Merken

A Solution to the Backup Inconvenience

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
right good morning dimensionality right on time so let's get started Alex now I want to introduce you to 1 of our smaller projects with colored arrested and then a well catch intact line at that enough to to drag all in here so this morning when we start to talk about the project and Alex is that be the developers so he will then drive you through that the tests and the light demonstration will talk a little bit about the project background away and we feel that we need another backup solution and what are the design principles along which we develop breast
and some of the features and the components and what we do differently compared to other solutions and mentioned Alex will go through a quick demonstration also explain the technical background and chunking in cryptography and then uh obviously we'll talk a little bit about project and alter how you can contribute a we have a new contribution as well and also we'll leave little room for some
questions and answers so why do we feel that we need another backup solution and monarchs and I we have some mutual infrastructure that we use and we have also thought about OK back of might be a good thing to do at least at 1 point in time so we share the backup server in a location that would help the only partially secure and then this securities OK we wanted to do cryptography but the service hosted in a location will tell you where where someone else might temple was the system or might even get access to the system so without that we would like to have a proper that solution so we tested the multitude of backup programs also tried like using an ANC FS over ssh Avestan and other things and they obviously they didn't work so well and this is where we started
the project and use case was nothing special so we didn't want to back up like a huge amount of pictures or data I was really like the usual working directory in this example it's 16 index of files 140 K 5 and 36 a cave directories which is some source source code images audio file at some pictures so really nothing special and the solutions that we tried out were most recent work flow and particularly over the bond connections the latency was an issue that we came across and and some of the programs especially the older ones that they don't have the a cryptography them or notification and that mostly and not any of the 2 and then the usability was also somewhat cumbersome with some of the programs that we tried so we had configuration files or they they had to read chunks of the data on the Crab did before you can append on new back up so these were the things that we came across and and that's why we started off with a discussing about the design principles of what we feel good backup solution and it should look
so um it should be easy so if it's not easy you you wouldn't do it most likely and I'm wondering who of you would say wholeheartedly that you have a current backup of all of data and then we also wanted to be a scriptable so you can run it when you log off and it should be a pretty fast as well so the only limitations should be like external factors network the true good through your hardware such and it shouldn't be limited them with the software itself it should be verifiable so we wanted to check the integrity of the the data at the back of the destination and that without decrypting
it so Alex for plane a spend little bit about and how we achieve that want to avoid for example be trucking would be drudging on the back location so you can only you can easily do it shot to some on the back up the destination you can verify and whether you're integrity is the error of someone only has but with the data and secure so as dimension initially we thought the back of location might not be trustworthy so that was 1 of the design principles so we assume that the back of the destination is never trustworthy another user might have improve on the system for and you have just physical access to the to the best software and we also want to thank at the back of and restore so when you read the back of it should be verified if the backup was modified and efficient and so in some of the backup solutions you would
need to take a decision whether you take a full bag of or increment and that this is something that we don't want to do are you do like regular full ups to not pile-up in increments when he eventually restore it so that's also something that we looked into and we do indeed application only on file-based but on chunks of data so and that's also that some of the solutions did lack yeah of course it's free software that's why we're here and we use happen have pretty transparent development
process and we haven't even internally uses a smaller development well so we're review and 1 of our fellow developers is going before committing it to the repository so it is a quick glance at the
features so we we support Linux them 0 6 on dominant on the eastern variant and windows this yet we do do deduplication has just mentioned the request for full mode and that was also 1 of the design principles is stable and extendable so it stays and so we can extend and without losing their compatibility with prior worse and so there was also 1 of the design called it's already pretty fast so our use case today that we will show user use these 3 and half drive that we have with us knowledge will show you what this true is still a little later and you can use 1 back of repository for multiple machine back up and you can when you have a different sources and eventually use this for restoring 2 different machines and later point in time or at the at the at the same data on the machines can back up to the same repository and that the deduplication is also working there we have pluggable back and so we feel that this is very important so you can use of local storage as if p is built in we have a AWS necessary support and some more to come and restoring backups can be a little cumbersome so and take a tower archive and you only want to restore 1 file so can be pretty lengthy process when you just want to
restore that 1 file so we thought we want to have a browsable back up so we have a few support knowledge loss show and how that world so you can browse at the back of and just copy that 1 file out of the back up to them you you restore destination and also small embedded web server is about to be committed so you can then browse and all to your backups with and this small web server so some of the key
