Multi-Model NoSQL Databases


Formal Metadata

Multi-Model NoSQL Databases
Title of Series
Part Number
Number of Parts
Hackstein, Michael
CC Attribution - NonCommercial 2.0 Germany:
You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal and non-commercial purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Free and Open Source software Conference (FrOSCon) e.V.
Release Date

Content Metadata

Subject Area
Multi-Model NoSQL Databases An introduction to Polyglot persistence and NoSQL In this talk i will introduce and discuss the term polyglot persistence in the NoSQL world. Afterwards i will present the multi-model approach of NoSQL database that try to overcome some drawbacks of polyglot persistence setups while keeping the benefits. The database i will use for this offers embedded Javascript for my examples. In many modern applications the database side is realized using polyglot persistence – store each data format (graphs, documents, etc.) in an appropriate separate database. This approach yields several benefits, databases are optimized for their specific duty, however there are also drawbacks: * keep all databases in sync * queries might require data from several databases * experts needed for all used systems A multi-model database is not restricted to one data format, but can cope with several of them. In this talk i will present how a multi-model database can be used in a polyglot persistence setup and how it will reduce the effort drastically. I will show with Javascript examples how to make use of such a multi-model database. ······························ Speaker: Michael Hackstein Event: FrOSCon 2014 by the Free and Open Source Software Conference (FrOSCon) e.V.
Free and Open Source Software Conference
Freeware Scientific modelling Open source Content (media) Line (geometry) Degree (graph theory) Database Personal digital assistant Term (mathematics) Database System programming Information Bimodal distribution
Trail Debugger Projective plane Core dump Grass (card game) Graph (mathematics) Degree (graph theory) Degree (graph theory) Process (computing) Video game Visualization (computer graphics) Database Database Touch typing System programming Information systems Scripting language Information
Read-only memory Addition Spacetime Sequel Relational database File format Cartesian coordinate system Code Table (information) Data model Single-precision floating-point format Database Normal (geometry) Quicksort Object (grammar) Social class Physical system
Data model Sequel Data storage device Database Similarity (geometry) Price index Object (grammar) Data type Subtraction Multiplication Table (information) Attribute grammar
Group action Theory of relativity Sequel Length Graph (mathematics) Attribute grammar Data model Subject indexing Data storage device Query language Database Object (grammar) Multiplication
Key (cryptography) Set (mathematics) Graph (mathematics) Binary file Data model Data storage device Term (mathematics) Database String (computer science) Operator (mathematics) Charge carrier Object (grammar) Multiplication
Query language Server (computing) Implementation Service (economics) Code State of matter Direction (geometry) Scientific modelling Finitary relation Process modeling Virtual machine Complete metric space Graph (mathematics) Attribute grammar Database Energy level Position operator Scaling (geometry) Constraint (mathematics) Key (cryptography) Bit Graph (mathematics) Process (computing) Hash function Database Data storage device Vertex (graph theory) Right angle Data structure
Point (geometry) Query language Theory of relativity Process (computing) Transformation (genetics) Finitary relation Process modeling Cartesian coordinate system Graph (mathematics) Graph (mathematics) Attribute grammar Data model Pointer (computer programming) Structured programming Database Data storage device Database Operator (mathematics) Mixed reality Query language Website Object (grammar) Data structure Data structure
Slide rule Decision tree learning Scientific modelling Finitary relation Process modeling Relational database Strukturierte Daten Theory Attribute grammar Database Query language Subtraction Multiplication Units of measurement Address space Form (programming) Physical system Product (category theory) Relational database Library catalog Cartesian coordinate system Table (information) Database Data structure Physical system Inductive reasoning
Home page Product (category theory) Decision tree learning Product (category theory) Sequel Relational database Library catalog Set (mathematics) Library catalog Graph (mathematics) Mereology Goodness of fit Database Physical system Physical system
Product (category theory) Asynchronous Transfer Mode Product (category theory) Theory of relativity Library catalog Graph (mathematics) Computer Attribute grammar Single-precision floating-point format String (computer science) Database Data type Subtraction
Product (category theory) Asynchronous Transfer Mode Observational study Code Transformation (genetics) Texture mapping File format Graph (mathematics) Focus (optics) Database Query language Logic Energy level Cuboid Data storage device Subtraction Overhead (computing) Product (category theory) Mapping Graph theory Single-precision floating-point format Category of being Data storage device Logic Natural number Synchronization Charge carrier Statement (computer science)
