I like the idea of comparing live instances; could we pose a test-instance
challenge, with some benchmarks, and invite different communities to take
it up, hosting their own demos of what a well-tuned instance of WD could
look like? (Could also be hosted by us / spun up by advocates for a tool
in our community; could also spur some kaggle interest)
The size of the community actively interested in the health of Wikidata
seems complementary information; alongside overall community size/health
(which appears on the existing metrics list). //S
On Fri, Aug 27, 2021 at 10:19 AM Kingsley Idehen via Wikidata <
wikidata(a)lists.wikimedia.org> wrote:
On 8/25/21 3:17 PM, Mike Pham wrote:
Thanks for all suggestions, and general enthusiasm in helping scale WDQS!
A number of you have suggested various graph backends to consider moving to
from Blazegraph, and I wanted to take a minute to respond more generically.
There are several criteria we need to consider for a Blazegraph
alternative. Ideally we would have this list of criteria ready and
available to share, so that the community can help vet alternatives with
us. Unfortunately, we do not currently have a full list of these criteria.
While the criteria we judged candidate graph backends on are available
here
<https://docs.google.com/spreadsheets/d/1MXikljoSUVP77w7JKf9EXN40OB-ZkMqT8Y5b2NYVKbU/edit?usp=sharing>,
it is highly unlikely these will be the exact set we will use in this next
stage of scaling, and should only be used as a historical reference.
It is likely that there is no silver bullet solution that will satisfy
every criteria. We will probably need to make compromises in some areas in
order to optimize for others. This is a primary reason for conducting the WDQS
user survey
<https://www.wikidata.org/wiki/Wikidata:Project_chat/Archive/2021/08#Wikidata_Query_Service_(WDQS)_User_Survey_2021>:
we would like a better understanding of what the overall community
priorities are, including from those who may be less vocal in existing
discussions. These priorities will then be a major component in distilling
the criteria (and weights) for a new graph backend.
The current plan is to share the (most up to date as we can) survey
results at WikidataCon
<https://www.wikidata.org/wiki/Wikidata:WikidataCon_2021> this year. I
appreciate the discussion around potential candidates so far, and welcome
the continued insight/help, but wanted to also be clear that we will not be
making any decisions about a new graph backend, or have a complete list of
criteria or testing process, at the moment — WikidataCon will be the next
strategic check-in point.
As always, your patience is appreciated, and I’m looking forward to the
continuing discussions and collaboration!
Best,
Mike
—
*Mike Pham* (he/him)
Sr Product Manager, Search
Wikimedia Foundation <https://wikimediafoundation.org/>
Hi Mike,
Here's a suggestion regarding this important matter, circa 2021:
At the very least, a candidate platform should be able to deliver on a
live instance of the Wikidata dataset accessible for interaction via SPARQL
Query Services Endpoint.
Based on the interesting list of suggestions presented in this mailing
list (and in the Google Spreadsheet
<https://docs.google.com/spreadsheets/d/1MXikljoSUVP77w7JKf9EXN40OB-ZkMqT8Y5b2NYVKbU/edit#gid=0&range=M1>
it's spawned), the larger goal of a vibrant LOD Cloud Knowledge Graph would
benefit exponentially if each platform actually offered a live instance.
Irrespective of the final decision made, we are always going to offer a
live Wikidata instance, just as we do a LOD Cloud Cache etc..
Also note, the WDQS and SPARQL loose-coupling suggested by Jerven is
ultra-important, making that cool Query Services App independent of SPARQL
Query Service backend will improve utility and general resilience,
immensely.
*Links*
[1]
https://wikidata.demo.openlinksw.com/sparql -- Wikidata instance
we've been hosting for quite some time
[2]
http://lod.openlinksw.com/sparql -- 40 Billion+ Triples instance
(used to be the largest live SPARQL Query Service instance until Uniprot
dethroned it!).
[3]
https://medium.com/virtuoso-blog/on-the-mutually-beneficial-nature-of-dbped…
-- On the Mutually Beneficial Nature of DBpedia and Wikidata
Kingsley
On 25August, 2021 at 09:41:28, Samuel Klein (meta.sj(a)gmail.com) wrote:
Aha, hello jerven :) I should have remembered your earlier comment,
delighted you are here.
Thank you again for sharing your promising experience + benchmarks +
suggestions -- and for highlighting both similarities and differences.
SJ
On Tue, Aug 24, 2021 at 2:18 AM jerven Bolleman
<jerven.bolleman(a)sib.swiss> <jerven.bolleman(a)sib.swiss> wrote:
Hi Samuel, All,
I am the software engineer responsible for
sparql.uniprot.org.
