Full Width [alt+shift+f] Shortcuts [alt+shift+k]
Sign Up [alt+shift+s] Log In [alt+shift+l]

Improve your reading experience

Logged in users get linked directly to articles resulting in a better reading experience. Please login for free, it takes less than 1 minute.

More from exist

When Imperfect Systems are Good, Actually: Bluesky's Lossy Timelines

Often when designing systems, we aim for perfection in things like consistency of data, availability, latency, and more. The hardest part of system design is that it’s difficult (if not impossible) to design systems that have perfect consistency, perfect availability, incredibly low latency, and incredibly high throughput, all at the same time. Instead, when we approach system design, it’s best to treat each of these properties as points on different axes that we balance to find the “right fit” for the application we’re supporting. I recently made some major tradeoffs in the design of Bluesky’s Following Feed/Timeline to improve the performance of writes at the cost of consistency in a way that doesn’t negatively affect users but reduced P99s by over 96%. Timeline Fanout When you make a post on Bluesky, your post is indexed by our systems and persisted to a database where we can fetch it to hydrate and serve in API responses. Additionally, a reference to your post is “fanned out” to your followers so they can see it in their Timelines. This process involves looking up all of your followers, then inserting a new row into each of their Timeline tables in reverse chronological order with a reference to your post. When a user loads their Timeline, we fetch a page of post references and then hydrate the posts/actors concurrently to quickly build an API response and let them see the latest content from people they follow. The Timelines table is sharded by user. This means each user gets their own Timeline partition, randomly distributed among shards of our horizontally scalable database (ScyllaDB), replicated across multiple shards for high availability. Timelines are regularly trimmed when written to, keeping them near a target length and dropping older post references to conserve space. Hot Shards in Your Area Bluesky currently has around 32 Million Users and our Timelines database is broken into hundreds of shards. To support millions of partitions on such a small number of shards, each user’s Timeline partition is colocated with tens of thousands of other users’ Timelines. Under normal circumstances with all users behaving well, this doesn’t present a problem as the work of an individual Timeline is small enough that a shard can handle the work of tens of thousands of them without being heavily taxed. Unfortunately, with a large number of users, some of them will do abnormal things like… well… following hundreds of thousands of other users. Generally, this can be dealt with via policy and moderation to prevent abusive users from causing outsized load on systems, but these processes take time and can be imperfect. When a user follows hundreds of thousands of others, their Timeline becomes hyperactive with writes and trimming occurring at massively elevated rates. This load slows down the individual operations to the user’s Timeline, which is fine for the bad behaving user, but causes problems to the tens of thousands of other users sharing a shard with them. We typically call this situation a “Hot Shard”: where some resident of a shard has “hot” data that is being written to or read from at much higher rates than others. Since the data on the shard is only replicated a few times, we can’t effectively leverage the horizontal scale of our database to process all this additional work. Instead, the “Hot Shard” ends up spending so much time doing work for a single partition that operations to the colocated partitions slow down as well. Stacking Latencies Returning to our Fanout process, let’s consider the case of Fanout for a user followed by 2,000,000 other users. Under normal circumstances, writing to a single Timeline takes an average of ~600 microseconds. If we sequentially write to the Timelines of our user’s followers, we’ll be sitting around for 20 minutes at best to Fanout this post. If instead we concurrently Fanout to 1,000 Timelines at once, we can complete this Fanout job in ~1.2 seconds. That sounds great, except it oversimplifies an important property of systems: tail latencies. The average latency of a write is ~600 microseconds, but some writes take much less time and some take much more. In fact, the P99 latency of writes to the Timelines cluster can be as high as 15 milliseconds! What does this mean for our Fanout? Well, if we concurrently write to 1,000 Timelines at once, statistically we’ll see 10 writes as slow as or slower than 15 milliseconds. In the case of timelines, each “page” of followers is 10,000 users large and each “page” must be fanned out before we fetch the next page. This means that our slowest writes will hold up the fetching and Fanout of the next page. How does this affect our expected Fanout time? Each “page” will have ~100 writes as slow as or slower than the P99 latency. If we get unlucky, they could all stack up on a single routine and end up slowing down a single page of Fanout to 1.5 seconds. In the worst case, for our 2,000,000 Follower celebrity, their post Fanout could end up taking as long as 5 minutes! That’s not even considering P99.9 and P99.99 latencies which could end up being >1 second, which could leave us waiting tens of minutes for our Fanout job. Now imagine how bad this would be for a user with 20,000,000+ Followers! So, how do we fix the problem? By embracing imperfection, of course! Lossy Timelines Imagine a user who follows hundreds of thousands of others. Their Timeline is being written to hundreds of times a second, moving so fast it would be humanly impossible to keep up with the entirety of their Timeline even if it was their full-time job. For a given user, there’s a threshold beyond which it is unreasonable for them to be able to keep up with their Timeline. Beyond this point, they likely consume content through various other feeds and do not primarily use their Following Feed. Additionally, beyond this point, it is reasonable for us to not necessarily have a perfect chronology of everything posted by the many thousands of users they follow, but provide enough content that the Timeline always has something new. Note in this case I’m using the term “reasonable” to loosely convey that as a social media service, there must be a limit to the amount of work we are expected to do for a single user. What if we introduce a mechanism to reduce the correctness of a Timeline such that there is a limit to the amount of work a single Timeline can place on a DB shard. We can assert a reasonable limit for the number of follows a user should have to have a healthy and active Timeline, then increase the “lossiness” of their Timeline the further past that limit they go. A loss_factor can be defined as min(reasonable_limit/num_follows, 1) and can be used to probabilistically drop writes to a Timeline to prevent hot shards. Just before writing a page in Fanout, we can generate a random float between 0 and 1, then compare it to the loss_factor of each user in the page. If the user’s loss_factor is smaller than the generated float, we filter the user out of the page and don’t write to their Timeline. Now, users all have the same number of “follows worth” of Fanout. For example with a reasonable_limit of 2,000, a user who follows 4,000 others will have a loss_factor of 0.5 meaning half the writes to their Timeline will get dropped. For a user following 8,000 others, their loss factor of 0.25 will drop 75% of writes to their Timeline. Thus, each user has a effective ceiling on the amount of Fanout work done for their Timeline. By specifying the limits of reasonable user behavior and embracing imperfection for users who go beyond it, we can continue to provide service that meets the expectations of users without sacrificing scalability of the system. Aside on Caching We write to Timelines at a rate of more than one million times a second during the busy parts of the day. Looking up the number of follows of a given user before fanning out to them would require more than one million additional reads per second to our primary database cluster. This additional load would not be well received by our database and the additional cost wouldn’t be worth the payoff for faster Timeline Fanout. Instead, we implemented an approach that caches high-follow accounts in a Redis sorted set, then each instance of our Fanout service loads an updated version of the set into memory every 30 seconds. This allows us to perform lookups of follow counts for high-follow accounts millions of times per second per Fanount service instance. By caching values which don’t need to be perfect to function correctly in this case, we can once again embrace imperfection in the system to improve performance and scalability without compromising the function of the service. Results We implemented Lossy Timelines a few weeks ago on our production systems and saw a dramatic reduction in hot shards on the Timelines database clusters. In fact, there now appear to be no hot shards in the cluster at all, and the P99 of a page of Fanout work has been reduced by over 90%. Additionally, with the reduction in write P99s, the P99 duration for a full post Fanout has been reduced by over 96%. Jobs that used to take 5-10 minutes for large accounts now take <10 seconds. Knowing where it’s okay to be imperfect lets you trade consistency for other desirable aspects of your systems and scale ever higher. There are plenty of other places for improvement in our Timelines architecture, but this step was a big one towards improving throughput and scalability of Bluesky’s Timelines. If you’re interested in these sorts of problems and would like to help us build the core data services that power Bluesky, check out this job listing. If you’re interested in other open positions at Bluesky, you can find them here.