Overhead (computing) Numerical digit Code Texture mapping Transport Layer Security Direction (geometry) System administrator File format Client (computing) Graph (mathematics) Mereology Focus (optics) Attribute grammar Linker (computing) Database Single-precision floating-point format Query language Logic Data structure Subtraction Formal grammar Physical system Overhead (computing) Product (category theory) Constraint (mathematics) Theory of relativity Content (media) Set (mathematics) Cartesian coordinate system Graph theory Database normalization Data storage device Query language Natural number Synchronization Normal (geometry) Object (grammar) Freeware Directed graph Library (computing)
Code File format Model theory Graph (mathematics) Cartesian coordinate system Mechanism design Data model Message passing Exterior algebra Database Query language Database Subtraction Multiplication Physical system
Ocean current Product (category theory) Model theory Auto mechanic Graph (mathematics) Event horizon Mechanism design Data model Data model Database Data storage device Database Multiplication
Database transaction Multiplication Database transaction Water vapor ACID Parameter (computer programming) Graph (mathematics) Bookmark (World Wide Web) Formal language Data model Data storage device Personal digital assistant Database Energy level Graph (mathematics) Subtraction Multiplication Reverse engineering
Group action Database transaction Mapping Code Interactive television ACID Client (computing) Complete metric space Raw image format Revision control Data model Process (computing) Database Energy level Software framework Right angle Graph (mathematics) Procedural programming Object (grammar) Quicksort Multiplication Resultant Library (computing)
Database transaction Digital filter Musical ensemble Scaling (geometry) File format Direction (geometry) Element (mathematics) Workstation Planning Graph (mathematics) Graph (mathematics) Workstation Formal language Human migration Flow separation Subject indexing Query language Game theory Logic gate Game theory Abstraction
Point (geometry) Operations research Default (computer science) Database transaction Group action Product (category theory) Database transaction Code Moment (mathematics) Bit ACID Group action Rollback (data management) Functional (mathematics) Query language Logic Natural number Function (mathematics) Database
Game controller Concurrency (computer science) Authentication Open set Information privacy Encapsulation (object-oriented programming) Thermodynamischer Zustand Revision control Goodness of fit Computer network Database Logic Encryption Energy level Directed set Software framework Information security Multiplication Constraint (mathematics) Information Usability Encapsulation (object-oriented programming) Functional (mathematics) Database Logic Revision control Charge carrier Object (grammar) Units of measurement Resultant
Authentication Java applet Encapsulation (object-oriented programming) Cartesian coordinate system Revision control Sic Database Computer network Database Password Revision control Module (mathematics) Directed set Data management
Polar coordinate system Overhead (computing) Authentication Limit (category theory) Client (computing) Cross-site scripting Insertion loss Computer network Database Energy level Directed set Office suite Subtraction Multiplication Authentication Algorithm File format Structural load Moment (mathematics) Java applet Login Representational state transfer Encapsulation (object-oriented programming) Cartesian coordinate system Local Group Tablet computer Inversion (music) Process (computing) Error message Intrusion detection system Database Revision control Resultant
Overhead (computing) Mapping Synchronization Texture mapping Database Synchronization Query language File format Graph (mathematics) Focus (optics)
Overhead (computing) Texture mapping System administrator File format Mereology Focus (optics) Linker (computing) Database Synchronization Query language Logic File viewer Object (grammar)
Batch processing Building Server (computing) Java applet Range (statistics) File format Graph (mathematics) Replication (computing) Formal language Revision control Device driver Strategy game Causality Database Single-precision floating-point format Operating system Authorization Subtraction Information security Addition Algorithm Uniqueness quantification Projective plane Open source Set (mathematics) Replication (computing) Vertex (graph theory) Backup
Trail Building Server (computing) Service (economics) State of matter Virtual machine 1 (number) ACID Point cloud Weight Replication (computing) Field (computer science) Theory Mathematics Bit rate Database Single-precision floating-point format Videoconferencing Energy level Damping Subtraction Scaling (geometry) Relational database Projective plane Basis (linear algebra) Bit Sequence Benchmark Connected space Word Integrated development environment Query language Phase transition Computer network Data center Backup Right angle Digitizing
and then all I would like to tell you some things about uh Polyglot Persistence and multimodal databases so this talk will be about a short introduction into this no secret term and then we'll find out what is this all about persistence and removal 2 features of which model databases and policy victims into this content concept of quality assistance so about myself my name is Chad lines as I'm working on during
the time of the year and the court team which is based in Cologne and I am responsible for or the front ends Graphs visualisation graph features and we have all that Britain and other scripts and i'm jobs developed and and every time and organizing Claudius letters the
