SQLite: 35% Faster Than the Filesystem

2 years ago (sqlite.org)

No file system attributes or metadata on records which also means no (xattrs/fattrs) being written or updated, no checks to see if it's a physical file or a pipe/symlink, no permission checks, no block size alignment mismatches, single open command.

Makes sense when you consider you're throwing out functionality and disregarding general purpose design.

If you use a fuse mapping to SQLite, mount that directory and access it, you'd probably be very similar performance (perhaps even slower) and storage use as you'd need to add additional columns in the table to track these attributes.

I have no doubt that you could create a custom tuned file system on a dedicated mount with attributes disabled, minimized file table and correct/optimized block size and get very near to this perf.

Let's not forget the simplicity of being able to use shell commands (like rsync) to browse and manipulate those files without running the application or an SQL client to debug.

Makes sense for developers to use SQLite for this use case though for an appliance-type application or for packaged static assets (this is already commonplace in game development - a cab file is essentially the same concept)

  • > tuned FS + dedicated mount

    For example, Ceph uses RocksDB as their metadata DB (and it's recommend to put it) directly on a block device, with the WAL on yet another separate raw device

    https://docs.ceph.com/en/latest/rados/configuration/bluestor...

    • More just this:

      mke2fs -t ext4 -b 1024 -N 100000 -O ^has_journal,^uninit_bg,^ext_attr,^huge_file,^64bit [/dev/sdx]

      (smaller block size, 100,000 inode file table entries (tuned to the number of blobs), no journal, no checksumming, no extended file attributes, use smaller integer file offset IDs, 32 bit padded vs 64 bit)

      Then mount it and run the same test.

      You could go even further and tune fopen BUFSIZE to be no greater than 12,000 bytes. You can even create this mount on a file inside your existing mount... which is essentially akin to having an sqlite file without needing a client library to read/write to it.

      Anyway - if the purpose is to speed up reads and save disk space on small blob files, there is little need to ditch the file system and it's many many upsides.

That's a very rigorously written article.

Let's also note the 4x speed increase on windows 10, once again underlining just how slow windows filesystem calls are, when compared to direct access, and other (kernel, filesystem) combinations.

I did some research in a database research lab, and we had a lot of colleagues working on OS research. It was always interesting to compare the constraints and assumptions across the two systems. I remember one of the big differences was the scale of individual records we expected to be working with, which in turn affected how memory and disk was managed. Most relational databases are very much optimized for small individual records and eventual consistency, which allows them to cache a lot more in memory. On the other hand, performance often drops sharply with the size of your rows.

This is precisely why I'm considering appending to a sqlite DB in WAL2 mode instead of plain text log files. Almost no performance penalty for writes but huge advantages for reading/analysis. No more Grafana needed.

  • Careful, some people will be along any second pointing out your approach limits your ability to use "grep" and "cat" on your log after recovering it to your pdp-11 running in your basement. Also something about the "Unix philosophy" :p

    Seriously though, I think this is a great idea, and would be interested in how easy it is to write sqlite output adaptors for the various logging libraries out there.

    • > some people will be along any second pointing out your approach limits your ability to use "grep" and "cat" on your log

      And they won’t be wrong.

      4 replies →

    • > Careful, some people will be along any second pointing out your approach limits your ability to use "grep" and "cat" on your log after recovering it

      I wish Splunk and friends would have an interface like that. Sure it does basic grep, and it is a much more powerful language, but sometimes you just needed some command line magic to find what you wanted.

      1 reply →

  • I've been doing this for years. I keep SQLite log databases at varying grains depending on the solution. One for global logs, one per user/session/workflow, etc. I've also done things like in-memory SQLite trace databases that only get written to disk if an exception occurs.

  • It could probably work. For a peculiar application I even used sqlite to record key frame-only video. (There was a reason)

    One could flip it around and store logs in a multimedia container, but then you won't have nice indices like with sqlite, just the one big time index

  • SQLite doesn't look like a good fit for large logs - nothing can beat liner write at least on HDD and with plain text logs you will have it (though linear write of compressed data even better but rare software supports it out of the box). With SQLite I would expect more write requests for the same stream of logs (may be not much more). Reading analysis will be faster than using grep over plain text log only if you'll create indices which add write cost (and space overhead).

    ClikcHouse works really well when you need to store and analyze large logs but compare to SQLite it would require to maintain a server(s). There is DuckDB which is embedded like SQLite and it could be a better than SQLite fit for logs but I have no experience with DuckDB.

    • SQLite is meant for transactional data, DuckDB for analytical data.

      I am not sure which one would be better for logs, I would need to play around with it. But i am not sure if SQLite wouldn’t be a better fit.