5 months ago 44 votes
Jetstream: Shrinking the AT Proto Firehose by >99%

Bluesky recently saw a massive spike in activity in response to Brazil’s ban of Twitter. As a result, the AT Proto event firehose provided by Bluesky’s Relay at bsky.network has increased in volume by a huge amount. The average event rate during this surge increased by ~1,300%. Before this new surge in activity, the firehose would produce around 24 GB/day of traffic. After the surge, this volume jumped to over 232 GB/day! Keeping up with the full, verified firehose quickly became less practical on cheap cloud infrastructure with metered bandwidth. To help reduce the burden of operating bots, feed generators, labelers, and other non-verifying AT Proto services, I built Jetstream as an alternative, lightweight, filterable JSON firehose for AT Proto. How the Firehose Works The AT Proto firehose is a mechanism used to keep verified, fully synced copies of the repos of all users. Since repos are represented as Merkle Search Trees, each firehose event contains an update to the user’s MST which includes all the changed blocks (nodes in the path from the root to the modified leaf). The root of this path is signed by the repo owner, and a consumer can keep their copy of the repo’s MST up-to-date by applying the diff in the event. For a more in-depth explanation of how Merkle Trees are constructed, check out this explainer. Practically, this means that for every small JSON record added to a repo, we also send along some number of MST blocks (which are content-addressed hashes and thus very information-dense) that are mostly useful for consumers attempting to keep a fully synced, verified copy of the repo. You can think of this as the difference between cloning a git repo v.s. just grabbing the latest version of the files without the .git folder. In this case, the firehose effectively streams the diffs for the repository with commits, signatures, and metadata, which is inherently heavier than a point-in-time checkout of the repo. Because firehose events with repo updates are signed by the repo owner, they allow a consumer to process events from any operator without having to trust the messenger. This is the “Authenticated” part of the Authenticated Transfer (AT) Protocol and is crucial to the correct functioning of the network. That being said, of the hundreds of consumers of Bluesky’s production Relay, >90% of them are building feeds, bots, and other tools that don’t keep full copies of the entire network and don’t verify MST operations at all. For these consumers, all they actually process is the JSON records created, updated, and deleted in each event. If consumers already trust the provider to do validation on their end, they could get by with a much more lightweight data stream. How Jetstream Works Jetstream is a streaming service that consumes an AT Proto com.atproto.sync.subscribeRepos stream and converts it into lightweight, friendly JSON. If you want to try it out yourself, you can connect to my public Jetstream instance and view all posts on Bluesky in realtime: $ websocat "wss://jetstream2.us-east.bsky.network/subscribe?wantedCollections=app.bsky.feed.post" Note: the above instance is operated by Bluesky PBC and is free to use, more instances are listed in the official repo Readme Jetstream converts the CBOR-encoded MST blocks produced by the AT Proto firehose and translates them into JSON objects that are easier to interface with using standard tooling available in programming languages. Since Repo MSTs only contain records in their leaf nodes, this means Jetstream can drop all of the blocks in an event except for those of the leaf nodes, typically leaving only one block per event. In reality, this means that Jetstream’s JSON firehose is nearly 1/10 the size of the full protocol firehose for the same events, but lacks the verifiability and signatures included in the protocol-level firehose. Jetstream events end up looking something like: { "did": "did:plc:eygmaihciaxprqvxpfvl6flk", "time_us": 1725911162329308, "type": "com", "commit": { "rev": "3l3qo2vutsw2b", "type": "c", "collection": "app.bsky.feed.like", "rkey": "3l3qo2vuowo2b", "record": { "$type": "app.bsky.feed.like", "createdAt": "2024-09-09T19:46:02.102Z", "subject": { "cid": "bafyreidc6sydkkbchcyg62v77wbhzvb2mvytlmsychqgwf2xojjtirmzj4", "uri": "at://did:plc:wa7b35aakoll7hugkrjtf3xf/app.bsky.feed.post/3l3pte3p2e325" } }, "cid": "bafyreidwaivazkwu67xztlmuobx35hs2lnfh3kolmgfmucldvhd3sgzcqi" } } Each event lets you know the DID of the repo it applies to, when it was seen by Jetstream (a time-based cursor), and up to one updated repo record as serialized JSON. Check out this 10 second CPU profile of Jetstream serving 200k evt/sec to a local consumer: By dropping the MST and verification overhead by consuming from relay we trust, we’ve reduced the size of a firehose of all events on the network from 232 GB/day to ~41GB/day, but we can do better. Jetstream and zstd I recently read a great engineering blog from Discord about their use of zstd to compress websocket traffic to/from their Gateway service and client applications. Since Jetstream emits marshalled JSON through the websocket for developer-friendliness, I figured it might be a neat idea to see if we could get further bandwidth reduction by employing zstd to compress events we send to consumers. zstd has two basic operating modes, “simple” mode and “streaming” mode. Streaming Compression At first glance, streaming mode seems like it’d be a great fit. We’ve got a websocket connection with a consumer and streaming mode allows the compression to get more efficient over the lifetime of the connection. I went and implemented a streaming compression version of Jetstream where a consumer can request compression when connecting and will get zstd compressed JSON sent as binary messages over the socket instead of plaintext. Unfortunately, this had a massive impact on Jetstream’s server-side CPU utilization. We were effectively compressing every message once per consumer as part of their streaming session. This was not a scalable approach to offering compression on Jetstream. Additionally, Jetstream stores a buffer of the past 24 hours (configurable) of events on disk in PebbleDB to allow consumers to replay events before getting transitioned into live-tailing mode. Jetstream stores serialized JSON in the DB, so playback is just shuffling the bytes into the websocket without having to round-trip the data into a Go struct. When we layer in streaming compression, playback becomes significantly more expensive because we have to compress outgoing events on-the-fly for a consumer that’s catching up. In real numbers, this increased CPU usage of Jetstream by 23% while lowering the throughput of playback from ~200k evt/sec to ~28k evt/sec for a single local consumer. When in streaming mode, we can’t leverage the bytes we compress for one consumer and reuse them for another consumer because zstd’s streaming context window may not be in sync between the two consumers. They haven’t received exactly the same data in the session so the clients on the other end don’t have their state machines in the same state. Since streaming mode’s primary advantage is giving us eventually better efficiency as the encoder learns about the data, what if we just taught the encoder about the data at the start and compress each message statelessly? Dictionary Mode zstd offers a mechanism for initializing an encoder/decoder with pre-optimized settings by providing a dictionary trained on a sample of the data you’ll be encoding/decoding. Using this dictionary, zstd essentially uses it’s smallest encoded representations for the most frequently seen patterns in the sample data. In our case, where we’re compressing serialized JSON with a common event shape and lots of common property names, training a dictionary on a large number of real events should allow us to represent the common elements among messages in the smallest number of bytes. For take two of Jetstream with zstd, let’s to use a single encoder for the whole service that utilizes a custom dictionary trained on 100,000 real events. We can use this encoder to compress every event as we see it, before persisting and emitting it to consumers. Now we end up with two copies of every event, one that’s just serialized JSON, and one that’s statelessly compressed to zstd using our dictionary. Any consumers that want compression can have a copy of the dictionary on their end to initialize a decoder, then when we broadcast the shared compressed event, all consumers can read it without any state or context issues. This requires the consumers and server to have a pre-shared dictionary, which is a major drawback of this implementation but good enough for our purposes. That leaves the problem of event playback for compression-enabled clients. An easy solution here is to just store the compressed events as well! Since we’re only sticking the JSON records into our PebbleDB, the actual size of the 24 hour playback window is <8GB with sstable compression. If we store a copy of the JSON serialized event and a copy of the zstd compressed event, this will, at most, double our storage requirements. Then during playback, if the consumer requests compression, we can just shuffle bytes out of the compressed version of the DB into their socket instead of having to move it through a zstd encoder. Savings Running with a custom dictionary, I was able to get the average Jetstream event down from 482 bytes to just 211 bytes (~0.44 compression ratio). Jetstream allows us to live tail all posts on Bluesky as they’re posted for as little as ~850 MB/day, and we could keep up with all events moving through the firehose during the Brazil Twitter Exodus weekend for 18GB/day (down from 232GB/day). With this scheme, Jetstream is required to compress each event only once before persisting it to disk and emitting it to connected consumers. The CPU impact of these changes is significant in proportion to Jetstream’s incredibly light load but it’s a flat cost we pay once no matter how many consumers we have. (CPU profile from a 30 second pprof sample with 12 consumers live-tailing Jetstream) Additionally, with Jetstream’s shared buffer broadcast architecture, we keep memory allocations incredibly low and the cost per consumer on CPU and RAM is trivial. In the allocation profile below, more than 80% of the allocations are used to consume the full protocol firehose. The total resident memory of Jetstream sits below 16MB, 25% of which is actually consumed by the new zstd dictionary. To bring it all home, here’s a screenshot from the dashboard of my public Jetstream instance serving 12 consumers all with various filters and compression settings, running on a $5/mo OVH VPS. At our new baseline firehose activity, a consumer of the protocol-level firehose would require downloading ~3.16TB/mo to keep up. A Jetstream consumer getting all created, updated, and deleted records without compression enabled would require downloading ~400GB/mo to keep up. A Jetstream consumer that only cares about posts and has zstd compression enabled can get by on as little as ~25.5GB/mo, <99% of the full weight firehose. Feel free to join the conversation about Jetstream and zstd on Bluesky.

