple logins, give
> each a different SSH key, and let people log in that way.
How about S3 though? From the looks of it, I can set up my own AWS
account, and then I can upload content. But what if you have a team of
devs producing content that you want to upload? Do you really have to
give them all the root password to do that? (This appears to violate the
AWS AUP.)
>> Of course, you get charged money for performing backups. ;-)
>
> No, but you get charged money for holding on to backups. Data from EC2
> to S3 is free, last I looked.
Oh *God*, I've lost track of which transfers are free and which ones get
charged at both friggin' ends. E.g., traffic from SQS to EC2 US is free,
but from SQS to EC2 EU is chargable one-way.
--
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Orchid XP v8 wrote:
> Looking at the protocol-level implementation, I'm guessing you can't
> include newline characters in an object key. ;-)
You might be able to if you encode them appropriately, as %0A for example.
> Yeah, you can do it with offline processing. But I found that when I did
> it that way, I almost *never* posted any blog entries. It was just too
> much effort to add a source file, run the processing tool, run a
> seperate tool that zips up the changed files, log in to my host, upload
> the zip file, unzip it over the WWWroot folder...
Insufficient automation. It should be one click.
> Currently I have a real blog interface. I can update my blog from any
> PC, anywhere on the Internet, without any special tools. The only
> problem is that it's 100% non-compliant as far as standards go. Oh, and
> buggy as hell! (It keeps "eating" my formatting instructions... and the
> CSS eats bullets.)
... and that it requires you to run a real server. ;-)
> Er, yeah. This is the problem part. It is almost impossible to get any
> SMTP server on earth to accept inbound SMTP messages. I've tried.
Um, no, it's not *that* hard, assuming you're subscribed to the ISP.
> How about S3 though? From the looks of it, I can set up my own AWS
> account, and then I can upload content. But what if you have a team of
> devs producing content that you want to upload? Do you really have to
> give them all the root password to do that?
You can give them write permission to the bucket on a user-by-user basis.
They'll still need some sort of AWS key, tho, yes.
> Oh *God*, I've lost track of which transfers are free and which ones get
> charged at both friggin' ends. E.g., traffic from SQS to EC2 US is free,
> but from SQS to EC2 EU is chargable one-way.
Probably because SQS is hosted in the USA. :-)
--
Darren New, San Diego CA, USA (PST)
Why is there a chainsaw in DOOM?
There aren't any trees on Mars.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
>> Looking at the protocol-level implementation, I'm guessing you can't
>> include newline characters in an object key. ;-)
>
> You might be able to if you encode them appropriately, as %0A for example.
Ah. So the keys are URL-encoded?
>> Yeah, you can do it with offline processing. But I found that when I
>> did it that way, I almost *never* posted any blog entries. It was just
>> too much effort to add a source file, run the processing tool, run a
>> seperate tool that zips up the changed files, log in to my host,
>> upload the zip file, unzip it over the WWWroot folder...
>
> Insufficient automation. It should be one click.
How do you log into a website and click some JavaScript-powered buttons
from a script?
Besides, even then, I can only update the blog from the one PC that has
the source code and all the automation scripts installed.
>> Currently I have a real blog interface. I can update my blog from any
>> PC, anywhere on the Internet, without any special tools. The only
>> problem is that it's 100% non-compliant as far as standards go. Oh,
>> and buggy as hell! (It keeps "eating" my formatting instructions...
>> and the CSS eats bullets.)
>
> ... and that it requires you to run a real server. ;-)
Well, no, I got it for free as part of my web hosting package. But if I
wanted to change to a different blog engine (say, one that doesn't SUCK
as much), I'd have to pay £15/month for a virtual server instead. (And
configure Apache by hand - which is pretty hard.)
>> Er, yeah. This is the problem part. It is almost impossible to get any
>> SMTP server on earth to accept inbound SMTP messages. I've tried.
>
> Um, no, it's not *that* hard, assuming you're subscribed to the ISP.
Well, I've never got it to work, let's put it that way.
>> How about S3 though? From the looks of it, I can set up my own AWS
>> account, and then I can upload content. But what if you have a team of
>> devs producing content that you want to upload? Do you really have to
>> give them all the root password to do that?
>
> You can give them write permission to the bucket on a user-by-user
> basis. They'll still need some sort of AWS key, tho, yes.
Right. So there *is* a capability to set up multiple user accounts in
AWS then? (Which is what I was trying to get at.)
>> Oh *God*, I've lost track of which transfers are free and which ones
>> get charged at both friggin' ends. E.g., traffic from SQS to EC2 US is
>> free, but from SQS to EC2 EU is chargable one-way.
>
> Probably because SQS is hosted in the USA. :-)
Well yeah, but you see my point. Figuring out what is and isn't free is
a complex minefield. The pricing structure is so complicated...
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Invisible wrote:
> Ah. So the keys are URL-encoded?
The keys *are* URLs. That's why they think they have a REST interface.
>> Insufficient automation. It should be one click.
>
> How do you log into a website and click some JavaScript-powered buttons
> from a script?
I meant that your offline solution was insufficiently automated, as you
described it just above where I said "insufficient automation."
As for your second question, remember that Javascript only sends stuff over
IP in the first place. You automate it the same way the browser does, if
it's worthwhile.
>>> Er, yeah. This is the problem part. It is almost impossible to get
>>> any SMTP server on earth to accept inbound SMTP messages. I've tried.
>> Um, no, it's not *that* hard, assuming you're subscribed to the ISP.
> Well, I've never got it to work, let's put it that way.
I wish *I* was that good.
> Right. So there *is* a capability to set up multiple user accounts in
> AWS then? (Which is what I was trying to get at.)
I'm not sure what you mean. Firstly, "AWS" isn't all one service. Different
services have essentially unrelated security setups. Secondly, I don't think
you can set up "multiple user accounts" other than setting up a bunch of
individual accounts. You *can* give specific people permissions to do
specific things with specific files or buckets. I.e., I think you can set up
a bucket such that only these specific three people can create files in it.
I'm not sure exactly what the permutations are, tho.
> Well yeah, but you see my point. Figuring out what is and isn't free is
> a complex minefield. The pricing structure is so complicated...
Most pay-as-you-go things are. If you're just interested in S3 and ECC, look
at just those prices.
--
Darren New, San Diego CA, USA (PST)
Why is there a chainsaw in DOOM?
There aren't any trees on Mars.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Orchid XP v8 wrote:
> How about S3 though? From the looks of it, I can set up my own AWS
> account, and then I can upload content. But what if you have a team of
> devs producing content that you want to upload? Do you really have to
> give them all the root password to do that?
Of course not. If you're root, you can create new accounts.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
Nicolas Alvarez wrote:
> Orchid XP v8 wrote:
>> How about S3 though? From the looks of it, I can set up my own AWS
>> account, and then I can upload content. But what if you have a team of
>> devs producing content that you want to upload? Do you really have to
>> give them all the root password to do that?
>
> Of course not. If you're root, you can create new accounts.
We're talking about "virtual" root users.
I think each user or group of users would have to have their own amazon
account, and you would have to set the ACLs on the buckets to make it work.
Yeah, you can do that. Give each person a different Amazon login, and then
set up the permissions on buckets to allow uploads/downloads.
--
Darren New, San Diego CA, USA (PST)
Why is there a chainsaw in DOOM?
There aren't any trees on Mars.
Post a reply to this message
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |
|
![](/i/fill.gif) |
| ![](/i/fill.gif) |
|
![](/i/fill.gif) |