      1 reply →

I recently had the idea to record every note coming out of my digital piano in real-time. That way if I come up with a good idea when noodling around I don’t have to hope I can remember it later.

I was debating what storage layer to use and decided to try SQLite because of its speed claims — essentially a single table where each row is a MIDI event from the piano (note on, note off, control pedal, velocity, timestamp). No transactions, just raw inserts on every possible event. It so far has worked beautifully: it’s performant AND I can do fun analysis later on, e.g. to see what keys I hit more than others or what my average note velocity is.

  • I wouldn't expect performance of pretty much any plausible approach to matter much. The notes just aren't going to be coming very quickly.

    • If you play ten note chords — one for each finger — in quick succession, that can rack up a lot of inserts in short time period (say, medium-worst case, 100Hz, for playing a chord like that five times per second, counting both “on” and “off” events).

      It’s also worth taking into consideration damper pedal velocity changes. When you go from “off” (velocity 0) to fully “on” and depressed (velocity 127), a lot of intermediate values will get fired off at high frequency.

      Ultimately though you are right; it’s not enough frequency of information to overload SQLite (or a file system), probably by several orders of magnitude.

      1 reply →

> The performance difference arises (we believe) because when working from an SQLite database, the open() and close() system calls are invoked only once, whereas open() and close() are invoked once for each blob when using blobs stored in individual files. It appears that the overhead of calling open() and close() is greater than the overhead of using the database

I wonder how io_uring compares.

When something built on top of the filesystem is "faster" than the filesystem, it just means "when you use the filesystem in a less-than-optimal manner, it will be slower than an app that uses it in a sophisticated manner." An interesting point, but perhaps obvious...

TLDR; don't do it.

I've used SQLite blob fields for storing files extensively.

Note that there is a 2GB blob maximum: https://www.sqlite.org/limits.html

To read/write blobs, you have to serialize/deserialize your objects to bytes. This process is not only tedious, but also varies for different objects and it's not a first-class citizen in other tools, so serialization kept breaking as my dependencies upgraded.

As my app matured, I found that I often wanted hierarchical folder-like functionality. Rather than recreating this mess in db relationships, it was easier to store the path and other folder-level metadata in sqlite so that I could work with it in Python. E.g. `os.listdir(my_folder)`.

Also, if you want to interact with other systems/services, then you need files. sqlite can't be read over NFS (e.g. AWS EFS) and by design it has no server for requests. so i found myself caching files to disk for export/import.

SQLite has some settings for handling parallel requests from multiple services, but when I experimented with them I always wound up with a locked db due to competing requests.

For one reason or another, you will end up with hybrid (blob/file) ways of persisting data.

  • > As my app matured, I found that I often wanted hierarchical folder-like functionality. Rather than recreating this mess in db relationships, it was easier to store the path and other folder-level metadata in sqlite so that I could work with it in Python. E.g. `os.listdir(my_folder)`.

    This is a silly argument, there's no reason to recreate the full hierarchy. If you have something like this:

        CREATE TABLE files (path TEXT UNIQUE COLLATE NOCASE);
    

    Then you can do this:

        SELECT path FROM files WHERE path LIKE "./some/path/%";
    

    This gets you everything in that path and everything in the subpaths (if you just want from the single folder, you can always just add a `directory` column). I benchmarked it using hyperfine on the Linux kernel source tree and a random deep folder: `/bin/ls` took ~1.5 milliseconds, the SQLite query took ~3.0 milliseconds (this is on a M1 MacBook Pro).

    The reason it's fast is because the table has a UNIQUE index, and LIKE uses it if you turn off case-sensitivity. No need to faff about with hierarchies.

    EDIT: btw, I am using SQLite for this purpose in a production application, couldn't be happier with it.

  • > To read/write blobs, you have to serialize/deserialize your objects to bytes. This process is not only tedious, but also varies for different objects and it's not a first-class citizen in other tools, so things kept breaking as my dependencies upgraded.

    I'm confused what you mean by this. Files also only contain bytes, so that serialization/deserialization has to happen anyway?

  • > I've used SQLite blob fields for storing files extensively. Note that there is a 2GB blob maximum: https://www.sqlite.org/limits.html

    Also note that SQLite does have an incremental blob I/O API (sqlite3_blob_xxx), so unlike most other RDBMS there is no need to read/write blobs as a contiguous piece of memory - handling large blobs is more reasonable than in those. Though the blob API is still separate from normal querying.

    • Do you have or know of a clear example of how to do this? I have to ask because I spent half of yesterday trying to make it work. The blob_open command wouldn't work until I set a default value on the blob column and then the blob_write command wouldn't work because you can't resize a blob. It was very weird but I'm pretty confident it's because I'm missing something stupid.