JavaScript User Group in Cologne and we also have a track frost con which is upstairs and the project rooms and I think today there will be 3 4 3 more books and attract so if you're interested transcript can come there and the my professional life of older master's degree was a specialization in databases and information systems and that is how I got in touch but enough about myself let's start with the my story
so sometime go and there was this kind of dark that used could the
use a relational database or it something wrong so actually we just created always the relational database put everything into 1
the formant and if you have an additional and they have it as an additional class or object in your application you just add another table connected somehow and if you do that a long time something like this in a quite large system really ends up and it's really hard to manage this and so and to get an overview of where sort but the idea why you do that is that the normalized monastery space efficient it's memory efficient and it's quite possible that every but the problem is you have to force your data formats always into this relational normalized formant and you have to let yourself so you have to do that in the application codes and you cannot just store the native data object that you have an application directly into his into the database let's with the rise of no sequel this actually change so give you just an overview
of the no sequel database that
around 3 major types the first one to the document database so the document database actually stores mostly Jason documents and you group them which mythological similarity into collections so to get the step from the relational were the collection is basically quite equal to a table but you do not have to define the column names so there no secret documents for can live with different attribute names in different objects in the same collection and then you can also add curious and indices on top of these documents so it's not a problem
if there is some attribute which is not filled that's simply ignored by the index and but all the others can be really getting really fast by the but if you have an index on certain actions thing that were then we have a graph databases the idea of graph databases is that there really focused on end-to-end relations between so an object is again some kind of document but to be cannot be that complex so documents for it's also possible to store nested objects and object and objects in objects and so on as 1 entry in the database and you can carry on the internal attributes also this is typically not possible
in graph databases the idea of graph databases simple documents connected to each other and carry on path length of 1 passes on arbitrary lengths so on graph database it is easy to get that from a to B with at most 3 steps in between so 1 step 2 step all 3 steps if you wanna do that in a sequel query you would have to write its receptor is 1
4 111 joint set to leverage on entry-level joints to get the same result so for these various graph databases are really really good this and the last 1 it's I would
say the simplest 1 key store the idea that offered to various that you have 1 unique that is unique in
your complete database and to this key to connect any arbitrary object the database doesn't know anything about the subject so it can store binary files or Jason all strings or whatever it doesn't care and you can only get this subject back using the key but you don't get carrier operations on the object and it can just very by you cannot say I know that in my objects is always that you name it can get that you have to put the name if you want to cry that into the key for efficient term
occurring in that because of this constraint that the database does not know anything of the continent's story it is really really easy to implement scaling and positioning because you just have this 1 piece that that is unique in your code the complete database use hash function that and then put it on any server you can just add millions of services and its unique which server stores the state for style for example it's quite hard to get because if if you shot you could still shot by any key for the owner was a document that's not a problem but if you want to carry it you want to carry on attribute level and maybe the attribute is not the 1 that you shot by then you have to connect the collects the data from all the services that's why it's a bit harder to and use shining on documents costs and the rest it's even worse because you have to and actually make sure that the vertices on the same machine as the edge to get frustrated and let in both directions so it's quite hard to really shot graph database that's like most graph databases don't even supported magnets talk about polyglot modeling the idea of polyglot monolingual Polyglot Persistence this to use the right
tool for the job so take a look at
your application what data you want to store and then you pick the database that can natively so data from so if you have structured data like Jason object then we should probably use documents for but just fire your uh Jason object to the documents for sausage and no transformations to be done neither on your site monitored and site and then if you have relations between these objects or and simple object and you want to carry them efficiently along the process then you should probably use a graph database and it is OK I don't care I just want maximum performance I don't want to carry on attribute that and I want to mention structure myself that you can use key-value operator of that is really fast at that point and you can scale
