Add to Watchlist

GPU Datenbanken

0 views

Citation of segment
Embed Code
Purchasing a DVD Cite video

Formal Metadata

Title GPU Datenbanken
Title of Series FOSSGIS 2019
Author Kunde, Felix
License CC Attribution 4.0 International:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
DOI 10.5446/40709
Publisher FOSSGIS e.V.
Release Date 2019
Language German

Content Metadata

Subject Area Computer Science
Abstract Was sind GPU-Datenbanken? Wie viel Geo steckt schon drin? Eine kurze Live-Demo mit OmniSci wirds euch zeigen.
Keywords Graphics Processor Unit (GPU)

Related Material

Video is cited by the following resource
Series
Annotations
Transcript
Loading...
now comes the felix customer too
Age at the post and now tell us about gpu databases will fit
There are actually everywhere in the
various technological always create some hypes gpu database is also such a hype
It has been around for a few years I was also surprised that it was not a bit older so, what is are geodaten had so because they why are very fast geo data so very fast because they
Damn many cores have and parallel algorithms are much faster can now calculate is injured one cpu so you have to show here that it such an older picture of a benchmark and there we have here all these beautiful systems from ddb to even iphones park and down here somewhere the more that looks right impressive from there is one more another blog from a user all over compared many different systems has and there are also the gpu databases actually very far ahead so they then that's very good both there are different approaches to it There is a quite current youtube playlist which is called has change rates I also have databases here below linked that's all very nice it explains once that approach you hard disk your data in ssd and then always depending on the query data is loaded to the gpu or just the approach you have everything in cpu ram and the query planner club then always okay that is this query can now also calculate the gto below must calculate faster upload the data also at the Of course you have the first two approaches always the disadvantage that you first the Data must be energized that is always a bottleneck you can save yourself if you like the data right now the figure has the burner and according to the representatives also at these benchmarks are always very far lie in front are just the the Everything in the gpu calculate that's why do not know if the speed ask because it is really so accurate if that really linear is what i learned too have these videos are the gpu tattoo loading banks can do very well they can be structured very well alike data is processed, for example The matrices are very much machine learning be used and in the machines would be synonymous very much and gts they are wonderful for that they can also be incredibly fast data sorting are actually predestined for vacation request that not so accurate understands there are always these distinction between we change a lot and our data if we look many transactions the olg or us want to analyze the data that makes my so data warehouse applications and there the data is a bit too usually structured differently one talks since most of karl m stores also a slightly different data compression rate and you can always count on that make quick analysis query and that is synonymous with gpu databases of case and of course they are also very good in the race the ring so i mean Everyone probably knows that too the usa play will be used by That's why that's actually synonymous pretty close so i had in the booklet or the
also a live demo that's not quite true now I have a live demo here traffic data or not so taxi data from dresden uploaded to the map
it was something like 70
score millions of GPS and that's the do this be that earlier this time map the and as I told you Just have a look done I've put my bag away now You can see a few different ones here
Unfortunately, filter is a bit too fast just do not know either
why exactly the green nicole in that has made So I have to for the times speeds and can be easy filter on it define here and in this month chart I have the weekdays how many gps points per weekday and can filter that easily
and probably the coolest thing was that I have such spatial filters in here
could define and then too really live queried everything directly So it was not my 70 points, so to speak almost instant displayed and queried in the database so you can see from time to time
still so little loading bar but it is Actually, she was almost in real time
everything on the scanable so it shows for me that would be a new kind and wise with geodata with many geodata So to work and this dashboard
are always what you the the the
users because twitter or wherever else that actually throws contrary to that the main application the database of this is also open source of this dashboard However , I always think that's it not open source and that
actually with many others Representatives so if you are there
a bit more that set apart then at some point you also notice the but that, for example, variable data not so good are lines and polygon as a rule, usually very variable so you can not do so well you have to work on a few tricks Apply gpu memory is also very much the only thing that is always limited is the restriction one has one can yes say that you start there so wonderful for big Data around findings from huge datasets to find this great dashboards do not know you have there maybe 50 gigabytes available Ram might be even bigger in future but you always have to consider that and the systems of course with the to do this bottleneck is also a disadvantage there are sql interfaces with you can work on at web.de or not be that Everything but quite limited what you ge function had it was proprietary representatives since I was already very much but everything was better anyway other than robust so the systems bother are still many occasions and Just stay in touch with queries So what is not the same feeling as we post now and yes as I said give scope of function The open source representatives are right low So yes, it will show if the in the next years will continue to grow So right now they are very up Anyhow, that's what you put customers out for firmly ok
yes felix thank you for your head
Database
Graphics processing unit
Database
Graphics processing unit
Machine learning
Query language
User interface
Coma Berenices
Strukturierte Daten
WINDOWS <Programm>
Matrix (mathematics)
Graphics processing unit
World Wide Web
Milan <Programmiersprache>
World Wide Web
World Wide Web
World Wide Web
Uniform space
RAM
Inference
Loading...
Feedback

Timings

  309 ms - page object

Version

AV-Portal 3.8.0 (dec2fe8b0ce2e718d55d6f23ab68f0b2424a1f3f)