Merken

Big Data meets Fast Data

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
that OK and so from the last part of the system would have you can think of this and the end of the the 2nd the the you the so thank we you work for so I want it old the interviewing or gender violence that name of the commercial rather than the region of the start of the year and the idea is to talk about all of the degree of system that is really focus on the data that has and cooperate with fasting memory using even within that really has to deal with the transactional workloads here for the use of machine learning or some more complex querying data into that in the book so problem you can make this integration easier
so we talk about the data right most of the time so you have a problem where you we have a single system or a single process or a single gene that has to do with all of these and it's that the problem right so let's imagine that the idea he wants to set out in all this balls and batted through them you right 1 approach
to this so vernalization is 1 way to do it it works but I love to eliminate and if we you thinking that it's a single guy doing all the work it would take a lot of time and so what you're trying to suggest here as an alternative is something like this the
so let's your wife right so I mean it's not essentially the ideal scenario and that's what you're trying to deal with this integration interval from so I'm not through everyone last
year when they listened to what happen but I am nicely to summarize is that it's the highly distributed also can be theories there is this idea of acid you will have both the full set of nodes and the data is spread across all these notes petition right here there is also the idea of using a specific protocol right all you get this it is also available and well then you can uh quarterly arise from files or from other systems so that very fast because the protocol right to wineries system smart enough to colony see in it by all needs available that machine and man those 2 files or something as needed and then lol that reading like of the or external temporary table for example and then set the dates of all the segments notes to media right so it's it's very fast but you're still doing that leveraging and used to be you can't write with the trial and the ideal across all by you that you have to build fire from from from this so we look at this and says OK but if the from right so the other problems that we have people or non sexually active rather to call the Vatican and that yield this are focused on the memory so you have this notion of clusters have multiple nodes and you have a lot of customers are using this conversion and we have of a transactional work load the deal with the data that the memory as well so why not leverage these other problems here any great book so there is the concern of works often in dealing with data more sensitive daylight and financial industries or or that's the resale right if you don't want to lose so in the case of yield scenes in the project actually started to do a work on the use of the financial industry right there are a lot of features states so in part of the project that that tries to solve this problem of the loss so for example whenever you wouldn't 3 inside the cluster you can specify the number of of correctly because it is 1 of the most let's see what the rest of the West of as wages share or a little bit more of a model of the forces and 1 of the most the number of this fired cluster is going to be established that talking it's 1 of the things right so with you that was used you happen to have some you you want to do in this data in memory right and you can do that using out of any that fails you can restart the this by the books and you are actually cooperating with the that level the over he deals with the network petition so the reason areas of this has been seen as more Fernandez enable lighting you have network for each detection it also supports the wind the right so this is very nice because 1 of the things about our many systems for many many customers are using system that doesn't have this teaches that OK I have this database nervous system running on 1 of the the but now I also want to have a job but it isn't really the center or actually want to operate all the centers at the same time so how do you do that so we offer when the litigation as well in the CCC memory right people sometimes get scared where we see now about half of the day of the crash mileage was already a soul although using memory you can also specify this to be actually precision on this and then we can have some more problems of to report any of these so even if your entire closes down you can over all the data from this in a in a in a very easy and fast way In other words we do know little about the work place and a so too any perspective of numbers so these are 2 uh use cases that we have at this using the little Jim fire but just for you to understand how the little part but entirely as a affected you still have the same people based sustained works the same way so China Radio ending array of if you actually do the math you have here a lot of transactions Computing users and what not but our way to look at those numbers used by the population so if you just go by year like joking fairly and say that that we're at responsible for