industry but if you have a mix of these and I think most
applications actually have that then you could use a multimodal database and 1 with model databases I will tell you a few minutes let's go to the use case example and e-commerce system so I think the that is theory quite simple you have some the product catalog they put all your products in the relational world you would have to normalize it because the T-shirts achieving these are 2 different entries to different tables relational what they different attributes induction installed you could put that is a collection because and it can handle that there different forms the same collection then you have your customers can also be that and relational database and they are quite similar and but also there's some problems so if you are only in Germany it's quite easy to make 1 customer and table and that but if you support German addresses and for example American addresses or whatever addresses they also quite differently so they need different attributes and you could say stolid in relational comet using different table forestall some Jason inside it but then you lose actually the benefit of a relational database because you cannot very Jason or other stuff inside your unit himself and the next thing you have is the state's history on the top left
slide which does just false which users bought which items and their average price and all this stuff and then you have the shopping cart the shopping cart you want to be really really false inaccessible because you need
that on every page of the system but the start of it is quite outside and the last thing is the most important thing of offer goods become a system is recommendations because you want to recommend good products to you bias so that by other products or also get more money for but recommendations I actually the graph problem because if you suspect products and
other users buying the same product and then you can suggest other products and that's a graph so probably would like to use the graph part of a graph databases but anyway you could do that international database like this but it's I think it's quite hard to manage and how to handle it quite hard to set up it can and it's well but had to get an overview that if we are now moving sequel world we have the same set of the same e-commerce systems we still have the product catalog which is
simply documents in same collection and the actually don't care for all these different attributes you just care for for the same attributes that are shared between your products like the price and stuff and I and this is they put putting in this then you have your customers where the documents so again I could live with different interests and different egress types and also misses documents but then you have sense 3 which is very much the same then you have shopping carts we actually just need the session ID and then some entries but if you want to carry that you will only cover that session ID whenever such for shopping carts containing a computer we just have 1 user ID and get it shopping and his shopping cart and that's it and then for the
recommendations already told you it's actually graph problems so once for the relations between product catalog and the customs in the graph database and if you do that been nowadays most famous technologies and we would use the documents for him up there and for the a string it could be used for for the shopping cart and the graph database
for the recommendation and was nowadays top leaders in these and different categories we could do this was among the Debian the documents for level was the of Aegean across for level and that is for example is shopping so these are just examples of different of secret database that supports for so
what are the benefits if you use this set up instead of 1 and
the creation of so the benefit is that you have a natural mapping of the data into database the study of graph graph databases store you products in the document store and a shopping cart became based store and you don't have to write any code to transform that ought to optimize that because databases rebuild for these kind of data then you also get the carriers that are tailored for your data for which so you don't have to create like huge it's SQL statements and to simulate a graph theory you get that out of the box in so benefits most of the benefit is that you can really focus on writing a business logic you don't have to write any transformation code and from your local you status to structure into the relational
structure and no normalization require could just simply use a data structure and that that yes just yes that is actually 1 of the problems I'm coming to that that in a few minutes ago because most documents for sexual don't support and OK so but what is the overhead that began with the set cost you guys don't get anything for free and so part of the letters that the data has to to be stored redundantly and has to be kept the link so for the recommendation examples we need the graph capabilities of the graph database but the actual content a stored in the document and we have to make sure that whatever is in the document and a document store this referenceable from the data graph database because we cannot connected those 2 together the the thing is you have to set up 3 different databases just to get you become a system running and that's a