10 months ago 27 votes
How HLS Works

Over the past few weeks, I’ve been building out server-side short video support for Bluesky. The major aim of this feature is to support short (90 second max) video streaming at a quality that doesn’t cost an arm and a leg for us to provide for free. In order to stay within these constraints, we’re considering making use of a video CDN that can bear the brunt of the bandwidth required to support Video-on-Demand streaming. While the CDN is a pretty fully-featured product, we want to avoid too much vendor lock-in and provide some enhancements to our streaming platform that requires extending their offering and getting creative with video streaming protocols. Some of the things we’d like to be able to do that don’t work out-of-the-box are: Track view counts, viewer sessions, and duration viewed to provide better feedback for video performance. Provide dynamic closed-caption support with the flexibility to automate them in the future. Store a transcoded version of source files somewhere durable to provide a “source of truth” for videos when needed. Append a “trailer” to the end of video streams for some branding in a TikTok-esque 3-second snippet. In this post I’ll be focusing on the HLS-related features above, namely view/duration accounting, closed captions, and trailers. HLS is Just a Bunch of Text files HTTP Live Streaming (HLS) is a standard established by Apple in 2009 that allows for adaptive-bitrate live and Video-on-Demand (VOD) streaming. For the purposes of this blog post, I’ll restrict my explanations to how HLS VOD streaming works. A player that implements the HLS protocol is capable of dynamically adjusting the quality of a streamed video based on network conditions. Additionally, a server that implements the HLS protocol should provide one or more variants of a media stream which accommodate varying network qualities to allow for graceful degradation of stream quality without stopping playback. HLS implements this by producing a series of plaintext (.m3u8) “playlist” files that tell the player what bitrates and resolutions the server provides so that the player can decide which variant it should stream. HLS differentiates between two kinds of “playlist” files: Master Playlists, and Media Playlists. Master Playlists A Master Playlist is the first file fetched by your video player. It contains a series of variants which point to child Media Playlists. It also describes the approximate bitrate of the variant sources and the codecs and resolutions used by those sources. $ curl https://my.video.host.com/video_15/playlist.m3u8 #EXTM3U #EXT-X-VERSION:3 #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=688540,CODECS="avc1.64001e,mp4a.40.2",RESOLUTION=640x360 360p/video.m3u8 #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=1921217,CODECS="avc1.64001f,mp4a.40.2",RESOLUTION=1280x720 720p/video.m3u8 In the above file, the key things to notice are the RESOLUTION parameters and the {res}/video.m3u8 links. Your media player will generally start with the lowest resolution version before jumping up to higher resolutions once the network speed between you and the server is dialed in. The links in this file are pointers to Media Playlists, generally as relative paths from the Master Playlist such that, if we wanted to grab the 720p Media Playlist, we’d navigate to: https://my.video.host.com/video_15/720p/video.m3u8. A Master Playlist can also contain multi-track audio directives and directives for closed-captions but for now let’s move onto the Media Playlist. Media Playlists A Media Playlist is yet another plaintext file that provides your video player with two key bits of data: a list of media Segments (encoded as .ts video files) and headers for each Segment that tell the player the runtime of the media. $ curl https://my.video.host.com/video_15/720p/video.m3u8 #EXTM3U #EXT-X-VERSION:3 #EXT-X-PLAYLIST-TYPE:VOD #EXT-X-MEDIA-SEQUENCE:0 #EXT-X-TARGETDURATION:4 #EXTINF:4.000, video0.ts #EXTINF:4.000, video1.ts #EXTINF:4.000, video2.ts #EXTINF:4.000, video3.ts #EXTINF:4.000, video4.ts #EXTINF:2.800, video5.ts This Media Playlist describes a video that’s 22.8 seconds long (5 x 4-second Segments + 1 x 2.8-second Segment). The playlist describes a VOD piece of media, meaning we know this playlist contains the entirety of the media the player needs. The TARGETDURATION tells us the maximum length of each Segment so the player knows how many Segments to buffer ahead of time. During live streaming, that also lets the player know how frequently to refresh the playlist file to discover new Segments. Finally the EXTINF headers for each Segment indicate the duration of the following .ts Segment file and the relative paths of the video#.ts tell the player where to load the actual media files from. Where’s the Actual Media? At this point, the video player has loaded two .m3u8 playlist files and got lots of metadata about how to play the video but it hasn’t actually loaded any media files. The .ts files referenced in the Media Playlist are where the real media is, so if we wanted to control the playlists but let the CDN handle serving actual media, we can just redirect those video#.ts requests to our CDN. .ts files are Transport Stream MPEG-2 encoded short media files that can contain video or audio and video. Tracking Views To track views of our HLS streams, we can leverage the fact that every video player must first load the Master Playlist. When a user requests the Master Playlist, we can modify the results dynamically to provide a SessionID to each response and allow us to track the user session without cookies or headers: #EXTM3U #EXT-X-VERSION:3 #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=688540,CODECS="avc1.64001e,mp4a.40.2",RESOLUTION=640x360 360p/video.m3u8?session_id=12345 #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=1921217,CODECS="avc1.64001f,mp4a.40.2",RESOLUTION=1280x720 720p/video.m3u8?session_id=12345 Now when their video player fetches the Media Playlists, it’ll include a query-string that we can use to identify the streaming session, ensuring we don’t double-count views on the video and can track which Segments of video were loaded in the session. #EXTM3U #EXT-X-VERSION:3 #EXT-X-PLAYLIST-TYPE:VOD #EXT-X-MEDIA-SEQUENCE:0 #EXT-X-TARGETDURATION:4 #EXTINF:4.000, video0.ts?session_id=12345&duration=4 #EXTINF:4.000, video1.ts?session_id=12345&duration=4 #EXTINF:4.000, video2.ts?session_id=12345&duration=4 #EXTINF:4.000, video3.ts?session_id=12345&duration=4 #EXTINF:4.000, video4.ts?session_id=12345&duration=4 #EXTINF:2.800, video5.ts?session_id=12345&duration=2.8 Finally when the video player fetches the media Segment files, we can measure the Segment view before we redirect to our CDN with a 302, allowing us to know the amount of video-seconds loaded in the session and which Segments were loaded. This method has limitations, namely that a media player loading a segment doesn’t necessarily mean it showed that segment to the viewer, but it’s the best we can do without an instrumented media player. Adding Subtitles Subtitles are included in the Master Playlist as a variant and then are referenced in each of the video variants to let the player know where to load subs from. #EXTM3U #EXT-X-VERSION:3 #EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="subs",NAME="en_subtitle",DEFAULT=NO,AUTOSELECT=yes,LANGUAGE="en",FORCED="no",CHARACTERISTICS="public.accessibility.transcribes-spoken-dialog",URI="subtitles/en.m3u8" #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=688540,CODECS="avc1.64001e,mp4a.40.2",RESOLUTION=640x360,SUBTITLES="subs" 360p/video.m3u8 #EXT-X-STREAM-INF:PROGRAM-ID=0,BANDWIDTH=1921217,CODECS="avc1.64001f,mp4a.40.2",RESOLUTION=1280x720,SUBTITLES="subs" 720p/video.m3u8 Just like with the video Media Playlists, we need a Media Playlist file for the subtitle track as well so that the player knows where to load the source files from and what duration of the stream they cover. $ curl https://my.video.host.com/video_15/subtitles/en.m3u8 #EXTM3U #EXT-X-VERSION:3 #EXT-X-MEDIA-SEQUENCE:0 #EXT-X-TARGETDURATION:22.8 #EXTINF:22.800, en.vtt In this case, since we’re only serving a short video, we can just provide a single Segment that points at a WebVTT subtitle file encompassing the entire duration of the video. If you crack open the en.vtt file you’ll see something like: $ curl https://my.video.host.com/video_15/subtitles/en.vtt WEBVTT 00:00.000 --> 00:02.000 According to all known laws of aviation, 00:02.000 --> 00:04.000 there is no way a bee should be able to fly. 00:04.000 --> 00:06.000 Its wings are too small to get its fat little body off the ground. ... The media player is capable of reading WebVTT and presenting the subtitles at the right time to the viewer. For longer videos you may want to break up your VTT files into more Segments and update the subtitle Media Playlist accordingly. To provide multiple languages and versions of subtitles, just add more EXT-X-MEDIA:TYPE=SUBTITLES lines to the Master Playlist and tweak the NAME, LANGUAGE (if different), and URI of the additional subtitle variant definitions. #EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="subs",NAME="en_subtitle",DEFAULT=NO,AUTOSELECT=yes,LANGUAGE="en",FORCED="no",CHARACTERISTICS="public.accessibility.transcribes-spoken-dialog",URI="subtitles/en.m3u8" #EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="subs",NAME="fr_subtitle",DEFAULT=NO,AUTOSELECT=yes,LANGUAGE="fr",FORCED="no",CHARACTERISTICS="public.accessibility.transcribes-spoken-dialog",URI="subtitles/fr.m3u8" #EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="subs",NAME="ja_subtitle",DEFAULT=NO,AUTOSELECT=yes,LANGUAGE="ja",FORCED="no",CHARACTERISTICS="public.accessibility.transcribes-spoken-dialog",URI="subtitles/ja.m3u8" Appending a Trailer For branding purposes (and in other applications, for advertising purposes), it can be helpful to insert Segments of video into a playlist to change the content of the video without requiring the content be appended to and re-encoded with the source file. Thankfully, HLS allows us to easily insert Segments into the Media Playlist using this one neat trick: #EXTM3U #EXT-X-VERSION:3 #EXT-X-PLAYLIST-TYPE:VOD #EXT-X-MEDIA-SEQUENCE:0 #EXT-X-TARGETDURATION:4 #EXTINF:4.000, video0.ts #EXTINF:4.000, video1.ts #EXTINF:4.000, video2.ts #EXTINF:4.000, video3.ts #EXTINF:4.000, video4.ts #EXTINF:2.800, video5.ts #EXT-X-DISCONTINUITY #EXTINF:3.337, trailer0.ts #EXTINF:1.201, trailer1.ts #EXTINF:1.301, trailer2.ts #EXT-X-ENDLIST In this Media Playlist we use HLS’s EXT-X-DISCONTINUITY header to let the video player know that the following Segments may be in a different bitrate, resolution, and aspect-ratio than the preceding content. Once we’ve provided the discontinuity header, we can add more Segments just like normal that point at a different media source broken up into .ts files. Remember, HLS allows us to use relative or absolute paths here, so we could provide a full URL for these trailer#.ts files, or virtually route them so they can retain the path context of the currently viewed video. Note that we don’t need to provide the discontinuity header here, and we could also name the trailer files something like video{6-8}.ts if we wanted to, but for clarity and proper player behavior, it’s best to use the discontinuity header if your trailer content doesn’t match the bitrate and resolution of the other video Segments. When the video player goes to play this media, it will continue from video5.ts to trailer0.ts without missing a beat, making it appear as if the trailer is part of the original video. This approach allows us to dynamically change the contents of the trailer for all videos, heavily cache the trailer .ts Segment files for performance, and avoid having to encode the trailer onto the end of every video source file. Conclusion At the end of the day, we’ve now got a video streaming service capable of tracking views and watch session durations, dynamic closed caption support, and branded trailers to help grow the platform. HLS is not a terribly complex protocol. The vast majority of it is human-readable plaintext files and is easy to inspect in the wild to how it’s used in production. When I started this project, I knew next to nothing about the protocol but was able to download some .m3u8 files and get digging to discover how the protocol worked, then build my own implementation of a HLS server to accommodate the video streaming needs of Bluesky. To learn more about HLS, you can check out the official RFC here which describes all the features discussed above and more. I hope this post encourages you to go explore other protocols you use every day by poking at them in the wild, downloading the files your browser interprets for you, and figuring out how simple some of these apparently “complex” systems are. If you’re interested in solving problems like these, take a look at our open Job Recs. If you have any questions about HLS, Bluesky, or other distributed, @scale social media infrastructure, you can find me on Bluesky here and you can discuss this post here.