36 per cent of the work of legitimate political segregation away so that the time that people may not know about them far dual so that was also 1 of the themes that started the trigger the idea of sharing the golden mean you open source of this proper so getting back to the original problem right the idea then was because the leverage did get this but instead of moving there from our this right from files just updated out from from that from that yield so this this concept of data which is like a little faster and also to then be doing things go over the ordinal then in cluster you so in the store and say I have formed here and assume that this connection established with all those notes from this point on each segment that problem will then start talking directly with the server it here on TOD and hopefully this is a work in progress by the way is still under development we have enough of it but it's better with the ball with use you working for example trying to solve the problem of passion or petition to the that you will we will partition is not exactly the same way as you were usually so the ideal solution would actually be if we will get the same scheme of partitioning is really match at 1 2 1 1 1 7 decided not to 1 serving the size so the talk about the same data set right so moving on a little bit this individual as you might know you see of the possible use the center-right are ideally this presentation will have available so I'm not able to do that because of technical reasons so this involves a lot of millions of you well but so imagine this area right so you have our pretty virtual sections so since the money but you want to do that very fast because of course you know people waiting in the waiting for the proposition that we have so we are going to operate and that everything in memory is to destabilize the others have actually been settled down pop art dealer then that they had this much the offloaded to bring book we will then you in the manner which is a data science machine learning of open-source source from as well it was a fairly new to the central part of managing the 2 so that in that the only thing I met recently in the last topic historical is said to be what you can do that in the 5 possible problems that works another way to actually and that is that right so 1st if you actually look at these user and its energy requirements you wouldn't want more people about the right so that's actually that because were sort of the source so that the use of resources solutions in much the same so the real time interval and 1st logo that when you have also have are possible adverse or POS simulator and also you have a to use the rest interface available and yield to Bush said something of deal then going to balance the distribution to be 1 of the things we know that we will be ready from time to time similar instances of Lake Placid precipitously or what's the probability of having a problem virtue was equal rights so that based on their let's say I'm I'm sure some of you might want write about all the time charged by a different a different amount of power and the difference in the rate is based on the the code doesn't know that you will have more possibilities of having brought or something like that so that you have here a lot in the seems to include the following process the new as well and then borrow at point which is just a matter of you know this is the 1st to show NBC poll dates from GEO through a feature called the the use of speech right so so you can leave whenever I want dataset from there is actually going to just monetary this cleared up as soon as a new entry Lindsay that data is actually going to push to the left line so here is I need to have all the y so was granted actually show that I think that a very is that we have like 1 of the green dots popping up in the wild signed in whenever section is the widest possible violence let you that that that's pretty much it and another way let's look
at that so here are the collection of sequence in the ones in red or if you want to write out the ones that actually had more problems actions the and so we are
of course but the source of looking forward you have a contributions and at any moment that might have evolved in solution variables to this problem or even for all the a curious thing the items that have been our years old for the old or 4 meter and so if you're in Java and with your own hours was was was really blood although we might still be using our native land for geo there's also some people lives there as well this article for the demo they're supposed to show you will be available on this the year and that all others and available even for dual form rate of and when we know to was of course we are tied so interested in working with us in this project as no some means references and yes ready here's the thing I know they much time for questions thank you of we for this integration of well the major problem they're not as good at catching so attractive and how it is that it's the only way we serialize the degree of the French might not work