lot so the Administration is really huge fought for such system outside concepts quite small the single thing of cases system with a with age with room with 8 or more based on reasonable 3 how to manage and now going question sinking requirements so actually you need relations between States is products customers and in many no secret documents force you can do some kind of a foreign key constraints and I would say that most often don't a joint so you just have to do to queries enjoined that on the client side so that is again code that you have to write yourself what hopefully there's library for that just library just hides the problem of the database and then be even worse sinking requirement is the recommendation already told you actually wants references to the real object in your graph database and if you want to very uncertain attributes also then you have to keep them redundancy in the graph database also so you just if someone searches for and to beasts and you would you would not of 1 to offer in a different to be for just what 1 but you would for example of the different DVD player or something and that should be part of the graph theory and you do not want to do 1 step in the graph database go back to the document always that a DVD player and go back to the graph direct database to suggest this so it's quite a lot of effort that you have to put in your application to get that
thing wrong but the good news is this alternative to the so called multimodal databases did you from 0 to 1 databases that accounts for different data formats data so it can store documents it can for graphs and it can store values and the idea of the graphs that it can stories that you simply use your documents from the documents for and connect them in the graph and the edges to the outside also appears documents so can query the edges as documents which gives you a different opportunities and you can queries documents messages still but you can use the same documents in a graph you don't have to write any application code to combine this and to keep
them redundantly because that's in the same system in the same database and it
offers you all the current mechanisms for data models so you get the current affairs for graphs you get current events for documents and get current affairs for QA pairs let's go back to the use case
so it's the same set up and we still want to use document stores graphs for and based on but because just use a multimodal database for all of them and this leads me to render so my
favorite features of rain would the 1st thing is that some which monitor-based so can store graphs documents and you values next thing is there's a clear language it is called a child and it is as powerful as SQL would say and it is offering joints and reversals that means that you can use this this language to premier graph database
inside and also get joints on the document level which is not offered by many documents costs and also across multiple collection and next thing this parameter to be is completely effort including water collection transactions so you can actually store the documents in different collections update them in 1 Assistance Act ends but also you can count wrote for features and this this 3 the
last 1 is the feature that you don't expect from database Fox is our JavaScript framework edges like comparable to no chance has some brought the built-in libraries that has a jobs for prime-time which is with the and you can use it to put JavaScript code on your database level and to adapt the API of the database to which actions need so if there's something missing and you need that you can just wrote right own API and JavaScript
that is and into the database entry X on the raw data objects so whenever you have something that you would put into your client analyzes and very again the database from the result without user interaction or something that you could put that into the database level which is much faster because you don't have to transform it from a to B can just execute on on the database level and put the complete result outside yeah this is the question I'm not sure that he had to yes and it is and the idea is that you that you make maps so you can put steps into and version control and you can then install them from your version control system into a random so it's not as good news 5 compared with the sort procedure but you can manage your versioning itself and you have it in your and version control system you can just install it and then I will have a slide on that later OK so how does it look like no AQL
that very language and on top of a document so in this country I would just get the users which actors and then I had to read over the gates and there are performing joint because I just get the scale of the games where the player has the same idea of music and this is how I could you join them and then I can return the user's name and for example the 2nd thing is a can to modify the documents again I iterate over all elements using this futile but if it is of course applied before you iterate over the documents because it fits in the index and we use the index fast and and this is very I would for example the migration so I get wherever the status has the old format not not active I replace it was active faults which is the new status in of and and the last thing is abrupt proposal which is a function and we're just say I have a graph which is called underground plant and in underground plan I have several document collections the next lectures so the abstractions connect all these documents and industry I want to start at the main station just want to go in the up on direction and check which is reachable within 2 of 5 stations which is for example important if you're driving
