Merken

The Evolution of Storage on Linux

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
I guess I can get started then I'm going to give this talk in English citizens also being recorded and broadcast and we might have visitors were not capable of speaking German on welcome cross content it's been awhile right I'm very happy to give the 1st uh talk in the 1st of the 1st day of my name is nancy and I've been with column since almost the very beginning I think missed 2 of the conferences in total I'm always happy to be here and I'm I'm glad to talk about an a modest topic this time around I'm a little bit of background to involve myself and by the currently work for a company and I G and they're located in food from and the they work yeah consulting and and projects based on open source technologies also combined with proprietary technologies so be they claim they are doing business open source and I'm in charge of product they develop which is called kinetic i'm I'm going to talk about it little later if there time evolution of storage and Linux and I thought that well 10 years of frost on it's time to go back in history and and reminisce about things that have and so on so what I'd like to do is to give you a kind of a trip down memory lane that that is a pun intended to remember that the topic that I'm going to talk about this storage and in this particular 1 2 given only a whole the whole concept of storage and this has evolved over the years I've been doing quite a lot of research this in the past few weeks from and when it comes to managing and providing storage for assistance with using linear system so your options are seemingly endless if you if you just look at how much how many file systems are supported on Linux today protein is pretty much ubiquitous and trying to cover all the aspects in in hours just the have hope them and I would be obviously lost so I try to pick out a few things so to say and starting with the history local file systems so this whole thing is called when talking about what the next generation of AI systems that came afterwards the most common services used to term limits system into a file server and then also try to give in and out look at today's landscape of all 5 systems and storage and where we're heading granted it's a very broad topic and I'm going to talk about lots of different technologies and I will not be able to cover them all in much detail and I've already been throwing out quite a lot of stuff while I was doing my research I found so many small technologies and process that I found interesting and started making knows about them and I end up with hundreds of sites basically and had to just get a step back and think OK this is not going to work on so yeah maybe at some point should be turned into workshop was something but so far it's more of a history no would you give you a broad picture fortunately of most of these topics are very well covered by other talks and presentations also in this conference and where appropriate and want to refer to a topic that is somewhat different of the technology um kappa will starting with local and then going out into the world world distributed justify systems yeah if you just look at local file systems to date on the Linux kernel has found 40 + has that adjusts capable of storing data on a local disk in various forms of just looking at what systems are aware of flesh devices for example would cover the top always on easy so I I try to pick up people pick out the prominent ones popular ones the ones that are most widely used to to give you the impression that and I also try to focus on on stuff about the device the the whole topic of storage of course also has a hardware component to the fastest in needs some means of tracking technology to store data and persistently some sorry we could also create a completely set of presentations about and the block storage layer how data is being stored what device to what technologies are available there out of scope and even though they are highly interesting so we keep it on the software side on the upper layers of the Linux kernel and also with the especially with the the the current technologies guess they run on Linux but most of the times they don't even need to a common component anymore person and also I wanted to do a few yeah conclusions upfront things that I kind of realized and learned that while I'm assembling the information for this talk and the interesting thing about this this fisher when you look at health systems of you end up was of to watch a lot of time and still nowadays and it something I found quite interesting and be the peak of of SourceForge was around the year 2000 I would say 2002 and many of the projects that are still part of the Linux kernel was initially started or evolved being hosted on sort and from a historical viewpoint if you if you look at it nowadays and you go back and try to figure out how this whole thing in working with what kind of history involved and sourceforge basically is a very important resource and even though nowadays everybody source which is that we're not going to use it moved to get but in my opinion it plays a kind of a crucial role in this whole history of Linux and it would be set to see this resource go away because you would lose quite a big part of history of Linux with 1 so this is a risk that you will always have an and you who will guarantee you that the test is going to be around in 10 or 20 years so the good thing about open source and hosting develop nowadays it's very easy to move and copy stuff around and headed somewhere else but million news archives backtracking history all those things in tend to get lost along the way which makes doing research on his sword purpose is quite high and also an interesting observation which is what more often and issue in previous times especially when it comes to things like storage and file systems new device drivers the Linux distributions in these days played a pretty important role because they employed to develop and they were driving the development and and bondsman's of these technologies for what and how many times they didn't wait for leanness actually incorporate the coding is currently distribution kernels spec then and sometimes the the amount of patches that were applied against the vanilla cone would be given the option comes also called and it was quite a maintenance burden and my work that's a loser from 98 2002 and who would model was the main color maintain a force was a he was yelling and screaming because you have he