components I was explaining the function so it's the functions that all the backup solutions might have so they are however is taking care of the the back up and restore obviously of the the the release of a few small then the check of the check the integrity cheque component and that can verify the back of his systems they use the the repository and the repository makes use of the chunk and I'll I'll explore expanded how we do chunking or unintelligible a content defined chanting here and the repository will also make use of the Cryptography before it writes it to 1 of the placable back and what's very important is the interface for the back and forth relatively simple so a new back can be integrated fairly and easily so them I think that's already the point when you wanted to start off with a quick demonstration Alexa late taking the limit
on the so yeah I and we just have 1 microphone so we have to switch it and you have to switch back again but I don't think that is a huge problem and so the rest implementation at the moment is written in go and that's a fairly new language developed by
Google to statically compiled language and it has a fairly good integrated testing support for example and you my original choice to goal was just to tempting to to play around with gold and so far it was a very good decision for that but for the topography we used a anti-police 1305 signature make all rhythm of that's also quite fast and many of the ideas in the resting come from things like it or candy store of candy stores a goal implementation of which the data sharing and data backup system and then the also that of ideas from from gets 1 of the ideas was that we address the things by yet we have content-addressable storage like the
address for storing something like a chunk of data is we got something like a the shot to some of the data itself we will see that in a moment for the metadata something like directory structures violence and so on we used a of that's I think not is not an obvious choice for a backup system that we found that it was fast enough and adjacent is really extendable so if you have new fields that out and present in later versions of the repository you can also read it with an older version of resting because the future just ignore 1st extendable on for the key derivation function at the the moment only backups they are secured by password but you can't really use a password as a direct input for the AES encryption function needed key derivation function and for that you need a script to be as secure as possible so 1 of the better choices for regularization function but we also do we do come to define tracking will come to that in a bit and that's done by Robin fingerprinting and road implementation go for forever fingerprinting and it's already really fast although it's own wouldn't go it's looking at the data it's written in assembler or C yet but that's an optimization that's quite possible but so for the court to define tracking for example have filed and and you would like to to back a bit and what
touristic does it's computes fingerprints of rapid fingerprint of 64 bit over a sliding window of 64 bytes so for each window of 64 bytes computes the fingerprints and in the fingerprint below you can see that it is always different for each for each of 64 we knew of bytes
at some point in time it will come it will compute a fingerprint that has the lowest 20 bits set to 0 and this is 1 of the this is the point where I've been defined at the current Chinese and that's the new China is about to start so
for example in this case the opera file already has its first-rank and yet the 1st trial already found and then it's computes all the other chunks and you can see that in the in the picture the the orange the light orange boxes in there is still 64 applied before the end of the trunk that defines the trunk point and in this case we have 4 contracts for this for this file and if you
modify the 5 for example you change something in the 1st chunk light there's an x and you make the wire or something like that then radical will detect that the 1st chunk has changed but the address of the end of the 1st time hasn't changed all the other chunks of the same again from the same as if you modify it for example the 2nd chunk and make sure the 1 but don't change the 64 bytes before the end of the chunk rested will attack the same point and all the other genes at the end of the file the same again so you can move tanks around and you can modified Charles and all the others will stay the same so that application read texts very well if you if you move why stuff in the final round something that that
so the basic repository structure like an insect example it's always the local storage in a directory but basically all the other begins on model like this you have a file called conflict and everything on the slides that is painted in with with this slide red background is encrypted and everything that's not only with a white background is in plain text so you have a config file that is interpreted so I'm and you have a number of key files since in the directory keys and then you can already see there that's the file names on the list of the 1st 3 bytes in this example the 1st 3 bytes of the shot to check some of the contents so but in reality it's longer
but on the slides I've shortened it to just the 1st 3 bytes or 6 characters so each repository has a number of key files so key file looks like this it has a day some structure in plain text there is some metadata on it and it's has a solvent and data and um usually I entered my password interested and arrested tries to decrease into rolled all key files and tries to decrease the data that this in the key file with the key that is derived ViaScribe from the past work and what's in the data basically is the master key for the repositories diner something like looks if you know a Linux file encryption and they already had also have a master key for a volume and we have a header with key slots and and apostrophes Keach key slot so once I've entered the correct password resting is able to be trip the config file in this case when trip that always meant the crypt and authenticate the cyphertext before trying to exclude the something to from attackers about tamper with Mexican finally exchanged and for something I can detected before I even try to interested so once I have the master password then I can for example be cooked 1 of the snapshots of snapshots is on meant something like
the structure file system structure at 1 point in time so each time you take back up you get a new snapshot will see that in a in a moment in this case the snapshot after decryption user based on structure and in their they have for example the path that has been back up in this case this whole after 0 well that's my my 1 of my work directories you can see it wasn't made of the host name Kuz'nia and that's what my user and so on and a the point in time at which it was taken there are some other fuses and this basically references to treat the trees and and then shot to the Hessian again uh 8 6 F and so on and then from the repository you can build this tree identified by the shot to some of the contents of the adjacent structure and in there you have a list of nodes in this case you can see that there is one directory it's called web and 1 file it's called firelock target is to and then for the for the web directory you have your another and In for a subtree 6 2 6 9 and so on so if you were to restore this web directory you need to fetch this subtree in the next step if you want to restore this file and you have like in this case 3 different uh shot to sums of chance and this is just the shock to some of the raw content trunk that has been there in this in this order so the file can be restored by just fetching these 3 Chinese from repository writing them to file in this order and then you will have the file again so I mean 1