a year ago 24 votes
An entire Social Network in 1.6GB (GraphD Part 2)

In Part 1 of this series, we tried to answer the question “who do you follow who also follows user B” in Bluesky, a social network with millions of users and hundreds of millions of follow relationships. At the conclusion of the post, we’d developed an in-memory graph store for the network that uses HashMaps and HashSets to keep track of the followers of every user and the set of users they follow, allowing bidirectional lookups, intersections, unions, and other set operations for combining social graph data. I received some helpful feedback after that post where several people pointed me towards Roaring Bitmaps as a potential improvement on my implementation. They were right, Roaring Bitmaps would be an excellent fit for my Graph service, GraphD, and could also provide me with a much needed way to quickly persist and load the Graph data to and from disk on startup, hopefully reducing the startup time of the service. What are Bitmaps? If you just want to dive into the Roaring Bitmap spec, you can read the paper here, but it might be easier to first talk about bitmaps in general. You can think of a bitmap as a vector of one-bit values (like booleans) that let you encode a set of integer values. For instance, say we have 10,000 users on our website and want to keep track of which users have validated their email addresses. We could do this by creating a list of the uint32 user IDs of each user, in which case if all 10,000 users have validated their emails we’re storing 10k * 32 bits = 40KB. Or, we could create a vector of single-bit values that’s 10,000 bits long (10k / 8 = 1.25KB), then if a user has confirmed their email we can set the value at the index of their UID to 1. If we want to create a list of all the UIDs of validated accounts, we can walk the vector and record the index of each non-zero bit. If we want to check if user n has validated their email, we can do a O(1) lookup in the bitmap by loading the bit at index n and checking if it’s set. When Bitmaps get Big and Sparse Now when talking about our social network problem, we’re dealing with a few more than 10,000 UIDs. We need to keep track of 5.5M users and whether or not the user follows or is followed by any of the other 5.5M users in the network. To keep a bitmap of “People who follow User A”, we’re going to need 5.5M bits which would require (5.5M / 8) ~687KB of space. If we wanted to keep bitmaps of “People who follow User A” and “People who User A follows”, we’d need ~1.37MB of space per user using a simple bitmap, meaning we’d need 5,500,000 * 1.37MB = ~7.5 Terabytes of space! Clearly this isn’t an improvement of our strategy from Part 1, so how can we make this more efficient? One strategy for compressing the bitmap is to take consecutive runs of 0’s or 1’s (i.e. 00001110000001) in the bitmap and turn them into a number. For instance if we had an account that followed only the last 100 accounts in our social network, the first 5,499,900 indices in our bitmap would be 0’s and so we could represent the bitmap by saying: 5,499,900 0's, then 100 1's which you notice I’ve written here in a lot fewer than 687KB and a computer could encode using two uint32 values plus two bits (one indicator bit for the state of each run) for a total of 66 bits. This strategy is called Run Length Encoding (RLE) and works pretty well but has a few drawbacks: mainly if your data is randomly and heavily populated, you may not have many consecutive runs (imagine a bitset where every odd bit is set and every even bit is unset). Also lookups and evaluation of the bitset requires walking the whole bitset to figure out where the index you care about lives in the compressed format. Thankfully there’s a more clever way to compress bitmaps using a strategy called Roaring Bitmaps. A brief description of the storage strategy for Roaring Bitmaps from the official paper is as follows: We partition the range of 32-bit indexes ([0, n)) into chunks of 2^16 integers sharing the same 16 most significant digits. We use specialized containers to store their 16 least significant bits. When a chunk contains no more than 4096 integers, we use a sorted array of packed 16-bit integers. When there are more than 4096 integers, we use a 2^16-bit bitmap. Thus, we have two types of containers: an array container for sparse chunks and a bitmap container for dense chunks. The 4096 threshold insures that at the level of the containers, each integer uses no more than 16 bits. These bitmaps are designed to support both densely and sparsely distributed data and can provide high performance binary set operations (and/or/etc.) operating on the containers within two or more bitsets in parallel. For more info on how Roaring Bitmaps work and some neat diagrams, check out this excellent primer on Roaring Bitmaps by Vikram Oberoi. So, how does this help us build a better graph? GraphD, Revisited with Roaring Bitmaps Let’s get back to our GraphD Service, this time in Go instead of Rust. For each user we can keep track of a struct with two bitmaps: type FollowMap struct { followingBM *roaring.Bitmap followingLk sync.RWMutex followersBM *roaring.Bitmap followersLk sync.RWMutex } Our FollowMap gives us a Roaring Bitmap for both the set of users we follow, and the set of users who follow us. Adding a Follow to the graph just requires we set the right bits in both user’s respective maps: // Note I've removed locking code and error checks for brevity func (g *Graph) addFollow(actorUID, targetUID uint32) { actorMap, _ := g.g.Load(actorUID) actorMap.followingBM.Add(targetUID) targetMap, _ := g.g.Load(targetUID) targetMap.followersBM.Add(actorUID) } Even