Hugh Glaser wrote:
Thanks, Kingsley and Aldo.
I have to say you raise quite a lot of concerns, or at least matters of
interest.
I really don't think it is a big deal that I asked someone to consider
resources when accessing my web site, and I am a bit uncomfortable that I
then get messages effectively telling me that my software is poor and I
should be using (buying?) something else.
Hugh,
You're losing me a little, I don't think Aldo or I were making any
comments about your software per se. or making suggestions about
alternatives.
Anyway more comments inline below.
On 26/11/2008 02:12, "Kingsley Idehen" <[EMAIL PROTECTED]> wrote:
Hugh Glaser wrote:
I thought that might be the answer.
So what is the ontology of the error, so that my SW application can deal with
it appropriately?
If it ain¹t RDF it ain¹t sensible in the Semantic Web.
;-|
And the ³entitlement² to spend lots of money by accident; a bit worrying,
although I assume there are services that allow me to find out at least
estimates of the cost.
If you are querying via iSQL or the Virtuoso Conductor you wont be
moving lots of data between your desktop and EC2. If you do large
constructs over the sparql protocol or anything else that produces large
HTTP workloads between EC2 and your location, then you will incur the
costs (btw - Amazon are quite aggressive re. the costs, so you really
have to be serving many client i.e., offering a service for costs being
a major concern).
Er, yes, that was the question we were discussing.
Large constructs over the sparql prototcol.
With respect to costs, I never mentioned Amazon, so I am not sure why that
is the benchmark for comparison.
But I don't want to have a go at the Openlink software (I often recommend it
to people); I was just asking about limitations.
All software has limitations.
Anyway, Virtuoso let's you control lots of things, including shutting
down the sparql endpoint. In addition, you will soon be able to offer
OAuth access to sparql endpoint etc..
Yes, and I didn't really want to have the overhead of interacting with
Ravinder to explain why I had shut down his access to the SPARQL endpoint.
I suspect that your comment about a bill is a bit of a joke, in that normal
queries do not require money?
But it does raise an interesting LOD question.
Ravinder asked for LOD sets; if I have to pay for the query service, is it
LOD?
You pay for traffic that goes in and out of your data space.
(effective November 26, 2008)
Fixed Costs ($)
<snip amazon costs/>
Here is a purchase link that also exposes the items above.
https://aws-portal.amazon.com/gp/aws/user/subscription/index.html?ie=UTF8&offe
ringCode=6CB89F71
Of course, you can always use the Open Source Edition as is and
reconstruct DBpedia from scratch, the cost-benefit analysis factors come
down to:
1. Construction and Commissioning time (1 - 1.5 hrs vs 16 - 22 hrs)
2. On / Off edition variant of live DBpedia instance that's fully tuned
and in sync with the master
Getting back to dealing with awkward queries.
Detecting what are effectively DoS attacks is not always the easiest thing to
do.
Has Bezzo really solved it for a SPARQL endpoint while providing a useful
service to users with a wide variety of requirements?
I believe so based on what we can do with Virtuoso on EC2. One major
example is the backup feature where we can sync from a Virtuoso instance
into S3 buckets. Then perform a restore from those buckets (what we do
re. DBpedia). In our case we offer HTTP/WebDAV or the S3 protocol for
bucket access.
I don't think this contributes to helping to service complex SPARQL queries,
or have I missed somthing?
Hugh: I certainly had my response above a little tangled :-(
To clarify, re. Bezos and DOS bit.
1. EC2 instances can be instantiated and destroyed at will
2. Virtuoso (and I assume other SPARQL engines) have DOS busting
features such as SPARQL Query Cost Analysis and Rate Limits for HTTP
requests.
In fact, people don¹t usually offer open SQL access to Open Databases for
exactly this reason.
I like to think the day will come when the Semantic Web is so widely used
that we will have the same problem with SPARQL endpoints.
The Linked Data Web is going to take us way beyond anything SQL could
even fantasize about (imho). And one such fantasy is going to be
accessible sparql endpoints without bringing the house down :-)
Now there I agree.
The power of LD/SW or whatever you call it will indeed take us a long way
further.
And I agree on the fantasy, which is actually what I was saying all along.
It is a fantasy to suggest that "you can do all the wrong you want".
Exactly!
But I think it is sensible to take the question to a new thread...
No problem :-)
Kingsley
Best
Hugh
Kingsley
--
Regards,
Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com