because we're going to see was there is a holiday and the loss of the sense of being the kind of relation We that will on opened the working on that the and and communities will not be not really you using screen as part of the field but there is a fire the field of the moral way for user to solving dual but I not really for this integration because very often about performance so we're going like low-level dollars in the yeah the this the but this is the very soul of the and would red so riddled with let's say the best 4 years of the good so what is the best thing I maybe a deal so there is this class of systems that have become very popular around 2 thousand of the of the vessel the project below the bottom part of it exists as a life for 4 years already right handed down from your address so that they were not going fast enough and financial issues like like someone in Africa on supporting particles actions so that we questions polynomially reiterated stable and operational similarly database but just you know what the worst thing you may not be enough because the data so that is not working with recognition institution and even how to proceed with this 1 was doing different way in a slightly different way than traditional databases but we will of the and that's the thing that the cost 1st the gestures like we did before and do you plan to extend this to you like to find the roots of humans and of the right to the beach where we actually start working on each vessel and 4 so the same way we are obligated to read to be lowered to this to do that with a fast now it's friendly in our policies explores the source code is available is quite of our field or anything already do but we're not really including our next we use quite a so that we actually start paying more feedback like this would be really interesting we can certainly were again and it will or something like that the city the present course part we have a spot on her as well for the field and its considered that is the right now just we or something that we can generate over like the more it but as of England and would you use to do all drafts and would like to think of our 1st time series definitely for graphs I guess when we don't we support or have any native support for vector in it so that we can something that let's say if we have the the the Grecians part I will definitely let that happen in this I you just some data from from from dual or something like that so I OK
Skalierbarkeit
Abfrage
Physikalisches System
Komplex <Algebra>
Hybridrechner
Computeranimation
Integral
Datenhaltung
Beanspruchung
Minimalgrad
Geschlecht <Mathematik>
Festspeicher
Mereologie
Algorithmische Lerntheorie
Prozess <Physik>
Äußere Algebra eines Moduls
Physikalisches System
Schwebung
Hybridrechner
Umsetzung <Informatik>
Sensitivitätsanalyse
Einfügungsdämpfung
Bit
Gemeinsamer Speicher
Atomarität <Informatik>
Mathematisierung
Systemzusammenbruch
Abgeschlossene Menge
Zahlenbereich
Computerunterstütztes Verfahren
Sommerzeit
Physikalische Theorie
Datenhaltung
Übergang
Demoszene <Programmierung>
Virtuelle Maschine
Knotenmenge
Informationsmodellierung
Perspektive
Prozess <Informatik>
Cluster <Rechnernetz>
Gammafunktion
Metropolitan area network
Umwandlungsenthalpie
Datennetz
Protokoll <Datenverarbeitungssystem>
Datenhaltung
Physikalisches System
Elektronische Publikation
Hybridrechner
Chipkarte
Integral
Beanspruchung
Transaktionsverwaltung
Flächeninhalt
Forcing
Menge
Rechter Winkel
Festspeicher
Hypermedia
Mereologie
Wort <Informatik>
Projektive Ebene
Tabelle <Informatik>
Aggregatzustand
Distributionstheorie
Bit
Subtraktion
Prozess <Physik>
Punkt
Sprachsynthese
Kombinatorische Gruppentheorie
Räumliche Anordnung
Code
Arithmetische Folge
Diskrete Simulation
Speicher <Informatik>
Softwareentwickler
Algorithmische Lerntheorie
Gerade
Leistung <Physik>
Einfach zusammenhängender Raum
Open Source
REST <Informatik>
Aussage <Mathematik>
Nummerung
Ähnlichkeitsgeometrie
Quellcode
Bitrate
Elektronische Publikation
Hybridrechner
Partitionsfunktion
Quick-Sort
Arithmetisches Mittel
Energiedichte
Skalarprodukt
Echtzeitsystem
Menge
Flächeninhalt
Rechter Winkel
Festspeicher
Mereologie
Server
Garbentheorie
Instantiierung
Rückkopplung
Demo <Programm>
Subtraktion
Folge <Mathematik>
Einfügungsdämpfung
Momentenproblem
Klasse <Mathematik>
Applet
Gruppenoperation
Ungerichteter Graph
Kombinatorische Gruppentheorie
Räumliche Anordnung
Datenhaltung
Eins
Quellcode
Variable
Bildschirmmaske
Zeitreihenanalyse
Meter
Wurzel <Mathematik>
Touchscreen
Videospiel
Datenhaltung
Relativitätstheorie
Physikalisches System
Vektorraum
Quellcode
Bitrate
Mustererkennung
Hybridrechner
Integral
Arithmetisches Mittel
Diskrete-Elemente-Methode
Minimalgrad
Datenfeld
Rechter Winkel
Mereologie
Räumliche Anordnung
Dualitätstheorie
Projektive Ebene
Partikelsystem
Computeranimation

Metadaten

Formale Metadaten

Titel Big Data meets Fast Data
Untertitel An scalable hybrid real transactional and analytics solution
Serientitel FOSDEM 2016
Teil 110
Anzahl der Teile 110
Autor Markito, William
Lizenz CC-Namensnennung 2.0 Belgien:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/30930
Herausgeber FOSDEM VZW
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik

Ähnliche Filme

Loading...