better if we want to compute the intersections of two sets (i.e. the people User A follows who also follow User B) we can do so in parallel: // Note I've removed locking code and error checks for brevity func (g *Graph) IntersectFollowingAndFollowers(actorUID, targetUID uint32) ([]uint32, error) { actorMap, ok := g.g.Load(actorUID) targetMap, ok := g.g.Load(targetUID) intersectMap := roaring.ParAnd(4, actorMap.followingBM, targetMap.followersBM) return intersectMap.ToArray(), nil } Storing the entire graph as Roaring Bitmaps in-memory costs us around 6.5GB of RAM and allows us to perform set intersections between moderately large sets (with hundreds of thousands of set bits) in under 500 microseconds while serving over 70k req/sec! And the best part of all? We can use Roaring’s serialization format to write these bitmaps to disk or transfer them over the network. Storing 164M Follows in 1.6GB In the original version of GraphD, on startup the service would read a CSV file with an adjacency list of the (ActorDID, TargetDID) pairs of all follows on the network. This required creating a CSV dump of the follows table, pausing writes to the follows table, then bringing up the service and waiting 5 minutes for it to read the CSV file, intern the DIDs as uint32 UIDs, and construct the in-memory graph. This process is slow, pauses writes for 5 minutes, and every time our service restarts we have to do it all over again! With Roaring Bitmaps, we’re now given an easy way to effectively serialize a version of the in-memory graph that is many times smaller than the adjacency list CSV and many times faster to load. We can serialize the entire graph into a SQLite DB on the local machine where each row in a table contains: (uid, DID, followers_bitmap, following_bitmap) Loading the entire graph from this SQLite DB can be done in around ~20 seconds: // Note I've removed locking code and error checks for brevity rows, err := g.db.Query(`SELECT uid, did, following, followers FROM actors;`) for rows.Next() { var uid uint32 var did string var followingBytes []byte var followersBytes []byte rows.Scan(&uid, &did, &followingBytes, &followersBytes) followingBM := roaring.NewBitmap() followingBM.FromBuffer(followingBytes) followersBM := roaring.NewBitmap() followersBM.FromBuffer(followersBytes) followMap := &FollowMap{ followingBM: followingBM, followersBM: followersBM, followingLk: sync.RWMutex{}, followersLk: sync.RWMutex{}, } g.g.Store(uid, followMap) g.setUID(did, uid) g.setDID(uid, did) } While the service is running, we can also keep track of the UIDs of actors who have added or removed a follow since the last time we saved the DB, allowing us to periodically flush changes to the on-disk SQLite only for bitmaps that have updated. Syncing our data every 5 seconds while tailing the production firehose takes 2ms and writes an average of only ~5MB to disk per flush. The crazy part of this is, the on-disk representation of our entire follow network is only ~1.6GB! Because we’re making use of Roaring’s compressed serialized format, we can turn the ~6.5GB of in-memory maps into 1.6GB of on-disk data. Our largest bitmap, the followers of the bsky.app account with over 876k members, becomes ~500KB as a blob stored in SQLite. So, to wrap up our exploration of Roaring Bitmaps for first-degree graph databases, we saw: A ~20% reduction in resident memory size compared to HashSets and HashMaps A ~84% reduction in the on-disk size of the graph compared to an adjacency list A ~93% reduction in startup time compared to loading from an adjacency list A ~66% increase in throughput of worst-case requests under load A ~59% reduction in p99 latency of worst-case requests under low My next iteration on this problem will likely be to make use of DGraph’s in-memory Serialized Roaring Bitmap library that allows you to operate on fully-compressed bitmaps so there’s no need to serialize and deserialize them when reading from or writing to disk. It also probably results in significant memory savings as well! If you’re interested in solving problems like these, take a look at our open Backend Developer Job Rec. You can find me on Bluesky here, you can chat about this post here.

a year ago 26 votes

More in programming

Executives should be the least busy people

If your executive calendar is packed back to back, you have no room for fires, customers, or serendipities. You've traded all your availability for efficiency. That's a bad deal. Executives of old used to know this! That's what the long lunches, early escapes to the golf course, and reading the paper at work were all about. A great fictional example of this is Bert Cooper from Mad Men. He knew his value was largely in his network. He didn't have to grind every minute of every day to prove otherwise. His function was to leap into action when the critical occasion arose or decision needed to be made. But modern executives are so insecure about seeming busy 24/7 that they'll wreck their business trying to prove it. Trying to outwork everyone. Sacrificing themselves thin so they can run a squirrel-brain operation that's constantly chasing every nutty idea. Now someone is inevitably going to say "but what about Elon!!". He's actually a perfect illustration of doing this right, actually. Even if he works 100-hour weeks, he's the CEO of 3 companies, has a Diablo VI addiction, and keeps a busy tweeting schedule too. In all of that, I'd be surprised if there was more than 20-30h per company per week on average. And your boss is not Elon. Wide open calendars should not be seen as lazy, but as intentional availability. It's time we brought them back into vogue.