had many vendors that wants to get their code into Linux like for example IBM with with the volume manager of a system and and they were all nagging Sousa to incorporate the code into their country even before it was actually in a 9 and was was was not and not the only company doing about it was also catching the current quite heavily so they ended up with a very unique unique on of a kernel that only they were shipping in this kind of form which made it hard for the upstream developers to accept but reports because well it was a distribution kind there had been no easy way to reproduce the problems that occurred there and also put a lot of burden on the distributed because they had to maintain and support these technologies by themselves so they have to employ be developed as the supporting if problems where in the court they 1st have to figure out OK so is this a problem based on caused by our patch set some modifications so is it a genuine maintenance problem so that the kernel engineers and distributions also weren't quite happy about that because they were basically swamped with problems and issues that they 1st have to figure out where they're actually and I'm looking from 0 and and by the way found that How do I stop this from I would be very interesting if to have additional insights or anecdotes anything to add to correct about the things that I'm talking about right away and are usually preferred doing a dialog instead of just giving a monolog up here and and you just not all fall asleep and on what you internally disagree because you know that I'm talking rubbish but if that's something that you know what the subject and I don't know much about most of this topics so few free to add your comments and if you have additional insights to share please do I think this would be much more interesting for the torque for itself OK but also an interesting observation is this guy here Christopher link however put this he left his mark so basically at all levels of the storage site and the trolling mailing lists I've been looking at patches and release announcements this name pops up quite frequently when it comes to file systems and and the storage layer so interest of of In my opinion he deserves to be kipedia entries just for the work that he's about and assembling all the contributions that he made to be quite challenging and since he has made such a broad contribution to the internet and storage layer the 1 thing about it he knows much more about these topics than I do and it's actually around so you don't miss his talk tomorrow at 2 PM and he will be giving an overview about the inner workings of the storage layer and show you how these pieces fit together and how they work together things that I'm not just going to gloss over very broadly but also I would like to thank Linux we can use for their continued work I we can use is still in the website layout well it still looks the same as that when they started painstaking but the the the content they produce is also very valuable and lots of the articles from there I think reading a lot of articles that were covering topics that have been doing research on continuous dialog also do a very important job and into keeping track of what's happening on the and Toscanini use I think he's here this week in this work from Heizer he has a very great job on also following the mailing lists and some arising what's going on land so a big thank you to him as well and of course well Wikipedia also was a good results when doing such a project for presentation and I found a few things that modified so as being also updating the sum of the Wikipedia entries along the way and if I had more time and more things to contribute by getting the top down in time at some point my important OK history the very early days this is where I'm going to start and links was developed all that the start of the development of was a 1991 willingness to was basically looked himself to his room for an entire summer and he got interested in the hot topic of operating systems where learning about in the University and 1 thing he hated about mimics was determined terminal-emulation he needed a tool that allowed him to contact his university so on and the mean external relations so that that he decided to write 1 on his own bold we he started he bought a 386 PCs remember correctly and started with a very simple scalar that nothing else imprinting and BC so we had to borrow phrase that we're doing independent task the next step you started expanding those tasks that 1 of the task was talking to the mode and the other task took the output of the model printed the screen so the very early beginnings of the terminal emulation he did the development on Munich's itself I think he spent a few months just and modifying and updating the point where it was useful development environment because for minute for the researcher operating system not so much for production work so we use the GCC compiler and all these things and started get going on at some point you realize that well my terminal-emulation needs to store stuff somewhere you need a persistent storage and well we the canonical choice was to create something that is mean compatible because the host OS that he was running with was semantics and I think you just have 1 modestly have to figure out a way to store the data where running his operating system and the same storage medium than the other operating system he was developing but it also helped that well since it was a research operating system meanings and 1st are well documented and and easy to reimplement in a way so this allowed him to have a terminal-emulation program that saved files on the file system and well this is how it started it has a lot of limitations 64 megabytes of maximum has systems science which was actually quite bearable for these times because well in the early nineties hard disks work that they can be honest but of course if it were limitations that at some point and made it clear that changes had to be made but what so yes it started as a terminal-emulation which eventually turned a operating system so 1 of the 1st things that were implemented wary early on was that of the yes S the virtual file systems which because on the 1 hand cleanness of course wanted to keep the meanings persist and road because he needed for practical purposes but he also wanted to make sure that there is an easy way to add another house so the we've as there was developed and Chris for internal provided the image patches for that mean integrated and publish them with 0 9 6 somewhere and 91 I mean but he heavily revolted in trying to look at those old consultative period where