thing I haven't told you up to now is that I don't save on our resting state doesn't say if any of you metadata and chunk to a file directly but attacks on bundles them together in a In in several you know several times a bonded together in fact file and this looks like this all the things in the in the in the text file there's tree this data data data and at the end there's a header and they had understood lists all the things that are in the file and the text file itself is an encrypted but transparent are interrupted so all by themselves so I can be all of them a structure them without having to re-encrypt everything in order to speed up the refining process because otherwise if I don't have like an index of something then drastic needs to fetch all the headers of all the text files to find where are the trees and where the data stored them in this case I have an index that's just not optimization and you don't really need but it's needed to to be efficient in the past and it lists like in in the Datapac 7 B and so on there are these blocks and the love type tree ofType data and if you have a look at the tree again on the right side you can see that the subtrees 6 2 is based on the slide but the content blocks 3 at 6 3 and D D they although in in this fact file so if you need to fetch that's been like story so this was a lot of theory
and then I would like to do some some demonstration for that I'm going to sit
down now and so let me
see OK can you see that OK great so the 1st thing I have attached this use the storage device here and and so on also note who sleeps find amount so it's just a USB devices connected all use be free and the 1st thing I need to do with the rest rest accusing it initializes repository that's really easy to do you have the stress a command already in solid for example and then you can do minus our all repository use B and then you do it and then you need to intervene to enter
profit that the similar to looks like looks form of of then you have this repositories has an idea that is saved the content just used for recognizing which repositories have been already
backed up on and then you can do a backup for for example of my directory data and yeah you just to give you a few hints that data is at the moment with we have made of something that's resembles a normal working directory it's 2 gigabytes and there are files and yields and the freezing everything in it and if you can't images and deviance resource and if we have if you
would like to back up this we need to give the rest at the back of location the speed and then back up and then just directory there should be backed up and offered and as you can see there it's quite fast like 17 megabytes per 2nd and this along with all speed if you have like AES and ideas and extensions for in the CPU that's a speedup AS computation and will take about like 10 seconds FIL occasional specialists takes longer so in this case is sensitive it's not we can do another back up again just to be sure and then you can see that it already knows all the structures it knows the files and modification times and it doesn't have to back up anything else so not the yeah like 3 backups and the command
snapshots what snapshots it's that I can list what was back ups and did I made in what directories have been backed off and so on but in order to not have to type this repository location again we can also exported in an environment valuable and they can just use this when you can also export the password but I will try that for for the effect the OK so what if I'm about to to do something with my data and these guys I have this for example are I have a large Vivian ISO the and let's let's say I have another
version of this ISO in the inner core faculty to and you can
see that it takes up the twice space because it's not the defecating file system but if I'm about to to back up this directory again
so we don't need that anymore to you can see this is still quite fast uh this 800 megabyte per 2nd something that's not really but using the already known chunks when the new ISAF I was right and to recognize that these trends have already been saved so they are not transmitted and not safe repository again and we can for example check that what's the the size it's all over yeah that 2 . 1 gigabytes and this is just the size of the data directory costs of even that even the larger because the 2nd that I put them OK so if I'm about to restore something
disturbed we all hope that
this point in time never comes that sometimes you have to restore things
from the back up and you can for example have a look at which snapshots are there and then you can see that for example this this last back up I would like to have a look at it that we can use L S with
so there you can see that these are all the files that indicates not wide enough so I'll just use
this yeah you can see the directory structure and for example there's this this last finds some of final my of my notes here and if I would like to restore that we can use rest think
restore to can restore and I can use either this complete back up or I can for example like include start and only and the files are restored OK and will have a look at and restore or they can see their state art and there are my notes OK
and this is quite cumbersome I think because you and I as a developer aphoristic need to remember all the command line switches and so on so we also added the possibility to mount this repository by a few and I will mounted in my home directory and file in a directory called an anti and I'm creating new tab a handful have just look at the image the thing that's snapshots and there are directly useful for all the snapshots and using the last 1 vacancy that's the stator what
part what part of Indian
catch and it will work and therefore demonstration purposes we also included largely due from the end project In the videos directories it's
quite large so that we can even play
that 1 however this we but it's
got I have no idea whether whether the when the wind father thing and see that
it's it's a for which the thing
and it's it's a place really really fast so that he's out of his
money from the USB drive 3
devices and you can also like
browser pictures but he's
pictures and have taken some
of it is a little and I can
even browse the older versions of files there like these are online holiday
pictures and you can view
