RE: Can we afford to offer SPARQL endpoints when we are successful? (Was "linked data hosted somewhere")

Hi
 

As noted before, we offer capacity for rent on EC2 with data and software
preloaded for those who want to have high traffic or complex queries.

But looking a bit further, there is even better coming down the pike.

When we host sets  significantly larger than Dbpedia, we will allow
absolutely unlimited queries but have a cap on resource  and/or time
After reaching the limit, we will return what was found in the time
allotted.  This is also for aggregating queries, trannsitive queries,
anything.

Take a use case:  You want to find people who have an uncommon interest in
common and do not know each other.  With 1M people and 100M interest triples
This takes a while.  If you ask this with a low quota, available for free,
you get something like John and Mary both like golf, which only 3 people
like and they do not know each other.  If you run longer, you would find
that 10000 people like golf and therefore this is not a distinctive interest
to begin with.

The point is, you get a feel for the data for free or at low cost and if you
want to run a one-off complex analytics question, you pay for it.  If you
want to run them all the time, you rent your own data center in the cklouds
or use one at your own facilities.

If a system is really out of capacity, it might refuse to continue with a
complex operation, in which case you would not pay for it. 

In this way, we have a continuum between the free for all simple query and
the expensive private data warehouse, no longer an either/or choice.

We do not believe that making query languages even more restricted than the
SPARQL  recommendation is a solution.  You can still make queries that run
forever in the most limited of languages.  On the other hand, we should
encourage people coming up with clever questions that can be answered
efficiently.  This stimulates intelligent use of the data.  For example, you
could run the 
distinctive interest query for free if you scope  it to people in your part
of town.  It is also likely that the answer with this qualification is more
interesting and relevant anyway.


Orri

Received on Friday, 5 December 2008 08:11:04 UTC