exactly and was introduced I think 0 9 6 was the 1st thing where I could find to be of s . h file and the source code so this is when we the development started like was was basically dumping tarball cement FTP site and infrequent basis and then published patches on top of that before he made a new talent show a few weeks later and the VFS basically it's kind of an abstraction there in the corner and on the user landside you system calls like open read and so on and the current takes those calls and figures out OK which file system or which file and that call the belongs where do we have to through and this allows you tool creates mom points using different processes so you could have your real fast system on the next and then Monson issues 0 antiques devices or whatever and this really was that the key component of or very 1st step in making the news modular and more attractive for others to contribute so the inclusion of the as quickly result in a whole slew of systems that have that we're developed and included in the current and the very 1st fastest to make use of the offense was speaks the extended system in the car was the guy who wrote it initially and it removed to 1 of the biggest limitations that many except in particular system size was increased to 2 gigabytes and is where all of a sudden 2055 Chancellor just 14 charts and and a much more more modern infrastructure and yeah just a bit more modern than what the news has system look like and thing edited in 0 9 6 see paper 92 Salinas has already been around for a year on it wasn't perfect but it was already a big step forward compared to what was there before so 1 of the key issues that the customer was that it was using on linked lists to keep track of the blocks and angles and the more you were using the crisis in the more extended to fragment itself and the performance just agreed horribly at some point so it was clear that excuse just a stopgap measure before something much later came along and it was the 2nd extended systems which also was implemented there any generate 93 also very early in the in in the early days of of the most come and this is 1st system was basically designed from scratch with extensibility in mind them so all the beyond this data structures already left some room in in the sense that you have free space in in in box and I know that you could possibly repurposed later if you want to extend the system I maybe that's where they called and system because it it really was quite extensive extensible of inspired by existing power systems again like the Beastie fast system and it had all the things that you you would expect from select the possibility to store times that's why files different box sizes you could terms fires mutual which makes 5 basically unmodifiable on disk and in quite a lot of modern concepts that can be used to seeing constant notice as well and again to terabytes that the size was again increased drastically and yeah you you later if you come them at k books you can store up to 32 terabytes in yeast each us so it started to at some point and had to live with the limitations that the Linux kernel rounded to give it on that say the view fist layer all the block device there but the extent to itself also further evolved over the years and and and get new features and and in fact exceed 2 was an ideal testbed for things or features that were developed much later in the time like for example politics as x is controlled from extended actually use things like that they were 1st added to the east to file system because it was well designed the developed new the court of it very mature and stable at some point and but made into a perfect candidate for starting with experimenting with new features that needed in the next and it's still has some purpose nowadays and extent 2 based because it's very simple and robust is still a good choices for example using minutes on embedded devices it's a good candidate for flash storage because it's in the presence of adjoining 1st and there's not much chance on the disk itself it doesn't from a lot of rights store data all well let's say special partition that only contains the current and the need to around that's such what is usually system where not much is really changing and do you really needed a big challenge has been told that if if spaces techniques to to estimate a good choice for I guess as a simple as they allow people to add a lot of new fast systems to expand they started importing existing 5 systems that made them capable of accessing existing data on a wide variety of formats and I'm just thinking that MS those for 1 example because you need DOS was probably the most common use operating system for people that were developing units that on their PCs so they need a way to access the funds there and then I understand introduce the initial and what support very early as well but was later extended using the fat which has some more extended features action groups of it has been initial call it X and this those that had to rename because of the trademark but but pattern is the support was also included very early on even before that based on my research they already included as a collection of 20 and tools which basically is it is to use the land to that was capable of accessing fat formatted file systems you could think of it like an FTP thank you start and tools pointed to a device and you could then get put individual file so it was not like you enter a directory and you directory structure but you had a client that talk to fight system over the user landscapes and and tools was around even before Linux was developed and it was very easy portable and Jim wins that included it the very early versions of the limits of properties that he maintained which in my opinion 1 of the 1st Linux distributions because they were floppy disks that included just current but also a show and and rudimentary tools to get started later on in the office came aboard Microsoft was also furthering the development of the 5 system SELinux people were looking at