pictures and so on but I'm about to look at what has changed
remember that added the who follows which was a copy of the dead in ISO file in the 1 of the latest snapshots we can for example look at the first one and have like data deviance ISO and their system 1 didn't ISO file and the from looking at the later 1 yeah I can see that there is also the from files and I can mount that a copy that way or have a look at that and browse the files used if for example but also possible OK afterwards don't
forget to my unbounded OK this
was a big demonstration
and the current status is that we have recently was the 1st version of a repository for months quite stable and we would like to have users using it now also have written in trying to make should using semantic versioning and then making sure that the positive form of always compatible so you when you do some backups now here and you you are able to restore them and access them many years and that we need to implement more and slightly cloud-storage nearline and inflation and so on and we'd like to use to user interface improvements for example the restore workflows too complicated at the moment of maybe a dialogue-based system for that we would like to add compression we would like to resume interrupted back absent yeah of course more performance of maybe you support the workflow to to do metric crypto on on the back up so you can you can have a server with an important data on it and so do some backups for the server and with win service had the attackers don't have access to all the all the back backups a unable to decrypt them because the private key used to decrypt data is not a but and other workflows like reading a backup from standard in think my square them for example I'm
getting started is really easy you can kill the repository of G. into the repository and just type go around that go on it will take care of all the usual stuff you have to figure out why well using our compiling a go program and then you can use in it and back up and that's it at the moment it so that it's
free software will stay free suffer of course it's it's clause licensing confined the repository their army using some some kind of like an agile workflow as far as you can talk about agile when developing open-source project with the redistributed developer base and at the moment we have 3 contributors for that you want velopment and 12 other contributors and some even present here thanks a lot for your contributions and you can help by you stress report box and we're sure that there are parts hidden in there we have and we really should be grateful for all the reports that point boxed up just because then we can fix them and also of course also request features that we might you may have seen up to now OK give
any questions yeah think so we have to
and like names like you you at the moment at the moment is not possible but there's an issue about adding like text to 2 snapshots and that's really easy to implement because you can you can also adjust the text fields of adjacent structure and you can do something like that well that yes they gave him as you yes and the he the yeah we can we can about the the Texas USA something like this adding a string to a backup while my making the back and we can we can implement that if you if you are interested in that which we take that as a plus 1 to implement this feature in might be it might become a little bit is hard to find a specific back up when you have a lot of back ups and so why is this a good idea yeah but and the question was it if it is possible to do automatic backups and like something like that and at the moment we only have the possibility to excessive rapport with the password so you can export the password in an environment variable and have residues that he's as a metric workflow that you're talking about like using a public key to encrypt data that cannot be decoded again with this public key but require the private key for decryption and I think that's what you're talking about is not implemented at the moment but its planned a case but if this is when you want to go back to the room and you know and this is where progress all the way to more work all of this work from that because of the the question was what would suffer requirements are needed for for a remote locations and at the moment we have several back and start stone requiring and installing anything else and so for example we have S FTP from implemented and you really only uses the F S FTP protocol so you just have to be able to use the S FTP clients to connect to a server and have them something like open as a running and providing server that's enough but it's not the most efficient way to do it because as a T P was designed in a time where people thought it was a good idea to hot cold like buffer size and so on so we have to in order to transfer a large amount of data we have to transfer really small office so it's not the most efficient back up the back and but it's possible it doesn't need any installing anything else the question yeah it would and you get the when the the moon uh so and your question was what is it possible to do this this in the hope that have of yeah is the how look at the moment it's it's not possible but you have to build something around it like for for example in LEM snapshot of a currently in use file system so you have a specific snapshot of the file system at some point in time I for Windows there's this shadow copy service provides the the same and it's not implemented rest moment but it's certainly possible to either do that the ideas around the backup program the around the backup process for example for the EM images of or even Canada invested but is not included at the moment on and when I was encryptor directory in your case so you have and on the yeah resting at the moment prosecute just use the normal cone interface to like Mr. directory and open files and then in although the company you yeah you is that you can see the back up the the intricate directory from the outside so that's the input the directory is saved in the back up and when you have when you restore then you can get you also restore the encrypted directory and need the user to enter a password for example decrepit again and or you can have a user like unlocking his home directory and encrypted call back capping the plain text files from there is no integrated like the corrupting something like an intricate home directory at the moment all yeah this 1 1 possibility if you were involved in the OK there was you I'm at the moment the repository in order to complete the project just use a single compiler as a dependency we use some some external libraries but they all are included in the repository as in in the vendor directory so we can have like a deterministic building for something if you if I have using this version of the go compiler under Linux and somebody else is the same Gold compiler under Linux the same binary then will result in the same