2 days ago 4 votes
Dispatch 012: Local-first talks, Automerge 3, and Scribbling on a Google Calendar

A secret master plan, the official launch of Automerge 3, and an update on Sketchy Calendars

2 days ago 2 votes
React Server Components with Vite and React-Router (tip)

Create a small example app and send payloads from the server to the client using RSC's

3 days ago 9 votes
2000 words about arrays and tables

I'm way too discombobulated from getting next month's release of Logic for Programmers ready, so I'm pulling a idea from the slush pile. Basically I wanted to come up with a mental model of arrays as a concept that explained APL-style multidimensional arrays and tables but also why there weren't multitables. So, arrays. In all languages they are basically the same: they map a sequence of numbers (I'll use 1..N)1 to homogeneous values (values of a single type). This is in contrast to the other two foundational types, associative arrays (which map an arbitrary type to homogeneous values) and structs (which map a fixed set of keys to heterogeneous values). Arrays appear in PLs earlier than the other two, possibly because they have the simplest implementation and the most obvious application to scientific computing. The OG FORTRAN had arrays. I'm interested in two structural extensions to arrays. The first, found in languages like nushell and frameworks like Pandas, is the table. Tables have string keys like a struct and indexes like an array. Each row is a struct, so you can get "all values in this column" or "all values for this row". They're heavily used in databases and data science. The other extension is the N-dimensional array, mostly seen in APLs like Dyalog and J. Think of this like arrays-of-arrays(-of-arrays), except all arrays at the same depth have the same length. So [[1,2,3],[4]] is not a 2D array, but [[1,2,3],[4,5,6]] is. This means that N-arrays can be queried on any axis. ]x =: i. 3 3 0 1 2 3 4 5 6 7 8 0 { x NB. first row 0 1 2 0 {"1 x NB. first column 0 3 6 So, I've had some ideas on a conceptual model of arrays that explains all of these variations and possibly predicts new variations. I wrote up my notes and did the bare minimum of editing and polishing. Somehow it ended up being 2000 words. 1-dimensional arrays A one-dimensional array is a function over 1..N for some N. To be clear this is math functions, not programming functions. Programming functions take values of a type and perform computations on them. Math functions take values of a fixed set and return values of another set. So the array [a, b, c, d] can be represented by the function (1 -> a ++ 2 -> b ++ 3 -> c ++ 4 -> d). Let's write the set of all four element character arrays as 1..4 -> char. 1..4 is the function's domain. The set of all character arrays is the empty array + the functions with domain 1..1 + the functions with domain 1..2 + ... Let's call this set Array[Char]. Our compilers can enforce that a type belongs to Array[Char], but some operations care about the more specific type, like matrix multiplication. This is either checked with the runtime type or, in exotic enough languages, with static dependent types. (This is actually how TLA+ does things: the basic collection types are functions and sets, and a function with domain 1..N is a sequence.) 2-dimensional arrays Now take the 3x4 matrix i. 3 4 0 1 2 3 4 5 6 7 8 9 10 11 There are two equally valid ways to represent the array function: A function that takes a row and a column and returns the value at that index, so it would look like f(r: 1..3, c: 1..4) -> Int. A function that takes a row and returns that column as an array, aka another function: f(r: 1..3) -> g(c: 1..4) -> Int.2 Man, (2) looks a lot like currying! In Haskell, functions can only have one parameter. If you write (+) 6 10, (+) 6 first returns a new function f y = y + 6, and then applies f 10 to get 16. So (+) has the type signature Int -> Int -> Int: it's a function that takes an Int and returns a function of type Int -> Int.3 Similarly, our 2D array can be represented as an array function that returns array functions: it has type 1..3 -> 1..4 -> Int, meaning it takes a row index and returns 1..4 -> Int, aka a single array. (This differs from conventional array-of-arrays because it forces all of the subarrays to have the same domain, aka the same length. If we wanted to permit ragged arrays, we would instead have the type 1..3 -> Array[Int].) Why is this useful? A couple of reasons. First of all, we can apply function transformations to arrays, like "combinators". For example, we can flip any function of type a -> b -> c into a function of type b -> a -> c. So given a function that takes rows and returns columns, we can produce one that takes columns and returns rows. That's just a matrix transposition! Second, we can extend this to any number of dimensions: a three-dimensional array is one with type 1..M -> 1..N -> 1..O -> V. We can still use function transformations to rearrange the array along any ordering of axes. Speaking of dimensions: What are dimensions, anyway Okay, so now imagine we have a Row × Col grid of pixels, where each pixel is a struct of type Pixel(R: int, G: int, B: int). So the array is Row -> Col -> Pixel But we can also represent the Pixel struct with a function: Pixel(R: 0, G: 0, B: 255) is the function where f(R) = 0, f(G) = 0, f(B) = 255, making it a function of type {R, G, B} -> Int. So the array is actually the function Row -> Col -> {R, G, B} -> Int And then we can rearrange the parameters of the function like this: {R, G, B} -> Row -> Col -> Int Even though the set {R, G, B} is not of form 1..N, this clearly has a real meaning: f[R] is the function mapping each coordinate to that coordinate's red value. What about Row -> {R, G, B} -> Col -> Int? That's for each row, the 3 × Col array mapping each color to that row's intensities. Really any finite set can be a "dimension". Recording the monitor over a span of time? Frame -> Row -> Col -> Color -> Int. Recording a bunch of computers over some time? Computer -> Frame -> Row …. This is pretty common in constraint satisfaction! Like if you're conference trying to assign talks to talk slots, your array might be type (Day, Time, Room) -> Talk, where Day/Time/Room are enumerations. An implementation constraint is that most programming languages only allow integer indexes, so we have to replace Rooms and Colors with numerical enumerations over the set. As long as the set is finite, this is always possible, and for struct-functions, we can always choose the indexing on the lexicographic ordering of the keys. But we lose type safety. Why tables are different One more example: Day -> Hour -> Airport(name: str, flights: int, revenue: USD). Can we turn the struct into a dimension like before? In this case, no. We were able to make Color an axis because we could turn Pixel into a Color -> Int function, and we could only do that because all of the fields of the struct had the same type. This time, the fields are different types. So we can't convert {name, flights, revenue} into an axis. 4 One thing we can do is convert it to three separate functions: airport: Day -> Hour -> Str flights: Day -> Hour -> Int revenue: Day -> Hour -> USD But we want to keep all of the data in one place. That's where tables come in: an array-of-structs is isomorphic to a struct-of-arrays: AirportColumns( airport: Day -> Hour -> Str, flights: Day -> Hour -> Int, revenue: Day -> Hour -> USD, ) The table is a sort of both representations simultaneously. If this was a pandas dataframe, df["airport"] would get the airport column, while df.loc[day1] would get the first day's data. I don't think many table implementations support more than one axis dimension but there's no reason they couldn't. These are also possible transforms: Hour -> NamesAreHard( airport: Day -> Str, flights: Day -> Int, revenue: Day -> USD, ) Day -> Whatever( airport: Hour -> Str, flights: Hour -> Int, revenue: Hour -> USD, ) In my mental model, the heterogeneous struct acts as a "block" in the array. We can't remove it, we can only push an index into the fields or pull a shared column out. But there's no way to convert a heterogeneous table into an array. Actually there is a terrible way Most languages have unions or product types that let us say "this is a string OR integer". So we can make our airport data Day -> Hour -> AirportKey -> Int | Str | USD. Heck, might as well just say it's Day -> Hour -> AirportKey -> Any. But would anybody really be mad enough to use that in practice? Oh wait J does exactly that. J has an opaque datatype called a "box". A "table" is a function Dim1 -> Dim2 -> Box. You can see some examples of what that looks like here Misc Thoughts and Questions The heterogeneity barrier seems like it explains why we don't see multiple axes of table columns, while we do see multiple axes of array dimensions. But is that actually why? Is there a system out there that does have multiple columnar axes? The array x = [[a, b, a], [b, b, b]] has type 1..2 -> 1..3 -> {a, b}. Can we rearrange it to 1..2 -> {a, b} -> 1..3? No. But we can rearrange it to 1..2 -> {a, b} -> PowerSet(1..3), which maps rows and characters to columns with that character. [(a -> {1, 3} ++ b -> {2}), (a -> {} ++ b -> {1, 2, 3}]. We can also transform Row -> PowerSet(Col) into Row -> Col -> Bool, aka a boolean matrix. This makes sense to me as both forms are means of representing directed graphs. Are other function combinators useful for thinking about arrays? Does this model cover pivot tables? Can we extend it to relational data with multiple tables? Systems Distributed Talk (will be) Online The premier will be August 6 at 12 CST, here! I'll be there to answer questions / mock my own performance / generally make a fool of myself. Sacrilege! But it turns out in this context, it's easier to use 1-indexing than 0-indexing. In the years since I wrote that article I've settled on "each indexing choice matches different kinds of mathematical work", so mathematicians and computer scientists are best served by being able to choose their index. But software engineers need consistency, and 0-indexing is overall a net better consistency pick. ↩ This is right-associative: a -> b -> c means a -> (b -> c), not (a -> b) -> c. (1..3 -> 1..4) -> Int would be the associative array that maps length-3 arrays to integers. ↩ Technically it has type Num a => a -> a -> a, since (+) works on floats too. ↩ Notice that if each Airport had a unique name, we could pull it out into AirportName -> Airport(flights, revenue), but we still are stuck with two different values. ↩