way on how to access the interface has systems emerging from this was the the 1st 1 take a stab at that some the problem with is and he this is a proprietary file system and had to be reverse engineered and which is a the case of some of the 1st that's included in and that the specs specifications were not open so people have to basically look at what the price system is doing on the rock layer and really reverse engineer what was going on there so although the interface travel that was included in the Linux kernel to my knowledge never fully supported right writing 20th this assistance but she was always able to mount and read data offered but and the kernel module I think they started experimental support and you had to look more with a force option 20 the right support and he was still on your own because they were there was always a risk that writing data to an interface a system which is angle everything and you will be lost later on the code was further developed under a project and team in g which replaced interface driving in the corner and in parallel there was a project going on and and 3 G where they simply created an interface driver and use the land and use the fused system which I'm also going to come a bit later on to to talk the the block device the open source project page on sourceforge nowadays redirects to company Teixeira and according to their website they have further developed their interface as driver to no support other operating systems like Mac OSX free the Solaris QNX and so on so it was probably a good idea to to implement this on on the fuse level because it was much more palatable to these other platforms that but pattern and if s just to prominent examples of what people were doing based on the capabilities they viewed as they are provided to them something on the next big step on my milestone in my opinion happens when 5 systems became generally 5 systems where everything that we've talked about before with the exception of into this may be of systems that maintain a very special structure on disk and but over this in in this talk here on this 5 system check versus journaling of you probably if you have been around with this for a while you probably know the message and file systems level where has been going for too long at check forced and Murphy's law mandates that this always happens when you have to read which you find over and people are waiting for it to come back so we extend to had several ways that all Sara criteria spirit forced system to be checked well the the 1st of course is an unclean amount if your system crashes the system goes down the street to have deflected wasn't clear and so it was complaining of you probably crashed I think it's better if I check my consistency and internal structures and of it went on but also depending on how many times you've multitude wanted the file system or how much time has been passed since the last 10 years mounted and unmounted 1st and system would also trigger check and administrators were yelling and screaming about that because distress the bigger storage requirements were increasing the 55 systems worse than spending hundreds of megabytes gigabytes think again this was in the nineties but processing checks could take a very long time and this was simply not an option if Linux and wanted to become an enterprise operating system that could compete with the other Commission's insistence that so this problem became more and more apparent as the size of this drives increased and and the amount of data stored on them an increased in another and but there were several projects in parallel that started to alleviate this problem or shortcomings Institute but the most prominent ones are the extreme 3 5 system is checks affair society instituted is and the rise of the system and these are the ones that I'm going to talk about in brief in the next few slides the cool who has ever cursive modifies is check taking hours OK so they really you know it's going to finish at some point you just don't know when that's just mind wrecking and very annoying and especially if it at the spotted an arrow and you were clueless about what was going on them because then it it's started to attempted fixing things the debug messages just got more confusing and it got really scary at that point and I have been in this situation must have way too many times OK the solution to this problem from a theoretical point of view is that you use so-called journal so instead of changing the data in the file system directly it being in place so to say you create what you basically mimic what databases and really in in the sense that you define transactions states and had to to have happened in atomic weight so that you will succeed only if for and you 1st write the intention to change something into a log file and from that the changes will be propagated to faster 1 so in the event of a crash and the recovery basically just included and 15 minutes of OK you still need 10 minutes that OK and each area anyway so journaling was the solution to this problem because well he just read the journal and figure out what was the last injection that has either completed or fail and you could get the file system that inconsistencies in much faster way and again there was a new layer created the GBD journaling block device that not all of the juggling 5 systems actually use it to 1 that does this is the each 2 3 5 systems DVD was developed in parallel with the file system and which is a generic service that any high system could use well to become transaction and and and journaling functionality so the system had to communicate with the GBD in order to announce that the transaction is taking place the data is going to be written out to disk and so on and this seems pretty develop this in parallel with 63 and Jimmy was also used by Oracle false if it's too and these 2 closest started by for King James Dean to and here we have the C. 3 basically in a nutshell they took speaks existing IIiX to to call because it was mature well written they know it all and headed John and capabilities because this was back then the most serious limitation the and then the way they did this using the GBD got there was actually quite smart and and since it didn't really change the