binary and having the dependencies vendor to in the archetype report also means that I'm able to like use get by 2nd all my dependencies that nice this means that the in the in this yeah this we are at the moment we are trying to build this for packages it's a slow process because you have to do for all the different distros at the moment there are packages like for archlinux and homebrew unlike and it will probably at and others so it's just that we have reduced the the 1st release in just like a few days ago and think we need to do some some more tests on her back ups and all the time it will inserted into the distributions all the special files already have a look at the start of the hands of the sky endowment so something's our back up as what they so and we can even go back up like the device files character devices and so on and so forth coupling so it's the implementation is and isn't finished at the of this we have this question yes and the question was if we have implemented a way to remove all backups I've heard that other backup systems like this possibility at the moment and at the moment it's not finished but it's would you already started a process that can and have shown you that there is for each snapshot there's a file in the repository you can remove that and run the check process which we detect that there are dangling blocks that aren't used anymore and in the future it will be able to remove them and repack the others so that I may be used in the same so there is a pull request outstanding and I'm working on it at the moment that's is but the question was if we already taken care of excellent attributes and security at this at in units for example I'm at the moment the answer unfortunately is no because we've just started but it's it's an issue and in the detector went yeah planning to implement that unfortunately at this from my point of view the ACL handling isn't standardized across different operating
systems so we have to do an implementation for like each operating system so it will take some time but it is planned to also back this expression that this is the overall they also so intricate and everything is interested except these things that you need to derive the master key password like everything except the Key files it is a start encrypted part of the but not they don't know anything in the Key files you can you can think of something like it looks the slot that's basically just holds the timestamp and sold and the encrypted data for the master that's it everything else is encrypted the of the world I want know the world I and the question was if you have a structure was subdirectories and yes we have destructive subdirectories to account for like having a repository of many files so there's 1 subdirectory level for the data files what yet at the moment all it's and no because when we have and the question was if we start to use all the different tastes files and in in in in in a lot of small files and nowhere bundling them together up to like 20 K Jason structure files if they're really small in 1 file so we have just bit bundling the files together so we don't have so many so many small parts of and you know it is going to be an he said that if I had well I the question was if we need to implement lecturing data laws that require distracting data all removing removing data from a back up and at the moment is not implemented but we have an issue about that exactly about the this thing like all of by accident I committed the password 5 of the sequence data into the back up and I need to remove that from every division and it is possible but we have to rewrite the data and there is 1 the 1 of the the the basic attributes of the repository structures that things don't they are just written once so if you have to like or change something you read it and you write a new version again and then remove the old version and need to correct all the references in adjacent files that it's not impossible but at the moment not implement at this the following and good question we don't have a plan for that right now that's completely ready we also have a really have a discussion in the in the issue describing this as a metric use case because we need to to understand what is the exact use case for it and it's not so easy because there might be many different use cases for example this I'm using resting to back up a server and afterwards we store it and if the service tax the backup system of the data in the back of the old data especially should not be accessible that's just 1 use case 1 of the ideas was that may be different keys are used to encrypt like data and metadata and so for example let's say when the when the system is doing because rest again is just has the private key for decrypting the metadata again in order to do efficient back up but it doesn't have the private key for the data on that but this depends on the use case so maybe you would also like to have the metadata encrypted only 1 way and you have like local cache for example we need to discuss that and afterwards plan would be doing but we're going to it is a large part of the the of the models along with all the people all you have good question question was and we advertise that it was possible to and verify that there was no 1 will be dropped or the backup data on the server hasn't changed and electricity basically does is everything that is saved on the server and remote location is the file name is always the shot to some of the contents of the encrypted content still on the server you can just run like shot 256 some on the files with the script doesn't make that read the files output shot to hash of it and you can can compare it to the to the file name even without decrypting anything that doesn't protect against attackers that are able to write the repository because they can like modified some filed there and this to a new file name that's we corresponds again and but you can we check for example if the hottest is faulty all the Rams faulty into during right if something went wrong you that and so so and understand the question in the so this is so of that I think question was that it is possible to use public or private cloud implementations store the backups and at the moment is for example possible to use S 3 bonds to store data the beckons implemented but some other back and I'm not implemented at moment for example for for cloud or electrical drive or something like that but I we try to make the interface for implementing a backup really small and really easy to implement so it should be easy to add that and for example for for the cloud or for as here there already a go libraries that do the interface for you you just have to adapt it to be used as the back investing by the time in which the all on all the question was what what happens if I back up something on Linux and restore its own and of asphalt system on windows and the general policy for for these kinds of things that