3 days ago 8 votes
Our $100M Series B

We don’t want to bury the lede: we have raised a $100M Series B, led by a new strategic partner in USIT with participation from all existing Oxide investors. To put that number in perspective: over the nearly six year lifetime of the company, we have raised $89M; our $100M Series B more than doubles our total capital raised to date — and positions us to make Oxide the generational company that we have always aspired it to be. If this aspiration seems heady now, it seemed absolutely outlandish when we were first raising venture capital in 2019. Our thesis was that cloud computing was the future of all computing; that running on-premises would remain (or become!) strategically important for many; that the entire stack — hardware and software — needed to be rethought from first principles to serve this market; and that a large, durable, public company could be built by whomever pulled it off. This scope wasn’t immediately clear to all potential investors, some of whom seemed to latch on to one aspect or another without understanding the whole. Their objections were revealing: "We know you can build this," began more than one venture capitalist (at which we bit our tongue; were we not properly explaining what we intended to build?!), "but we don’t think that there is a market." Entrepreneurs must become accustomed to rejection, but this flavor was particularly frustrating because it was exactly backwards: we felt that there was in fact substantial technical risk in the enormity of the task we put before ourselves — but we also knew that if we could build it (a huge if!) there was a huge market, desperate for cloud computing on-premises. Fortunately, in Eclipse Ventures we found investors who saw what we saw: that the most important products come when we co-design hardware and software together, and that the on-premises market was sick of being told that they either don’t exist or that they don’t deserve modernity. These bold investors — like the customers we sought to serve — had been waiting for this company to come along; we raised seed capital, and started building. And build it we did, making good on our initial technical vision: We did our own board designs, allowing for essential system foundation like a true hardware root-of-trust and end-to-end power observability. We did our own microcontroller operating system, and used it to replace the traditional BMC. We did our own platform enablement software, eliminating the traditional UEFI BIOS and its accompanying flotilla of vulnerabilities. We did our own host hypervisor, assuring an integrated and seamless user experience — and eliminating the need for a third-party hypervisor and its concomitant rapacious software licensing. We did our own switch — and our own switch runtime — eliminating entire universes of integration complexity and operational nightmares. We did our own integrated storage service, allowing the rack-scale system to have reliable, available, durable, elastic instance storage without necessitating a dependency on a third party. We did our own control plane, a sophisticated distributed system building on the foundation of our hardware and software components to deliver the API-driven services that modernity demands: elastic compute, virtual networking, and virtual storage. While these technological components are each very important (and each is in service to specific customer problems when deploying infrastructure on-premises), the objective is the product, not its parts. The journey to a product was long, but we ticked off the milestones. We got the boards brought up. We got the switch transiting packets. We got the control plane working. We got the rack manufactured. We passed FCC compliance. And finally, two years ago, we shipped our first system! Shortly thereafter, more milestones of the variety you can only get after shipping: our first update of the software in the field; our first update-delivered performance improvements; our first customer-requested features added as part of an update. Later that year, we hit general commercial availability, and things started accelerating. We had more customers — and our first multi-rack customer. We had customers go on the record about why they had selected Oxide — and customers describing the wins that they had seen deploying Oxide. Customers starting landing faster now: enterprise sales cycles are infamously long, but we were finding that we were going from first conversations to a delivered product surprisingly quickly. The quickening pace always seemed to be due in some way to our transparency: new customers were listeners to our podcast, or they had read our RFDs, or they had perused our documentation, or they had looked at the source code itself. With growing customer enthusiasm, we were increasingly getting questions about what it would look like to buy a large number of Oxide racks. Could we manufacture them? Could we support them? Could we make them easy to operate together? Into this excitement, a new potential investor, USIT, got to know us. They asked terrific questions, and we found a shared disposition towards building lasting value and doing it the right way. We learned more about them, too, and especially USIT’s founder, Thomas Tull. The more we each learned about the other, the more there was to like. And importantly, USIT had the vision for us that we had for ourselves: that there was a big, important market here — and that it was uniquely served by Oxide. We are elated to announce this new, exciting phase of the company. It’s not necessarily in our nature to celebrate fundraising, but this is a big milestone, because it will allow us to address our customers' most pressing questions around scale (manufacturing scale, system scale, operations scale) and roadmap scope. We have always believed in our mission, but this raise gives us a new sense of confidence when we say it: we’re going to kick butt, have fun, not cheat (of course!), love our customers — and change computing forever.

3 days ago 11 votes