on-disk structures of a migration from east to to 63 was very easily doable because well you just to enable the journal and off you go and you could also disable the journal again in case that you had to move mounted from and all this and that was only capable of launching the future for assistance and it increases the file size terrabytes and and it if you have a new features that we're in the 2 thousands already some development is going on there as well the IBM also realize that there is a need and as the solution waiting time they have been working on Germany 5 systems for the AIX operating systems and the early nineties already so they know this topic quite well on back in the States idea made a very big announcement that they are going to support the development of limits with 1 billion dollars on sure they wanted to make sure that their technologies were pushed into Linux and and they remain relevant and dominant and had a foothold in the dorsal saying but I been contributed to quite a lot of useful things and and the jf s is probably 1 of the of the key contributions that which we call them I've been working with david Steve Drumet times its is on making sure that the so the news was always shipping the latest version of G as and so on so they were very you go on on making Jefferson and its success was quite fun working with these guys just sizes were increased significantly again of and interestingly the code base later then found its way into the x as to again so but they basically we looked at the existing GFS implementation decided to scrap because it didn't support SMP systems and was not very scalable portable and just took the concepts and ideas and started from scratch again with the support for multiple operating systems in mind and then and x and later all where is it always to have a definite mentioned somewhere else whatsoever yeah so but was also capable of using the fastest and so on a Linux books using Jervis could also moment and always to socialist universe for assistance with problems was another contestants isolated from it was quite heavily supported by effect and I think so that was the only distributed really putting a lot of energy and resources into Gaza and size of the the the original creator conditional 10 so again already anyway just let me know when and under what he was quite a controversial person and in the way he pushed sizes and and His while what I call it stubbornness I don't know but uh he had lots of issues with the current community there was a lot of friction in in in the development process going on and so for example employed Chris Mason who developed the journaling part of of the size of fist and then they were the 1st ones to really claim that a stable and can be used and that's an observation forgot to mention the beginning 5 systems on Linux I haven't really really hard time because people read about somebody losing his data on systems and would never touched therefore nobody tests it therefore it doesn't get better if it's a catch 22 and it's especially worse if it's a lot of that is isn't you part of maintenance so that is that is really have to go through quite a lot of iterations and lots of negotiations and discussions until it was finally included in the eyes of his 3 . 5 students part of the common nowadays but its star is thinking I would say that it is pretty mature but the development basically stopped and just fixing issues that the spring of because the codebase evolving or if if this reduces by I'm not aware that anybody is actively working on 3 . 5 however something that really surprised me is that there a force to learn to develop and they have separate patch sets for Linux kernel afforded 1 already they have a website and mailing list so these guys haven't given up on the whole project yet but but you would ever be part of mainland Linux question mark animals but it's it's interesting to see that some developments still going on there and it if it was really the 1st generate high system that was actually included 2 minutes at some point on the beat all the other implementations when it comes to the timing of the intended improve modern structure and that made it quite suitable especially for small it's like a service may have stolen server or web so small page and things like that but it had issues in other areas as checks surface on the other hand came from the opposite direction files high-throughput lots of concurrency him she I was using the fastest among the Irix Unix labor for quite some time development started in the mid to early nineties and they announced it in 1999 but I think it took them years to go through all the legal process and talking to all the developments that involved in the file system to get their part of the history and making sure that all the 3rd party components be included were taken out so that review process 2 for about a year and I was quite impressed reading about how many iterations that you go through and and that they really persevered and push this forward until they came to a point where a Linux version was available these guys were the employers from his ji working on it and a number of high very popular figures from community started supporting the only on like and the key from the user or young cows and again just OK because also I think a key contributor to accept end up being made its way into the kernel in 2004 I think and we need many of the distributions to the default to accept this as the key file system and I think looking at this in the corner where there and I I think it's the you're most systems as well Linux deployments I remember that we were recommended for my stress that is what you get in all right In this talk about the theory of status man the idea of the all of these laws actually poses the question but I'm going to repeat it for the audience I was I think this is about his experience and if he was still suggest nowadays and yes they were customer deploys using 63 and they had time the formant issues which switching to accept as basically solved so it's very mature that's robust from think from from all of the joining 5 systems exists it has the biggest range of support and user based on most of the employs work for red head node incentives under active development so it's a safe choice alright persistence and nice but you need to put them somewhere and in the early days