rested block a warning that is not possible to like extract assembling and in the case of assembling it it it might be possible to just add a a symbolic link or something like that in in in the interface false system but again policy too often error about it is not possible to restore this this type of file and this file system and continue to restore things the you the yeah that's that's no problem at all the question was is and if it is possible to back up the system in a linear system like likely byte order is starting with no almost bite and restorative began system and that's no problem at all because on the data side which is moving moving bytes from the files to the backup repository and back again and for all the for everything else that it's metadata that's either J. Sun like ASCII encoded so numbers and that's doesn't depend on the byte order and there's 1 exception for the text files are we have a little yeah small-format for for the for bombing the files together and 1 of the few theories the lengths of the trunk and this is defined as to be so that's that's no problem it that change the structure of the but the question whether you can change the encryption key on the question that to you would encryption key you mean the master key password at the end of the month of the year 1 of the things that people use the ViaScribe virtually and this is the smallest of the following well known mine we assume that it was in the you so so the question was what happens if somebody gets my password or is is is possible to to change my password for the back-up remotes back at some point in time like every half year or so and the the answer is that the master key is always but we have for
repository and you can easily change that but we have this 1 level of indirection by using the password that user supplies to derive a key to decrypt the master master so the design is really similar to 2 loops again we have a password and you can use that to decrease the master key and the master key decrease the price of the device and it's possible to have multiple passwords and it's possible to change the possibility that if an attacker has access to the back application the the attacker is able to just make it back out of the old key file with the same for looks if we if have an attacker was copied my my complete device including the and afterwards learns my password he can use that with the with the that he has already saved so changing the password in this case it doesn't give you anything but you can do it this sentiment
expression if this is a
summary of the news media how long are you going to college the question was how do we initialize the counters for AES encryption because a counter mode encryption when use that you have to make sure that you use really random encounters in this case it's generated so random confidence values generated so we have like we generate an IV 16 bytes of of random with the group of reference a cryptographically secure pseudorandom number generator and we use stats and flip the bits and increment them for condom over it's a standard practice for that so we have this really really small chance that 2 intonation vectors will be completely different for 16 bytes of random but it's really unlike what you've done be on the children of this is very the rest of us the top of the other part of the world that's all you review because they would lose slower the that of was the question was whether the migration strategies from if i have a remote location where I've already packed up my all my data to hide to back up this again in order to use rustic and my recommendation was that would be too if the remote location is somewhat trustworthy so you trusted to store your data for a short amount of time you can 1 extract you data there but it depends on the form of what you have learned to could structured data there undomestic ones with all the data there remove it again and then use domestic from the from your new location because then you haven't repository that is initialized with all data already all the terms are there only 30 structures their and radical just just use that for the d'application so you always do not only transmitting the you know the data that is new and hasn't been stored of the complication all right we all know it especially the data so if a data type this in the repository of always use that 1 in the repository and not stored again the question was it is possible to to exclude files or directories and yes that's this possible we have a simple executes the command line switch for the back up uh already and there are plans to like and implementers and you more sophisticated 1 we you can have a file with with patterns or directories that I want to be back up or something like that it's yeah it's possible and it will be possible future to do more things it is used as the basis of the size of the of the of the of the of the all interesting use case the question was it if it's possible to exclude files based on the size like that in the back of small files what look that up especially large files I haven't thought about that up to now but maybe if if it's interesting that were implemented and it should be really easy to do some of these issues is the book was written in and then use those along the also what's really population is the part of the whole In the 1st of all the little so we so for example when you initialize it back up and then repeating the question for the for the others later but when you want to do an initial back out of your like you working directory would like to exclude all the large deviance ISO files because they don't need to to be better and is not possible but we will consider adding it we appreciate each of the that the k a questions in the but they also mentioned again because of the change and the so question was how to how can what can I make it's really fast back up on the local hard drive but replicate this repository to another location and as you already suggested you can just use are assumed to just transfer the files and hence for the updated files and it's even possible to just use the new 5 because resting always will access files and we do not change them so once they are in the repository they're not change but there are new files created and maybe files removed to remove for example so just moving the files to remote location that is sufficient you can just restored and from the remote location that's possible OK any other questions that this is the most all laws and the question was interested candles sparse files at the moment rest it doesn't handle sparsify specifically as paraphrase of fires that have like holes in them when you read them you will get the whole lunch and a bunch of zeros and but the deed application already is able to handle this this case as well because uh whenever it's a large amount of heroes then I will recognize that this is a block of zeros so I already have that 1 and don't save it to the