      2 replies →

    • Most RDBMS's have streaming blob apis:

      MS SQL Server: READTEXT, WRITETEXT, substring, UPDATE.WRITE

      Oracle: DBMS_LOB.READ, DBMS_LOB.WRITE

      PG: Large Objects

      Most of my experience is with SQL server and it can stream large objects incrementally through a web app to browser without loading the whole thing into memory at 100's Mbytes/sec on normal hardware.

      2 replies →

    • I wish the API was compatible with iovec though. As that's what all the c standard lib APIs use for non-contiguous memory

  • > so i found myself caching files to disk for export/import

    Could use a named pipe.

    I’m reminded of what I often do at the shell with psub in fish. psub -f creates and returns the path to a fifo/named pipe in $TMPDIR and writes stdin to that; you’ve got a path but aren’t writing to the filesystem.

    e.g. you want to feed some output to something that takes file paths as arguments. We want to compare cmd1 | grep foo and cmd2 | grep foo. We pipe each to psub in command substitutions:

        diff -u $(cmd1 | grep foo | psub -f) $(cmd2 | grep foo | psub -f)
    

    which expands to something like

       diff -u /tmp/fish0K5fd.psub /tmp/fish0hE1c.psub
    

    As long as the tool doesn’t seek around the file. (caveats are numerous enough that without -f, psub uses regular files.)

    • > I’m reminded of what I often do at the shell with psub in fish.

      ksh and bash too have this as <(…) and >(…) under Process Substitution.

      An example from ksh(1) man page:

          paste <(cut -f1 file1) <(cut -f3 file2) | tee >(process1) >(process2)

  • > As my app matured, I found that I often wanted hierarchical folder-like functionality. Rather than recreating this mess in db relationships, it was easier to store the path in sqlite and work with it in Python. E.g. `os.listdir(my_folder)`

    This makes total sense and it is also "frowned upon" by people who take a too purist view of databases

    (Until it comes a time to backup, or extract files, or grow a hard drive etc and then you figure out how you shot yourself in the foot)

    • To make it more queryable, you can have different classes for dataset types with metadata like: file_format, num_files, sizes

  • > As my app matured, I found that I often wanted hierarchical folder-like functionality.

    In the process of prototyping some "remote" collaborating file systems, I always wonder whether it is a good idea maintaining a flat map from path concatenated with "/" like an S3 to the file content, in term of efficiency or elegancy.

  • > As my app matured, I found that I often wanted hierarchical folder-like functionality

    (1) Slim table "items"

    - id / parent_id / kind (0/1 file folder) integer

    - name text

    - Maybe metadata.

    (2) Separate table "content"

    - id integer

    - data blob

    There you have file-system-like structure and fast access times (don't mix content in the first table)

    Or, if you wish for deduplication or compression, add item_content (3)

I was looking at self hosted RSS readers recently. The application is single user. I don't expect it to be doing a lot of DB intensive stuff.

It surprised me that almost all required PostgreSQL, and most of those that didn't opted for something otherwise complex such as Mongo or MySQL.

SQLite, with no dependencies, would have simplified the process no end.

Depends, depends.. but just of logic:

All fs/drive access is managed by the OS. No DB systems have raw access to sectors or direct raw access to files.

Having a database file on the disc, offers a "cluster" of successive blocks on the hard drive (if it's not fragmented), resulting in relatively short moving distances of the drive head to seek the necessary sectors. There will still be the same sectors occupied, even after vast insert/write/del operations. Absolutely no change of DB file's position on hard drive. It's not a problem with SSDs, though.

So, the following apply:

client -> DB -> OS -> Filesystem

I think, you already can see the DB part is an extra layer. So, if one wouldn't have this, it would be "faster" in terms of execution time. Always.

If it's slower, then you use the not-optimal settings for your use case/filesystem.

My father did this once. He took H2 and made it even more faster :) incredible fast on Windows in direct comparison of H2/h2-modificated with same data.

So having a DBMS is convenient and made in decisions to serve certain domains and their problems. Having it is convenient, but that doesn't mean it's the most optimized way of doing it.

SQLite can be faster than FileSystem for small files. For big files, it can do more than 1 GB/s. On Pack [1], I benchmarked these speeds, and you can go very fast. It can be even 2X faster than tar [2].

In my opinion, SQLite can be faster in big reads and writes too, but the team didn't optimise it as much (like loading the whole content into memory) as maybe it was not the main use of the project. My hope is that we will see even faster speeds in the future.

[1] https://pack.ac [2] https://forum.lazarus.freepascal.org/index.php/topic,66281.m...

Let's assume that filesystems are fairly optimized pieces of software. Let's assume that the people building them heard of databases and at some point along the way considered things like the costs of open/close calls.

What is SQLite not doing that filesystems are?