you just have physical disks and partitions where you could put a 1st Monday HTA death D 1 on if you ran out of space so what did you do he took a bigger desk created a new file system copy the data across model somewhere and which was not very flexible and once again was something that cause downtime which she didn't want to have in in cell operating system so the idea was OK how could be abstract that layer in the medium term introduced in 1988 by small-time what's the solution to that where basically said a hard disk is just a physical volume summer underneath and I'm going to put a layer on top of that that basically hides the underlying physical structure just create logical volumes on top and make them so flexible that it doesn't matter what kind of physical storage assignments and the the and then allow them to create volumes put assessment of of that space was getting tired just put another disk in my rec added it to the physical world expanded my logical 1 member good to go so this was really going on at end really made starch management and Linux so much easier and not just in the 7 moment but also on on a home computer pieces so if your home directory it's on separate volume suddenly becomes too small it's node is much easier to to increase the size and even though that some of the installers the thing you want to install a studio for us to put in 1 2 on physical partitions if you don't force you to use logical volumes which is kind of odd the lesson yes so another layer a very crudely actually because it allowed me to do lots of all things and the 2nd version of the logical volume manager used device today to get the job done so the 1st version of the game was on yeah hard coded was just as in between the 1st system and the disk with a view to encounter to the 6 you it was using the device that device that they really did the brunt of the work and device-mapper allows you to define the virtual devices that you can stay on top of each other you have applied to infrastructure that allows you to modify the data while it travels from 1 block the to another 1 which gives you unlimited opportunities and and and and ways to to to mess with the data where the travels across this layer and yeah the multipath is 1 of the the interesting aspects here basically it's the a way where the storage layer can access the same storage devices through different storage path like 5 channel control or I scuzzy or whatever and it's just a where OK you can take this route it can take this will but you will end up in the same device and enter the multipath knows about this and can even balance between those 2 roles in order to distribute the load or if 1 of the connection fails it wasn't just the other ones to you get both scalability and high-availability for his device that uh the into all these components initially were created at the company insisting on which was then later acquired by ready to commence also an interesting topic that many of the key components like except Steve resonant ponds or not primarily minutes and you look at that and nowadays but it doesn't seem to be the case that this company's going down soon the intention of something that was recently developed another nice trick using device-mapper you could create a separate flesh drive catching device and so if if you're a rotating disk to slow intention allows you to put flesh discuss a cash there into a distortion of so to see also developed and most Hardin who now works for that is part of the team that might be the coveted elevated to basically be improved for the LVM 1 was already provided by adding more flexibility and interesting things that happened reports and string provisioning that allows you to basically create this that a bigger than the physical storage actually provide and it also had a class W management options if you have multiple machines sharing storage and the ends capable of talking to this this will make sure that you do something messed around the and of project by IBM that was quite not worth back then was the EVM as the enterprise quality management system where they basically took the concept of how they were managing storage on the ported to Linux but they initially also wanted to create kind support for that but the Linux developers thought that device-mapper and into our a much saner choice and lots of politics and probably and lots of lengthy discussions the interesting thing here although not with the was that the IBM developers at that point then decided to take we're not getting our cake so we are going to adapt UBMs to use and I really don't but the effect what what time is it I don't have a clicker tenants OK so I let me go on for 5 minutes and then the dual the unit but yes so Linux decided every into was the better approach the idea in their accepted this decision and modified to use 100 get the device instead of using their own implementations the area of union is interestingly in the sense that it does it does more than just volume image and it is capable or it knows about 5 systems that are being put on top of a 1 year and it was only possible to grow and shrink systems along with the the underlying volumes that's the capability that true got much later in the process for example but nowadays the VMS has no more relevance and IBM basically stop development in 2006 so store services sigh going to run through this no I'm sorry I think it's sort of an entity running out of time that's embarrassing and benefit so
we have been talking about local file systems all the time but and in order to run a so we need to share the data with the the world and and if this was that the very 1st implementation on that the 1st way of how to exchange data with with other systems and if its head is both a client and server part and it has gone through several iterations and it's Ford and 1 is currently the most of the most active version and handling also developers are really on the forefront of making sure that in benefits of 1 in Linux is up to speed there was a time when the limits implementation of and this was lagging behind this has changed dramatically there's a lot of effort being put into making this 1 of the reference implementation but they're the the initial this implementation then started very early on then of course we have some time you all have heard of it is take a file system on the Linux disk run some bond of of it and share it with the windows and by and this is basically house on understand with