back up again but we can we have a maybe will access sparse support by recognizing the special holds in the files and handle them special especially because we don't have to like read all those zeros if we know there are only zeros we can just add that to the metadata information you always of you know this optimal moment no support for restoring when you restore a sparse file you will end up with it a non sparse file with filled with zeros that's a limitation at the moment we will we will at that natural do you also have a question OK with all variance that are not of the people of the man and the mother of all of the functions of money good and the question was when whenever are really a lot of files and repository what it gets to know if 1 wants and the some 5 system and if there are something like 10 thousand files in a directory and for that we have in the in the data directory we already have 1 level of subdirectories and stultifies there so the the 1st by this actually used for creating a subdirectory and will probably add more if this becomes a problem in the future but at the moment is just 1 level and um and we can know the repository is also idea if they see it under the repository is also structured in a way that's more like snapshots and for example data files are in different directories because the snapshots and I will arrested must be able to like efficiently list them and read them in order to spread into the use of find latest snapshot the current back up is based on and so we moved that to a different directory and the data files and and another tree
it just so many people usually use long bottles do you have to use to model and this is the level of the people yeah yeah we could have also use like 2 bytes but at the moment it's it's 1 and this becomes a problem with the will that's all make configurable or something like that bad hand games have fled have what is that you use what the
question was if we needed to to specify some for Protestant yeah so if you need is specified system for the repository that holds the backup data and I don't think that we have a limitation like this but the limit you mentioned like 32 K files for 4 x 2 it might become a problem if you up to x 2 but yes please please give it give it a try if you run into this limitation right initiated would you wouldn't what what is the best not use like the year when the when the directory change that you are you backed up or a lot more than you that's what is the value the change of in the world I mean so what we have the the the question was what happens if a file is changed while the session is running for a a while this uh why this snapshot has finished for example when you back up a large directory sometimes might change in between for example in the world idioms on example was was moved to the file has been back up already but the backup process wasn't finished and the moving why the Beckhams running to a different directory that is going back up that may come later on in this case rested will include the new new file in the back of a 2nd time but it won't save any new data because of the the dedication because all the data in fact has already been saved to the repository it would just include like a reference to the state in the metadata again OK any other questions good thanks a lot for listening to give it a try and report all paths that you may find
that
Softwaretest
Offene Menge
Bit
Güte der Anpassung
Computeranimation
Freeware
Software
Rechter Winkel
Kryptologie
Vorlesung/Konferenz
Projektive Ebene
Zusammenhängender Graph
Softwareentwickler
Gerade
Server
Punkt
Extrempunkt
Audiodatei
Computeranimation
Eins
Metropolitan area network
Verzeichnisdienst
Kryptologie
Weitverkehrsnetz
Optimierung
Konfigurationsraum
Bildgebendes Verfahren
Einfach zusammenhängender Raum
Elektronische Publikation
Benutzerfreundlichkeit
Computersicherheit
Güte der Anpassung
Physikalisches System
Quellcode
Elektronische Publikation
Datenfluss
Dienst <Informatik>
Automatische Indexierung
Server
Projektive Ebene
URL
Datenfluss
Verzeichnisdienst
Ebene
Elektronische Publikation
Hausdorff-Dimension
Güte der Anpassung
Regulärer Graph
Extrempunkt
Teilbarkeit
Computeranimation
Integral
Metropolitan area network
Freeware
Funktion <Mathematik>
Software
Inverser Limes
URL
Informationssystem
Fehlermeldung
Normalvektor
Prozess <Physik>
Freeware
Dokumentenserver
Kartesische Koordinaten
Intranet
Extrempunkt
Computeranimation
Entscheidungstheorie
Metropolitan area network
Funktion <Mathematik>
Software
Vorlesung/Konferenz
Softwareentwickler
Informationssystem
Normalvektor
Basisvektor
Subtraktion
Einfügungsdämpfung
Prozess <Physik>
Punkt
Euler-Winkel
Hyperbelverfahren
Dokumentenserver
Datensicherung
Computeranimation
Metropolitan area network
Virtuelle Maschine
Multiplikation
Benutzerbeteiligung
Bildschirmfenster
Vorlesung/Konferenz
Turm <Mathematik>
Speicher <Informatik>
Formale Grammatik
ATM
Dokumentenserver
Stellenring
Quellcode
Elektronische Publikation
Packprogramm
Modallogik
Filetransferprotokoll
Server
Schlüsselverwaltung
Lineares Funktional
Punkt
Dokumentenserver
Momentenproblem
Formale Sprache
Implementierung
Physikalisches System
Computeranimation
Integral
Metropolitan area network
Kryptologie
Inverser Limes
Vorlesung/Konferenz
Zusammenhängender Graph
Hacker
Implementierung
Schnittstelle
Bit
Momentenproblem
Element <Mathematik>
Gemeinsamer Speicher
Minimierung
Adressraum
Formale Sprache
Versionsverwaltung
Implementierung
Derivation <Algebra>
Datensicherung
Computeranimation
Metropolitan area network
Metadaten
Weg <Topologie>
Regulärer Graph
Elektronischer Fingerabdruck
Skript <Programm>
Vorlesung/Konferenz
Passwort
Datenstruktur
Speicher <Informatik>
Auswahlaxiom
Implementierung
Softwaretest
Lineares Funktional
Dokumentenserver
Assembler
Speicher <Informatik>
Physikalisches System
Ein-Ausgabe
Elektronische Unterschrift
Entscheidungstheorie
Datenfeld
Chiffrierung
Schlüsselverwaltung
Verzeichnisdienst
Metropolitan area network
Bit
Punkt
Quader
Bildschirmfenster
Elektronischer Fingerabdruck
Computerunterstütztes Verfahren
Elektronische Publikation
Computeranimation
Design by Contract
Punkt
Dokumentenserver
Dokumentenserver
sinc-Funktion
Adressraum
Mathematisierung
Versionsverwaltung
Regulärer Ausdruck
Zahlenbereich
Vorzeichen <Mathematik>
Mailing-Liste
Unrundheit
Kartesische Koordinaten
Elektronische Publikation
Computeranimation
Rechenschieber
Metropolitan area network
Informationsmodellierung
Inhalt <Mathematik>
Speicher <Informatik>
Datenstruktur
Verzeichnisdienst
Konfigurationsraum
Schlüsselverwaltung
Gammafunktion
Dualitätstheorie
Gewichtete Summe
Punkt
Momentenproblem
Dokumentenserver
Chiffre
Zahlenbereich
Computeranimation
Netzwerktopologie
Metropolitan area network
Metadaten
Benutzerbeteiligung
Knotenmenge
Modul <Datentyp>
Dateiverwaltung
Vorlesung/Konferenz
Passwort
Spezifisches Volumen
Inhalt <Mathematik>
Datenstruktur
Konfigurationsraum
E-Mail
Dokumentenserver
Content <Internet>
Mailing-Liste
Elektronische Publikation
