Amazon EC2

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
24 messages Options
12
Reply | Threaded
Open this post in threaded view
|

Re: Amazon EC2

Avi  Bryant

On Sep 7, 2006, at 11:06 AM, Jason Johnson wrote:

>
> Right, but what I am saying is, the users (e.g. AOL users) are  
> using some ISP for internet access.  When they try to hit your  
> page, their PC sends a DNS request to whoever it's configured for,  
> which will be the ISP (e.g. AOL) servers.  The ISP servers will ask  
> the root servers, find you and give the answer, but they (or they  
> used to) ignore the TTL field.  They just run a modified version of  
> BIND or whatever with the cache time hard coded to 2 days.  So for  
> the next 2 days all users that use the effected ISP server will hit  
> that cache.  That wouldn't mean all of AOL for example, but some  
> percentage.
>
> Now I don't know how systems that us Dynamic DNS are getting around  
> this, but I guess they are so it probably wont be a problem.  All I  
> know is I changed over my domain some months back and I couldn't  
> get to my site for 2 days by name because of it.
>
> If this reminder is irrelevant for whatever reason, I apologize.  I  
> was trained for nearly a decade to point such things out. :)

No, thanks for explaining, I understand now.  That seems like  
obnoxious behavior on the part of AOL, but it would definitely pose a  
problem for the strategy I proposed.  Does anyone have any more data  
on whether this still happens and how DynDNS etc get around it (if  
they do)?

Avi
_______________________________________________
Seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
Reply | Threaded
Open this post in threaded view
|

Re: Amazon EC2

Klaus D. Witzel
On Fri, 08 Sep 2006 18:42:09 +0200, Avi Bryant wrote:

> On Sep 7, 2006, at 11:06 AM, Jason Johnson wrote:
>>
>> Now I don't know how systems that us Dynamic DNS are getting around  
>> this, but I guess they are so it probably wont be a problem.  All I  
>> know is I changed over my domain some months back and I couldn't get to  
>> my site for 2 days by name because of it.
>>
>> If this reminder is irrelevant for whatever reason, I apologize.  I was  
>> trained for nearly a decade to point such things out. :)
>
> No, thanks for explaining, I understand now.  That seems like obnoxious  
> behavior on the part of AOL, but it would definitely pose a problem for  
> the strategy I proposed.  Does anyone have any more data on whether this  
> still happens and how DynDNS etc get around it (if they do)?

IIRC even registrar's behavior can cause 2-3 days ttl, see for example

- http://www.dyndns.com/support/kb/archives/glue_records.html

And from own experience, when troubleshooting my client's IP / DNS / TTL /  
MTU problems, I know with confidence that there are IProviders who have  
"pass TTL of answers to customer's DNS query according to our greediest  
business plan" as default business rule for their non-static IP addressed  
customers ...

/Klaus

> Avi


_______________________________________________
Seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
Reply | Threaded
Open this post in threaded view
|

Re: Amazon EC2

Jason Johnson-3
In reply to this post by Avi Bryant
Avi Bryant wrote:

>
> On Sep 7, 2006, at 11:06 AM, Jason Johnson wrote:
>>
>> Right, but what I am saying is, the users (e.g. AOL users) are using
>> some ISP for internet access.  When they try to hit your page, their
>> PC sends a DNS request to whoever it's configured for, which will be
>> the ISP (e.g. AOL) servers.  The ISP servers will ask the root
>> servers, find you and give the answer, but they (or they used to)
>> ignore the TTL field.  They just run a modified version of BIND or
>> whatever with the cache time hard coded to 2 days.  So for the next 2
>> days all users that use the effected ISP server will hit that cache.  
>> That wouldn't mean all of AOL for example, but some percentage.
>>
>> Now I don't know how systems that us Dynamic DNS are getting around
>> this, but I guess they are so it probably wont be a problem.  All I
>> know is I changed over my domain some months back and I couldn't get
>> to my site for 2 days by name because of it.
>>
>> If this reminder is irrelevant for whatever reason, I apologize.  I
>> was trained for nearly a decade to point such things out. :)
>
> No, thanks for explaining, I understand now.  That seems like
> obnoxious behavior on the part of AOL, but it would definitely pose a
> problem for the strategy I proposed.  Does anyone have any more data
> on whether this still happens and how DynDNS etc get around it (if
> they do)?
>
> Avi
> _______________________________________________
> Seaside mailing list
> [hidden email]
> http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
>
>
> --No virus found in this incoming message.
> Checked by AVG Free Edition.
> Version: 7.1.405 / Virus Database: 268.12.2/441 - Release Date: 9/7/2006
>
>

Yea, it does seem obnoxious.  But nearly all (if not all) ISP's in the
US used to do this (and I would imagine still do).  Earlier bandwidth
was small and expensive and the ISP's wanted to eliminate any traffic
they could.  DNS was a low hanging fruit because most of the ISP
customers will wind up going to the same sites.
_______________________________________________
Seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
Reply | Threaded
Open this post in threaded view
|

Re: Re: Amazon EC2

tblanchard
In reply to this post by Chris Muller

On Sep 7, 2006, at 7:23 PM, Chris Muller wrote:

> S3, Turk and now EC2.  What a fascinating set of distributed computing
> services!  Do we have a new species of computer here (low-cost virtual
> computer)?  What a hopeful enablement for small players!

That is the idea.

> Wouldn't it be neat if, like Turk, anyone could contribute their own
> computer to sell their CPU cycles for a few pennies.  Does annyone  
> know
> how it works?

Yes, but if I told you, I'd have to kill you. :-)

> How else can Amazon quickly procure enough hardware to
> meet demand fluctuations, while still keeping prices fixed and so
> reasonable, I dunno?

By relentlessly yelling at their developers to improve the efficiency  
of their code.  :-)

Seriously, what they are doing is renting out spare capacity in their  
data centers and providing a public interface into their provisioning  
infrastructure.  Its the same stuff they use to run their own website  
and manage their own fleet, but demand fluctuates seasonally (you  
know - Christmas) and they end up with spare capacity.  So they've  
decided to rent it.



_______________________________________________
Seaside mailing list
[hidden email]
http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside
12