an yes later on is in the was expanded to be the common Internet fast system even though I doubt that anyone is really sharing disks over 6 over the public Internet anyway Microsoft has also been furthering the development and was catching up as well at some point you had to kernel modules that allowed you to mount to share from what know so the old in BFS module and the mom on insists module and is in the office still mentioned in many how tools and documentation so that shouldn't be used anymore and that it has been dropped quite a while ago already which gets us to current systems 64 the latest version of some still consider it to them since they belong not ready for production interestingly as a set of systems are very tricky subject excuse for also started basically as a branch of 63 that was then gradually extended and expanded and improved yeah it added the function more modern concepts and especially the new extend on-disk format allowed it to watch the much more scale and the faster but it also was the uh as a step that made it incompatible speaks the 3 D structure so when it says it is capable of mounting the 3 5 systems you wouldn't make use of any of the benefits of 64 if you don't really converge faster and there was a market that you can set and any new data would be stored on on the pulmonary 53 system would be induced extends which gives you an mixed on disk moment of both these 3 and 4 at that point you can go back basically it's a well I think it's pretty solid the default system for some distributions it's been around it's another option for example this is but yeah the better 1st and I think giving talks about but if it's a few years ago we come basically it's not justify system it also includes rate and volume management so it covers the entire stack and if the system is capable of talking to the disk directly you can do a lot of very interesting things checksums for example and the whole of the rate and volume management all these things wouldn't be possible without budget is being include charges that the school if you should use but affairs don't put node the in between or a DM rate or whatever give raw physical this this is the best approach cool thing about but if s you can convert existing east of the 405 systems on the fly and it would create a snapshot of the old seeks to fast so that you can you can go back to in case you want to go back but so it gives you a way to keep your existing data you would have to copy them all around Chris Mason is too busy working on its notice he works for Facebook vise and all the stuff of development and Oracle then moved to fusion i altogether some other guys and know faced was sponsoring his development still ongoing features of being added but of is basically is the future local file system from the in my my opinion and we have the effect of Sun open-source Solaris OpenSolaris in 2 thousand and 6 or something and and part of OpenSolaris was z of support system which was a revolutionary break them because similar to what about services do nowadays it's not justify system with it does volume management it's very scalable provides a lot of robust features and In the vote never be part of the Linux kernel due to the licensing its lessons and CDL which is incompatible with the GPL on there is a very active community around the codebase opens nowadays and z effects on Linux which need to download and install separately you have a very solid star solution that and solves a lot of problems here I'm sorry that have to run through this quickly but in BT
was something that I found interesting have a remote server to share block device mounted somewhere else but look there's a talk about it tomorrow if you need to know more DIB allows you to create kind of a rate over networking events over here so that there every data you right on the 1 side is being replicated to the other sets block identical copy of the data on the 2nd system due d usually works in in synchronous mode but you can also be switching to as increments model you have latency between the nodes and it supports catching up in case the conduction goes down and it needs to replicated data that has been sold locally and you can change roles so as provide lots of and high-availability capabilities as well as currently there work on RB gene 9 the next incarnation which will support more than 2 notes in like the size of a studio nobody developed parts like to get involved in the discussions this is that politically correct anyway there's lots of discussions going on about the review right now also they're being they are having a hard time getting coding testify
systems so this suffer systems that you can use to share a disk between 2 different systems I ran out of time it's a bummer but I'm going to give you well you can get the pdf slides later on but let's keep minutes for questions with tell you what I'm going to flip through the slides were quick would have talked about also this Jesus requirements why we need storage clusters set lasted and yet I apologize that was way too much but I didn't know about I got just too excited about all these different technologies and what you can do with them so be with me and I hope it was still insightful in some way and is it gave you back fond memories of things that you hope to have forgotten already any comment question anything that you want to leave before I leave but it didn't overwhelmed with too much stuff OK things I hope you still enjoying it and negative
Offene Menge
Distributionstheorie
Bit
Punkt
Prozess <Physik>
Kinematik
Extrempunkt
Computeranimation
Eins
Kernel <Informatik>
Metropolitan area network
Freeware
Notepad-Computer
Dateiverwaltung
Vorlesung/Konferenz
Inklusion <Mathematik>
Softwaretest
Hardware
Güte der Anpassung
Stellenring
p-Block
Biprodukt
Konfiguration <Informatik>
Arithmetisches Mittel
Softwarewartung
Software
Dienst <Informatik>
Generator <Informatik>
Forcing
Festspeicher
Projektive Ebene
Information
Zeitabhängigkeit
Web Site
Subtraktion
Total <Mathematik>
EDV-Beratung
Kombinatorische Gruppentheorie
Term
Kappa-Koeffizient
Bildschirmmaske
F-Test
Software
Rhombus <Mathematik>
Mini-Disc
Inverser Limes
Luenberger-Beobachter
Datenspeicherung
Zusammenhängender Graph
Inhalt <Mathematik>
Softwareentwickler
Speicher <Informatik>
Ereignishorizont
Datei-Server
Drucksondierung
Zeitabhängigkeit
Open Source
Anwendungsspezifischer Prozessor
Physikalisches System
Packprogramm
Patch <Software>