Rechenschieber
Chiffrierung
Ordnung <Mathematik>
Schlüsselverwaltung
Verzeichnisdienst
Tabusuche
Prozess <Physik>
Dokumentenserver
Minimierung
Mailing-Liste
p-Block
Elektronische Publikation
Physikalische Theorie
Computeranimation
Netzwerktopologie
Rechenschieber
Metropolitan area network
Metadaten
Diskrete-Elemente-Methode
Automatische Indexierung
Datentyp
Bildschirmsymbol
Inhalt <Mathematik>
Ordnung <Mathematik>
Datenstruktur
E-Mail
Faserbündel
Gammafunktion
Aggregatzustand
Konfiguration <Informatik>
Interrupt <Informatik>
Freeware
Dokumentenserver
Vorlesung/Konferenz
Speicher <Informatik>
Normalspannung
Ext-Funktor
Computeranimation
Elektronische Publikation
Konfiguration <Informatik>
Dokumentenserver
Momentenproblem
Dokumentenserver
Zwei
Gefrieren
Computerunterstütztes Verfahren
Zentraleinheit
Elektronische Publikation
Computeranimation
Physikalisches System
Bildschirmmaske
Verzeichnisdienst
Interrupt <Informatik>
Front-End <Software>
Passwort
URL
Inhalt <Mathematik>
Maßerweiterung
Normalvektor
Datenstruktur
Verzeichnisdienst
Ext-Funktor
Bildgebendes Verfahren
Soundverarbeitung
Dokumentenserver
Dokumentenserver
Fakultät <Mathematik>
Versionsverwaltung
Computeranimation
Videokonferenz
OISC
Verzeichnisdienst
Total <Mathematik>
Datentyp
Passwort
Speicherabzug
Passwort
URL
Ordnung <Mathematik>
Verzeichnisdienst
Ext-Funktor
Innerer Punkt
Videokonferenz
Physikalisches System
Mailing-Liste
Verzeichnisdienst
Elektronische Publikation
Twitter <Softwareplattform>
Dokumentenserver
Dokumentenserver
Total <Mathematik>
Dateiverwaltung
Passwort
Verzeichnisdienst
Raum-Zeit
Computeranimation
Schnittstelle
Punkt
Elektronische Publikation
Dokumentenserver
Singularität <Mathematik>
Oval
Elektronische Publikation
Bildschirmfenster
Computeranimation
Metropolitan area network
Physikalisches System
Mailing-Liste
Verzeichnisdienst
Total <Mathematik>
Vorlesung/Konferenz
Passwort
Datenstruktur
Versionsverwaltung
Verzeichnisdienst
Dokumentenserver
Dokumentenserver
Browser
Total <Mathematik>
Passwort
Softwareentwickler
Elektronische Publikation
Verzeichnisdienst
Datensicherung
Bildgebendes Verfahren
Computeranimation
Aggregatzustand
Offene Menge
Server
Dokumentenserver
Regulärer Ausdruck
Datensicherung
Computeranimation
Videokonferenz
Spezialrechner
Verzeichnisdienst
Total <Mathematik>
Demo <Programm>
Tropfen
Gerichtete Menge
Elektronische Publikation
Stichprobe
Vektorpotenzial
Codec
Videokonferenz
Quader
Filetransferprotokoll
Mereologie
Attributierte Grammatik
Projektive Ebene
Reelle Zahl
Verzeichnisdienst
Versionsverwaltung
Offene Menge
Gerichtete Menge
Elektronische Publikation
Browser
Versionsverwaltung
Stichprobe
Physikalisches System
Elektronische Publikation
Codec
Computeranimation
Videokonferenz
Verzeichnisdienst
Eigentliche Abbildung
Total <Mathematik>
Public-Key-Kryptosystem
Benutzeroberfläche
Euler-Winkel
Momentenproblem
Dokumentenserver
Seidel
Dokumentenserver
Kryptologie
Versionsverwaltung
Benutzeroberfläche
Dateiformat
Computeranimation
Physikalisches System
Bildschirmmaske
Dienst <Informatik>
Interrupt <Informatik>
Total <Mathematik>
Server
Passwort
Quellencodierung
Standardabweichung
Portscanner
Open Source
Explosion <Stochastik>
Punkt
Quader
Dokumentenserver
Momentenproblem
Software
Dokumentenserver
Klon <Mathematik>
Mereologie
Speicherabzug
Datenmanagement
Projektive Ebene
Softwareentwickler
Optimierung
Verkehrsinformation
Computeranimation
Distributionstheorie
Bit
Prozess <Physik>
Punkt
Momentenproblem
Compiler
Versionsverwaltung
Datensicherung
Computeranimation
Einheit <Mathematik>
Bildschirmfenster
Datenreplikation
Dateiverwaltung
Vorlesung/Konferenz
Schnittstelle
Differenzenrechnung
Softwaretest
Sichtenkonzept
Dokumentenserver
Computersicherheit
Gebäude <Mathematik>
Güte der Anpassung
Systemaufruf
p-Block
Störungstheorie
Ein-Ausgabe
Chiffrierung
Datenfeld
Server
Projektive Ebene
Ordnung <Mathematik>
Verzeichnisdienst
Zeichenkette
Public-Key-Kryptosystem
Filetransferprotokoll
Automatische Handlungsplanung
Implementierung
Puffer <Netzplantechnik>
Variable
Arithmetische Folge
Determiniertheit <Informatik>
Programmbibliothek
Abschattung
Passwort
Optimierung
Datenstruktur
Bildgebendes Verfahren
Drucksondierung
Attributierte Grammatik
Physikalisches System
Elektronische Publikation
Offene Menge
Residuum
Normalvektor
Verkehrsinformation
Punkt
Momentenproblem
Dokumentenserver
Versionsverwaltung
Kartesische Koordinaten
Gesetz <Physik>
Computeranimation
Übergang
Metropolitan area network
Metadaten
Arithmetischer Ausdruck
RPC
Bildschirmfenster
Skript <Programm>
Dateiverwaltung
Vorlesung/Konferenz
Zeitstempel
Funktion <Mathematik>
Schnittstelle
Multifunktion
Dicke
Dokumentenserver
Güte der Anpassung
Stellenring
Ausnahmebehandlung
p-Block
Linearisierung
Dienst <Informatik>
Rechter Winkel
Server
URL
Ordnung <Mathematik>
Schlüsselverwaltung
Ext-Funktor
Fehlermeldung
Public-Key-Kryptosystem
Subtraktion
Folge <Mathematik>
Fächer <Mathematik>
Automatische Handlungsplanung
Implementierung
Zahlenbereich
Physikalische Theorie
Division
Data Mining
Loop
Mailing-Liste
Multiplikation
Netzbetriebssystem
Datentyp
Programmbibliothek
Passwort
Passwort
Inhalt <Mathematik>
Datenstruktur
Attributierte Grammatik
Linienelement
Symboltabelle
Physikalisches System
Elektronische Publikation
Binder <Informatik>
Caching
Mereologie
Streuungsdiagramm
Bit
Momentenproblem
Mathematisierung
Gruppenkeim
Automatische Handlungsplanung
Implementierung
Kartesische Koordinaten
Gesetz <Physik>
Term
Computeranimation
Eins
Übergang
Netzwerktopologie
Metadaten
Arithmetischer Ausdruck
Bildschirmmaske
Bereichsschätzung
Migration <Informatik>
Datentyp
Mustersprache
Randomisierung
Inverser Limes
Vorlesung/Konferenz
Datenstruktur
Pseudozufallszahlen
Varianz
Metropolitan area network
ATM
Lineares Funktional
Dokumentenserver
Computersicherheit
Güte der Anpassung
Vektorraum
Schwach besetzte Matrix
p-Block
Elektronische Publikation
Chiffrierung
Basisvektor
Hypermedia
Mereologie
Strategisches Spiel
Information
URL
Ordnung <Mathematik>
Verzeichnisdienst
Prozess <Physik>
Momentenproblem
Dokumentenserver
Güte der Anpassung
Mathematisierung
Physikalisches System
Elektronische Publikation
Systemaufruf
Übergang
Metadaten
Informationsmodellierung
Spieltheorie
Rechter Winkel
Inverser Limes
Vorlesung/Konferenz
Verzeichnisdienst
Verkehrsinformation
Aggregatzustand
Offene Menge
Software
Freeware
Computeranimation

Metadaten

Formale Metadaten

Titel A Solution to the Backup Inconvenience
Serientitel FrOSCon 2015
Teil 45
Anzahl der Teile 80
Autor Neumann, Alexander
Daniel, Florian
Lizenz CC-Namensnennung 3.0 Unported:
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 10.5446/19545
Herausgeber Free and Open Source software Conference (FrOSCon) e.V.
Erscheinungsjahr 2015
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Secure and efficient backups can be cumbersome. "restic" is a new, easy-to-use, very fast backup application that does deduplication and crypto right. And it's free software! Alexander Neumann, Florian Daniel

Ähnliche Filme

Loading...