with the subway and you buy a ticket which allows it to be right at most 4 stations can check we can go next thing as a transaction so everything you do in H U this
rejection by default but if you
want to do a bit more if you want to write your own code to some logic that you cannot express nature so you can just right JavaScript function which is put in in the action you can get a lot on the collections that you need so in this query would uh right and uses of products and would only read on recommendations so whenever I start this rejects and and this transaction I get whatever stored and at that moment then I do everything that is stored in action as a JavaScript function if that works through everything will be stored and it's possibly inside the database and if there is at some point a for whatever reason so the internal function fails then it will automatically do World and I don't have to care for anything that has broken my database good this head of In not really we have
Motivation concurrency control so you just get the only thing that you really need them and you try to get but you don't see newer versions of these objects know let's talk about Fox the JavaScript framework that being included so the
idea of folks that you can encapsulate microstate in micro-services right such a database of them by an API and what you can do for example this aggregate data and enforcing privacy open security constraints so for example you can cut out some data that you would not want to deliver them to use all you can for example apply some and decryption encryption functionality and at some further information from other carriers or whatever directly on the database level and then just put up the complete results to the user good and then you could also use it to trigger logic on data modification so whenever you want to spoil something we just go over Fox wrote and it adds some further information so that for
example encrypting the user password or whatever and then sells it to the database and and beauty of the Fox that you write in small so you put them in your own version control and you can be quite them from other folks at so for example if you have a user management to 1 you write this 1 once and required from different applications that direct and you always get the features of this 1 user module this and other things you can do with that is to improve your Monte device set up so nowadays you just
have like iPhones or
whatever tablets or PCs and and if you for example the moment set up then you typically do not want this whole load of data pushed to the your client if an office set up you maybe want that and so you can actually decide on this level already but I have to crunch something down right that 1 time in my application and then just ship-to-shore results and they do not have to write the same thing on my I fall on my Android and on my wherever uh wanted them ever devised to just conscious that all trigger different groups you can just use the same route and write it once and for all in a database so you
get the correct data format for the best then then you can also secure your data and Fox so you can just use and further the authentication algorithms or whatever authentication you need so we have built in also to all HTTP basic out then 1 benefit you get a you have low network traffic and directed Texas so you just push out the the data piece that you actually need in your application you cannot push out anymore and have to create an artificial then and that's why you can add your own customized inversion REST API and jobs strips want to arrange the which is then the API that you would actually expect from the data for which you need from data OK so that now let's go back to the benefits and overhead of the Polyglot Persistence said so benefits
native mapping into enough data
into of data into your database it's still given because you can store documents against all graphs accounts for values native thing of databases that optimized for the formants in the and you get trace for all of them around but on the other
side we can have a close enough so data has to be stored redundantly and has to be stored there has to be kept in sync and actually we don't need that anymore because that will be done in the database of so you don't have to and
false that it is in the link or something because you reuse the same object don't have to copy them itself
several technologies involved while also true use 1 multimodal database and that's it and therefore administration for the future it's also not true anymore so we could get for this part of the viewer into keeping the benefits but all these so I had told you it's not valid for multimodal data anymore yes and that depends
on that only a set of so we have replication algorithm if you want to replicate it for for yes 1 yes and so the the idea is that a graph Korean can access documents and edges to the onset of P also documents so you actually have in your batch have pointed to the starting document and the target document and you do not have additional documents stored in the if you have a different graph database you need to store these vertices in addition but now we can use the document security and that's the benefits so there's actually no redundancy unique for that end yeah then given the chance to give you an overview of some more features that the author was surrender to be particular so the complete project is open-source and freely was Apache 2 license you can just download it check it out from the top of yourself and just try it and we want to console follows for all the operating systems including Windows Linux Mac OS and a single set respite version if you like it and then the building charting of