I already offered to help in
https://phabricator.wikimedia.org/T206561.
So no need to ask Andra or Egon ;)
While we are good users of virtuoso, and strongly suggest it is
evaluated. As it is in general a good product that does scale.[1]
One of the things we did differently than WDQS is to introduce a
controlled layer between the "public" and the "database".
To allow things like query rewriting/redirection upon data model
changes, as well as rewriting some schema rediscovery queries to a known
faster query. We also parse the queries with RDF4J before handing them
to virtuoso. This makes sure that the queries that we accept are only
valid SPARQL 1.1. Avoiding users getting used to almost SPARQL dialects
(i.e. retain the flexiblity to move to a different endpoint). We are in
the process of updating this code and contributing it to RDF4J, with the
first contribution in the develop/4.0.0 branch
I think a number of current customizations in WDQS can be moved to a
front RDF4J layer. Then the RDF4J sail/repository layer can be used to
preserve flexibility. So that WDQS can more easily switch between
backend databases in the future.
One large difference between UniProt and WDQS is that WikiData is
continually updated while UniProt is batch released a few times a year.
WDQS is somewhat easier in some areas and more difficult in others
because of that.
Regards,
Jerven
[1] No Database is perfect, but it does scale a lot better than
Blazegraph did. Which we also evaluated in the past. There is still a
lot of potential in Virtuoso to scale even better in the future.
On 23/08/2021 21:36, Samuel Klein wrote:
Ah, that's lovely. Thanks for the update,
Kingsley! Uniprot is a good
parallel to keep in mind.
For Egon, Andra, others who work with them: Is there someone you'd
recommend chatting with at uniprot?
"scaling alongside uniprot" or at least engaging them on how to solve
shared + comparable issues (they also offer authentication-free SPARQL
querying) sounds like a compelling option.
S.
On Thu, Aug 19, 2021 at 4:32 PM Kingsley Idehen via Wikidata
<wikidata(a)lists.wikimedia.org <mailto:wikidata@lists.wikimedia.org>>
wrote:
On 8/18/21 5:07 PM, Mike Pham wrote:
>
> Wikidata community members,
>
>
> Thank you for all of your work helping Wikidata grow and improve
> over the years. In the spirit of better communication, we would
> like to take this opportunity to share some of the current
> challenges Wikidata Query Service (WDQS) is facing, and some
> strategies we have for dealing with them.
>
>
> WDQS currently risks failing to provide acceptable service quality
> due to the following reasons:
>
> 1.
>
> Blazegraph scaling
>
> 1.
>
> Graph size. WDQS uses Blazegraph as our graph backend.
> While Blazegraph can theoretically support 50 billion
> edges <https://blazegraph.com/>, in reality Wikidata is
> the largest graph we know of running on Blazegraph (~13
> billion triples
> <
https://grafana.wikimedia.org/d/000000489/wikidata-query-service?viewPanel=…
),
> and there is a risk that we will reach a size
> <
https://www.w3.org/wiki/LargeTripleStores#Bigdata.28R.29_.2812.7B.29
limit
of what it can realistically support
<https://phabricator.wikimedia.org/T213210>. Once
Blazegraph is maxed out, WDQS can no longer be updated.
This will also break Wikidata tools that rely on WDQS.
2.
Software support. Blazegraph is end of life software,
which is no longer actively maintained, making it an
unsustainable backend to continue moving forward with long
term.
Blazegraph maxing out in size poses the greatest risk for
catastrophic failure, as it would effectively prevent WDQS from
being updated further, and inevitably fall out of date. Our long
term strategy to address this is to move to a new graph backend
that best meets our WDQS needs and is actively maintained, and
begin the migration off of Blazegraph as soon as a viable
alternative is identified
<https://phabricator.wikimedia.org/T206560>.
Hi Mike,
Do bear in mind that pre and post selection of Blazegraph for
Wikidata, we've always offered an RDF-based DBMS that can handle
current and future requirements for Wikidata, just as we do DBpedia.
At the time of our first rendezvous, handling 50 billion triples
would have typically required our Cluster Edition which is a
Commercial Only offering -- basically, that was the deal breaker
back then.
Anyway, in recent times, our Open Source Edition has evolved to
handle some 80 Billion+ triples (exemplified by the live Uniprot
instance) where performance and scale is primary a function of
available memory.
I hope this helps.