How much more performance could you get by bypassing the filesystem and writing directly to the block device? Of course, you'd need to effectively implement your own filesystem, but you'd be able to optimise it more for the specific workload.

  • Oracle, some other databases did this back in a day in 00s by wrong with block devices directly.

    I am not sure if this is done anymore, because the performance gains were modest compared to the hassle of a custom formatted partition.

i.e. opening and closing many files from disk is slower than opening and closing one file and using memory.

It's important. But understandable.

Why hasn’t someone made sqlitefs yet?

  • Because SQLite not being an FS is apparently the reason why it’s fast:

    > The performance difference arises (we believe) because when working from an SQLite database, the open() and close() system calls are invoked only once, whereas open() and close() are invoked once for each blob when using blobs stored in individual files.

    • Additionally, it may also be that accesses to a single file allows the OS to efficiently retrieve (and IIRC in the case of Windows, predict) the working set allowing the reduction of access times; which is not the case if you open multiple files.

      1 reply →

  • Here you go:)

    https://github.com/narumatt/sqlitefs

    And it seems quite interesting:

    "sqlite-fs allows Linux and MacOS to mount a sqlite database file as a normal filesystem."

    "If a database file name isn't specified, sqlite-fs use in-memory-db instead of a file. All data will be deleted when the filesystem is closed."

  • Proxmox puts the VM configuration information in a SQLite database and exposes it through a FUSE file system. It even gets replicated across the cluster using their replication algorithm. It’s a bespoke implementation, but it’s a SQLite-backed filesystem.

  • I remember reading someone's comments about how instead of databases using their own data serialisation formats for persistence and then optimizing writes and read over that they should just utilize the FS directly and let all of the optimisations built by FS authors be taken advantage of.

    I wish I could find that comment, because my explanation doesn't do it justice. Very interesting idea, someone's probably going to explain how it's already been tried in some old IBM database a long time ago and failed due to whatever reason.

    I still think it should be tried with newer technologies though, sounds like a very interesting idea.

    • > instead of databases using their own data serialisation formats for persistence and then optimizing writes and read over that they should just utilize the FS directly and let all of the optimisations built by FS authors be taken advantage of.

      The original article effectively argues the opposite: if your use case matches a database, then that will be way faster. Because the filesystem is both fully general, multi-process and multi-user, it's got to be pessimistic about its concurrency.

      This is why e.g. games distribute their assets as massive blobs which are effectively filesystems - better, more predictable seek performance. Starting from the Doom WAD onwards.

      For an example of databases that use the file system, both the mbox and maildir systems for email probably count?

    • ReiserFS was built on the premise that doing a filesystem right could get us to a point where the filesystem is the database for a lot of use cases.

      It's now "somewhat possible" in that modern filesystem are overall mostly less broken about handling large number of small (or at least moderately small) files than they used to be.

      But databases are still far more optimized for handling small pieces of data in the ways we want to handle data we put into databases, which typically also includes a need to index etc.

    • As far as I can remember MongoDB did not have any dedicated block caching mechanism in its earlier releases.

      They basically mmap’ed the database file and argued that OS cache should do its job. Which makes sense but I guess it did not perform as well as any fune tuned caching mechanism.

      1 reply →

  • > Why hasn’t someone made sqlitefs yet?

    What do you expect the value proposition of something loosely described as a sqlitefs to be?

    One of the main selling points of SQLite is that you can statically link it into a binary and no one needs to maintain anything between the OS and the client application. I'm talking about things like versioning.

    What benefit would there be to replace a library with a full blown file system?

  • There's quite a number of sqlite FUSE implementations around, if you want to head in that direction.

  • No mention of how it performs when you need random access (seek) into files. Perhaps it underperforms the file system at that?

  • POSIX interfaces (open, read, write, seek, close, etc) are very challenging to implement in an efficient/reliable way.

    Using SQLite let's you tailor your data access patterns in a much more rigorous way and side step the POSIX tarpit.

  • There are at least two for macOS. But they run into trouble nowadays because FUSE wants kernel extensions.

  • Would you put sqlite in the kernel? Or using something like FUSE?

    It seems to me that all the extra indirection from using FUSE would lead to more than a 35% performance hit.

    Statically linking an sqlite into a kernel module and providing it with filesystem access seems like something non trivial to me.

> Reading is about an order of magnitude faster than writing

not a native speaker, what does it mean?

  • Order of magnitude usually refers to a 10x difference. Two orders of magnitude would be 100x difference.

    (Sometimes the phrase is casually used to just mean "a lot", but here I think they mean 10x).

  • An order of magnitude is for example from 10 to 100, or from 1,000 to 10,000, so typically an increase by 10x or similar.

* for certain operations.

Which is a bit d'oh, since being faster for some things is one of the main motivations for a database in the first place.