replication so replication for backup strategy and shouting if you have more data than it would fit on 1 server of then we have grounds for a wide range of languages including Java Ruby JavaScript cause and Tyson and Ph.D. and then we have regret front which is shipped and was rendered and you don't have to pay extra for that just get it free and then we have documentation which is mates with spitballs which I think is quite nice now and and therefore you professional those community support for the database the
and well if you like it you can be invited to join our growing
community by using a rendered
your contributing to it if you have further questions just contact me at and all these green fields as any spy contribution coming from from all over the world entering so we're quite low enough OK thank you very much if
you have further questions I will be around today days In the JavaScript track which I'm organizing was sort of and all of a video so a person without this is phase of the the the basis of this yes so the acid is you can enable it and you can do that on a book in and collection level and on the current level so if you wanna say this 1 theory has to be superfast but I don't care if it's possible not to save that for the current and if you have like financial data or something just way that it's good but then it it takes of course a bit longer but in typical use case you actually for as the 1st a connection weights because they are not time critical and 4 times crisis you don't wanna wait so that is you can configure that this please you yeah yeah but you can and so if you have a rate of something that you could just you surrender to be all the states and set up a different database for each of them then you have the same setup as before OK which does not really help you but it's OK you can do that and things but the other thing is that we have this building chanting and replication so if you use this set up to use 1 large database in some cloud environment or chartered environment or something and then it appears like 1 digit large database project should get this there again the benefit that you have different service for new data and you could just throw away that 1 server and replace it by the other 1 so yeah what 1 yes so I'm not sure if I really get got point and that you can use the same at the idea of these is micro-services that you could apply them on all the machines so if you throw away 1 of them it would still be available on the other ones and you just have to make a backup of the data and then you still have the same service available on machines and then it's actually a transparent to which 1 of these machines you connect but it would be false if you click the right 1 because then you don't have enough for trafficking and so our target is is also all the Service into in the same dataset so if you are in the same data center you can use all the machines that are there and actually the targets we would suggest would be like for a change something in that that region but you could do more so there is no parliament and using like 200 machines but if you're using different data centers across the the words and then you always have this network latency but have to take into account for various that actually use more than 1 server consider but you could do that but you have to live with that what you the shots new data such that everything that would be requested from Europe states in Europe and all the things that are requested from the US are in the US and then you could connect them in the in the global scale have some more advanced features if you really have some queries that are across these countries then you could do that yeah for example then he he could still get the the data from from yes in Europe or whatever this is was yeah and and now I think not yet but that especially on our robust but I'm pretty sure we don't have the book right we don't have OK yeah my the the idea for and future you catch them right now and there is nothing that published but we're working on recipes I will publish them in the future and that is 1 of the use cases that we consider in this red uh recipes and then it's the yes you is the so so so my opinion is and that I don't think it would get as fast as single sequence databases are because they are built from scratch to support these features and so for a relational database you just build something on top of that simulates the most equally diverse that's what about this is a half half must admit and never trust these benchmarks because most of them are really all we have all yeah yes it's right or left so the on yes right yeah yeah right on yeah but I think yes yeah I think both parties really learn from each other and try to get to the 1 solution and that's it so it is both of them are approaches but I must say if you really have no secret data than the best using no secret database and not something that simulated that's my opinion and thank you very much mostly


  671 ms - page object


AV-Portal 3.9.1 (0da88e96ae8dbbf323d1005dc12c7aa41dfc5a31)