Druckertreiber
Mereologie
Codierung
Datenspeicherung
Kantenfärbung
Distributionstheorie
Gewichtete Summe
Virtualisierung
Extrempunkt
Programm
Raum-Zeit
Formale Semantik
Festplattenlaufwerk
Client
Dämpfung
Mustersprache
Radikal <Mathematik>
Dateiverwaltung
Vorlesung/Konferenz
E-Mail
Analytische Fortsetzung
Auswahlaxiom
Addition
Sichtenkonzept
Kategorie <Mathematik>
Winkel
Güte der Anpassung
Biprodukt
Rechter Winkel
Plateau-Problem
Gcc <Compiler>
Programmierumgebung
Lesen <Datenverarbeitung>
Subtraktion
Mathematisierung
Digital Rights Management
Flash-Speicher
Methodenbank
Informationsmodellierung
Bildschirmmaske
Weg <Topologie>
Spezifisches Volumen
Inhalt <Mathematik>
Datenstruktur
Maßerweiterung
Inklusion <Mathematik>
Binärdaten
Diskettenlaufwerk
Binder <Informatik>
Elektronische Publikation
Partitionsfunktion
Patch <Software>
Resultante
Bit
Punkt
Prozess <Physik>
Versionsverwaltung
Gruppenkeim
Skalarfeld
Übergang
Internetworking
Kernel <Informatik>
Metropolitan area network
Einheit <Mathematik>
Prozess <Informatik>
Figurierte Zahl
Einflussgröße
Funktion <Mathematik>
ATM
Abstraktionsebene
Systemaufruf
Betriebssystem
Quellcode
p-Block
Frequenz
Arithmetisches Mittel
Softwarewartung
Dateiformat
Projektive Ebene
Verzeichnisdienst
Varietät <Mathematik>
Web Site
Filetransferprotokoll
Quader
Gruppenoperation
Kombinatorische Gruppentheorie
Term
Code
Task
Polarkoordinaten
Mini-Disc
Luenberger-Beobachter
Inverser Limes
Datenspeicherung
Speicher <Informatik>
Softwareentwickler
Grundraum
Bildgebendes Verfahren
Gammafunktion
Touchscreen
Leistung <Physik>
Mailing-Liste
Physikalisches System
Office-Paket
Arithmetisch-logische Einheit
Moment <Stochastik>
Basisvektor
Emulator
Verkehrsinformation
Distributionstheorie
Momentenproblem
Landau-Theorie
Reibungskraft
t-Test
Iteration
Extrempunkt
Gesetz <Physik>
Homepage
Richtung
Dämpfung
Reverse Engineering
Notepad-Computer
Dateiverwaltung
Vorlesung/Konferenz
E-Mail
Auswahlaxiom
Schnittstelle
Metropolitan area network
Managementinformationssystem
Sichtenkonzept
Gruppe <Mathematik>
Winkel
Ereignishorizont
Dienst <Informatik>
Forcing
Menge
Ablöseblase
Server
Ordnung <Mathematik>
Lesen <Datenverarbeitung>
Mathematisierung
Systemzusammenbruch
Systemplattform
Knotenmenge
Spannweite <Stochastik>
Modul <Datentyp>
Zeitrichtung
Datenstruktur
Soundverarbeitung
Open Source
Elektronische Publikation
Modul
MIDI <Musikelektronik>
Patch <Software>
Wiederherstellung <Informatik>
Bit
Punkt
Prozess <Physik>
Datenparallelität
Versionsverwaltung
Kernel <Informatik>
Eins
Übergang
Metropolitan area network
Arbeit <Physik>
Mehrrechnersystem
Schwebung
Figurierte Zahl
Parallele Schnittstelle
Default
Inklusion <Mathematik>
Umwandlungsenthalpie
Lineares Funktional
Datenhaltung
Betriebssystem
p-Block
Speicherbereichsnetzwerk
Konfiguration <Informatik>
Rechenschieber
Softwarewartung
Transaktionsverwaltung
Projektive Ebene
Schlüsselverwaltung
Normalspannung
Message-Passing
Aggregatzustand
Web Site
Implementierung
Code
Physikalische Theorie
Multiplikation
Benutzerbeteiligung
Mini-Disc
Inverser Limes
Luenberger-Beobachter
Zusammenhängender Graph
Speicher <Informatik>
Softwareentwickler
Drei
Grundraum
Widerspruchsfreiheit
Schreib-Lese-Kopf
Dateiverwaltung
Trennungsaxiom
NP-hartes Problem
Systemverwaltung
Mailing-Liste
Physikalisches System
Vorhersagbarkeit
Energiedichte
Druckertreiber
Flächeninhalt
Injektivität
Mereologie
Unternehmensarchitektur
Resonanz
Prozess <Physik>
Punkt
Momentenproblem
Versionsverwaltung
Iteration
Raum-Zeit
Eins
Metropolitan area network
Festplattenlaufwerk
Client
Einheit <Mathematik>
Skalierbarkeit
Prozess <Informatik>
Dateiverwaltung
Vorlesung/Konferenz
Emulator
Große Vereinheitlichung
Auswahlaxiom
Sichtenkonzept
Physikalischer Effekt
Stellenring
Hochverfügbarkeit
Betriebssystem
Ausnahmebehandlung
p-Block
Entscheidungstheorie
Konfiguration <Informatik>
Dienst <Informatik>
Forcing
Datenverarbeitungssystem
Ablöseblase
Server
Projektive Ebene
Ordnung <Mathematik>
Verzeichnisdienst
Schlüsselverwaltung
Zeichenkette
Subtraktion
Physikalismus
Digital Rights Management
Klasse <Mathematik>
Implementierung
Zellularer Automat
Mathematische Logik
Term
Virtuelle Maschine
Informationsmodellierung
Spieltheorie
Mini-Disc
Digital Rights Management
Inverser Limes
Installation <Informatik>
Datenspeicherung
Zusammenhängender Graph
Spezifisches Volumen
Speicher <Informatik>
Datenstruktur
Softwareentwickler
Bildgebendes Verfahren
Gammafunktion
Schreib-Lese-Kopf
Soundverarbeitung
Trennungsaxiom
Einfach zusammenhängender Raum
Routing
Physikalisches System
Partitionsfunktion
Quick-Sort
Flächeninhalt
Mereologie
Gamecontroller
Verzerrungstensor
Unternehmensarchitektur
Verkehrsinformation
Distributionstheorie
Stereometrie
Facebook
Abstimmung <Frequenz>
Punkt
Momentenproblem
Gemeinsamer Speicher
Physikalismus
Digital Rights Management
Versionsverwaltung
Extrempunkt
Service provider
Kernel <Informatik>
Internetworking
Metropolitan area network
Informationsmodellierung
Knotenmenge
Mini-Disc
Nichtunterscheidbarkeit
Dateiverwaltung
Kontrollstruktur
Vorlesung/Konferenz
Spezifisches Volumen
Softwareentwickler
Datenstruktur
Default
Soundverarbeitung
ATM
Lineares Funktional
Zentrische Streckung
Open Source
Hochverfügbarkeit
Verzweigendes Programm
Stellenring
p-Block
Physikalisches System
Biprodukt
Bitrate
Modul
Ereignishorizont
Office-Paket
Konfiguration <Informatik>
Dienst <Informatik>
Menge
Mereologie
Server
Dateiformat
Wärmeleitfähigkeit
Rechenschieber
Offene Menge
Metropolitan area network
Freeware
Software
Festspeicher
Mini-Disc
Vorlesung/Konferenz
Dichte <Stochastik>
Physikalisches System
Speicher <Informatik>
Cluster <Rechnernetz>
Computeranimation

Metadaten

Formale Metadaten

Titel The Evolution of Storage on Linux
Serientitel FrOSCon 2015
Teil 05
Anzahl der Teile 80
Autor Grimmer, Lenz
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 10.5446/19590
Herausgeber Free and Open Source software Conference (FrOSCon) e.V.
Erscheinungsjahr 2015
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Linux and Open Source Software have always played a crucial role in data centers to provide storage in various ways. In this talk, Lenz will give an overview of how storage on Linux has evolved over the years, from local file systems to scalable file systems, logical volume managers and cluster file systems to today's modern file systems and distributed, parallel and fault-tolerant file systems. Lenz Grimmer

Ähnliche Filme

Loading...