Related:
[1]
https://wikidata.demo.openlinksw.com/sparql
<https://wikidata.demo.openlinksw.com/sparql>-- Our Live Wikidata
SPARQL Query Endpoint
[2]
https://docs.google.com/spreadsheets/d/15AXnxMgKyCvLPil_QeGC0DiXOP-Hu8Ln97f…
<
https://docs.google.com/spreadsheets/d/15AXnxMgKyCvLPil_QeGC0DiXOP-Hu8Ln97f…
-- Google Spreadsheet about various Virtuoso Configurations
associated with some well-known public endpoints
[3]
https://t.co/EjAAO73wwE <https://t.co/EjAAO73wwE> -- this query
doesn't complete with the current Blazegraph-based Wikidata endpoint
[4]
https://t.co/GTATPPJNBI <https://t.co/GTATPPJNBI> -- same query
completing when applied to the Virtuoso-based endpoint
[5]
https://t.co/X7mLmcYC69 <https://t.co/X7mLmcYC69> -- about
loading Wikidata's datasets into a Virtuoso instance
[6]
https://twitter.com/search?q=%23Wikidata%20%23VirtuosoRDBMS%20%40kidehen&am…
<
https://twitter.com/search?q=%2523Wikidata%20%2523VirtuosoRDBMS%20%2540kide…
-- various demos shared via Twitter over the years regarding
Wikidata
--
Regards,
Kingsley Idehen
Founder & CEO
OpenLink Software
Home
Page:http://www.openlinksw.com <http://www.openlinksw.com>
Community
Support:https://community.openlinksw.com <
https://community.openlinksw.com>
https://medium.com/openlink-software-blog>
https://medium.com/virtuoso-blog>
Data Access Drivers Blog:
https://medium.com/openlink-odbc-jdbc-ado-net-data-access-drivers <
https://medium.com/openlink-odbc-jdbc-ado-net-data-access-drivers>
Personal Weblogs (Blogs):
Medium
Blog:https://medium.com/@kidehen <
https://medium.com/@kidehen>
http://www.openlinksw.com/blog/~kidehen/>
http://kidehen.blogspot.com>
https://www.pinterest.com/kidehen/>
https://www.quora.com/profile/Kingsley-Uyi-Idehen>
https://plus.google.com/+KingsleyIdehen/about>
http://www.linkedin.com/in/kidehen>
Web Identities (WebID):
Personal:
http://kingsley.idehen.net/public_home/kidehen/profile.ttl#i <
http://kingsley.idehen.net/public_home/kidehen/profile.ttl#i>
:
http://id.myopenlink.net/DAV/home/KingsleyUyiIdehen/Public/kingsley.ttl#this
<
http://id.myopenlink.net/DAV/home/KingsleyUyiIdehen/Public/kingsley.ttl#this
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
<mailto:wikidata@lists.wikimedia.org>
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
<mailto:wikidata-leave@lists.wikimedia.org>
--
Samuel Klein @metasj w:user:sj +1 617 529
4266
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
--
*Jerven Tjalling Bolleman*
Principal Software Developer
*SIB | Swiss Institute of Bioinformatics*
1, rue Michel Servet - CH 1211 Geneva 4 - Switzerland
t +41 22 379 58 85
Jerven.Bolleman(a)sib.swiss -
www.sib.swiss
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
--
Samuel Klein @metasj w:user:sj +1 617 529 4266
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
--
Regards,
Kingsley Idehen
Founder & CEO
OpenLink Software
Home Page:
http://www.openlinksw.com
Community Support:
https://community.openlinksw.com
Weblogs (Blogs):
Company Blog:
https://medium.com/openlink-software-blog
Virtuoso Blog:
https://medium.com/virtuoso-blog
Data Access Drivers Blog:
https://medium.com/openlink-odbc-jdbc-ado-net-data-access-drivers
Personal Weblogs (Blogs):
Medium Blog:
https://medium.com/@kidehen
Legacy Blogs:
http://www.openlinksw.com/blog/~kidehen/
http://kidehen.blogspot.com
Profile Pages:
Pinterest:
https://www.pinterest.com/kidehen/
Quora:
https://www.quora.com/profile/Kingsley-Uyi-Idehen
Twitter:
https://twitter.com/kidehen
Google+:
https://plus.google.com/+KingsleyIdehen/about
LinkedIn:
http://www.linkedin.com/in/kidehen
Web Identities (WebID):
Personal:
http://kingsley.idehen.net/public_home/kidehen/profile.ttl#i
:
http://id.myopenlink.net/DAV/home/KingsleyUyiIdehen/Public/kingsley.ttl#this
_______________________________________________
Wikidata mailing list -- wikidata(a)lists.wikimedia.org
To unsubscribe send an email to wikidata-leave(a)lists.wikimedia.org
--
Samuel Klein @metasj w:user:sj +1 617 529 4266