On Nov 6, 7:30 am, bFlood <[EMAIL PROTECTED]> wrote:
> andrew - that is quick but I don't see it as quick as GAE spawning a
> new request handler (~1 second) on a machine somewhere in their
> cluster. I guess it's just a different model, autoscaling at the
> request level vs. the machine level.
I'm sure Google's network scales both the 'machines' (plugging in more
hardware for customers) as well as 'requests' (shifting processing
between machines that are already started).
EC2 images/software can already do this also...
>
> EBS - I recognize that NFS (or something similar) would work but I
> could not find anyone using it when they needed to place database
> files on EBS (for performance reasons). please let me know if you find
> anything that works. that being said, how would I host a large DB on
> EC2 where I would likely need more then a single instance to handle
> spikes in traffic? from people i've talked to, you can run the DB
> locally on the instance but it's gone when it dies and the bigger the
> DB, the longer the startup which affected how they auto-scaled
Just have all your instances point to the same Amazon-managed-DB
instance, which scales for you. (The same you are doing on GAE) Or I
might be missing something.
>
> security - greg's comments above should not be overlooked, the
> flexibility of EC2 comes with the headaches as well
As a tradeoff for the GAE limitations, and there are many. Just take
one limitation for example: Python
In particular, I *love* Python, probably the most efficient language
to work in. But I'll attest than none of my clients - all fortune 500
companies - are using Python on the server for anything. And we don't
have the ability to install/use another language on GAE, so from their
perspective, GAE is not 'less headache', as re-constructing their
companies to have lots of Python coders as opposed to J2EE/.NET/etc.
developers is very,very difficult. (Not to mention existing multi-
billion dollar investments in existing J2EE/.NET technologies already,
that would have to be scrapped... purchasing things that work in
Python instead)
My point is just - we could sell Python in the same way we could
justify Oracle Stored Procedure programming, its a language that the
vendor uses in their product. But to do this we need justification...
in Oracle's case it gave better performance. In the case of Oracle no
product compared enough that 'didn't' have custom 'stored procedure
languages' so people used Oracle. In the case of GAE though, people
can just choose EC2 instead.
>
> thanks for all the input, I'm still not sure which is the best
> platform but its always good to hear everyone thoughts
agreed!
>
> also, I did look at SimpleDB and it does seem very robust (and based
> on this post [1], quite performant as well). One thing that was
> missing that I currently need is Text/Blob attributes. You could use
> SDB for the simple attribute queries and S3 for the Text/Blob storage
> but then you obviously need the second query. Something to keep an eye
> on though, thanks
Seems SDB supports keys in the DB which reference Blobs in S3. I
think due to Amazon's pricing model, its more cost-efficient as you
pay per-DB-traffic usage, and downloading from S3 is cheaper.
>
> cheers
> brian
>
> [1] -http://www.daemonology.net/blog/2008-06-29-high-performance-
> simpledb.html
>
> On Nov 6, 4:42 am, Greg <[EMAIL PROTECTED]> wrote:
>
> > On Nov 5, 9:25 am, sal <[EMAIL PROTECTED]> wrote:
>
> > > On Nov 4, 3:08 pm, yejun <[EMAIL PROTECTED]> wrote:
>
> > > > On EC2 you can make your own image, but that's not the point. I
> > > > believe no one in their right mind would use a public image for their
> > > > production system, it is just not safe by any means.
>
> > > You have to trust the Google admins with your data and code also. At
> > > least with EC2 you have direct access and create an encrypted
> > > filesystem / etc. That even the admins wouldn't be able to 'poke
> > > around' into. So I don't buy the security argument...
>
> > Having admined a commercial linux cluster, I DO buy the security
> > argument. Even with my experience, I wouldn't flatter myself that I
> > was anywhere near the expertise of Google's admins.
>
> > Just taking one aspect, you should have iptables set up to firewall
> > your traffic. Do you know how to drop any SSH connections from the
> > same ip address that makes more than three attempts in a minute? If
> > not, it'll take you a couple of hours to get your head around that,
> > and build it into your EC2 image.
>
> > Multiply that by a couple of hundred times - database admin, database
> > replication, mail server admin, web server admin, managing DOS attacks
> > with routing tables... or you can have Google do the whole lot for you
> > - for next to nothing. I'm buying it.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Google App Engine" group.
To post to this group, send email to google-appengine@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/google-appengine?hl=en
-~----------~----~----~----~------~----~------~--~---
