foundationdb VS .NET Runtime

Actually, Apple does this for iCloud! They use FoundationDB[1] to store billions of databases, one for each user (plus shared or global databases).

See: https://read.engineerscodex.com/p/how-apple-built-icloud-to-...

Discussed on HN at the time: https://news.ycombinator.com/item?id=39028672

[1]: https://github.com/apple/foundationdb https://en.wikipedia.org/wiki/FoundationDB

Is this the language in question? https://github.com/apple/foundationdb/blob/main/flow/README....

> I wondered if perhaps this WASM use case for a cross-language library was already just as possible and ergonomic using language bindings and maybe thats why this use case doesnt seem like a big deal to people.

Yeah that’s the reason. You don’t notice it a lot of the time, but FFIs are everywhere already. The most common foreign function interface is basically the ability to call C code, or have functions made available to C code. C is used because everyone knows it and it’s simple. And most languages either compile to native code (eg rust) - which makes linking to C code easy. Or the runtime is implemented in C or C++ (eg V8, Ruby). In languages like that, the standard library is already basically implemented via a FFI to C/C++ code.

I’ve got an iOS app I’m working on that’s half rust and half swift, with a touch of C in the middle. The bindings work great - the whole thing links together into one binary, even with link time optimizations. But the glue code is gross, and when I want to fiddle with the rust to Swift API I need to change my code in about 4 different places.

Most FFIs are a one to many relationship in that if you write a clean C API, you can probably write bindings in every language. But you don’t actually want to call naked C code from Ruby or Javascript. Good bindings will make you forget everything is done via ffi. Eg numpy. I haven’t looked at the wasm component model proposal - I assume it’s trying to make this process cleaner, which sounds lovely.

I maintain the nodejs bindings for foundationdb. Foundationdb bindings are all done via ffi linking to their C code. And the API is complex, using promises and things. I find it really interesting browsing their official bindings to go, Java, Python and Ruby. Same bindings. Same wrapped api. Same team of authors. Just different languages. And that’s enough to make the wrapper wildly different in every language. From memory the Java ffi wrapper is 4x as much code as it is in Ruby.

https://github.com/apple/foundationdb/tree/main/bindings

Isn’t apple server on C/C++?

They have their own db in that

https://github.com/apple/foundationdb

> Any meaningful technology is open source.

Clearly untrue, however FoundationDB is open source, with a permissive license.

https://github.com/apple/foundationdb

So is much of the operational tooling for it:

https://github.com/FoundationDB/fdb-kubernetes-operator

I don't recall any of those details but the test involved injecting a bogus block device that always returns garbage, and noting that this results in garbage records returned from client queries. And I don't think those kinds of issues have been eradicated, browsing through their github issues there are people trying to recover corrupted clusters. https://github.com/apple/foundationdb/issues/2480

For my current job, I have to install FoundationDB (they have a github repo) as a dependency for the project's code.

To clarify, FoundationDB has some documented multi-region features, but it's not at all clear that anybody who runs it at scale relies on those multi-region features [1]. Even if they do, it's not obvious they run this Fearless DR mode at any appreciable latency.

Geo-replication is an area where Anna really shines, and foundation pays severe costs. All of the concurrency control problems have contention footprints on the order of round-trips. Fully optimistic concurrency control needs to expose backoffs quite a few round-trips to be live. Even pessimistic concurrency control requires some number of round trips, probably at least 1.5 optimally, but in practice in most tuned systems, probably 3. A heck of a lot of use cases make sense at 0 global RTTs and don't at 1+. The ability to tell the database how to manage concurrency, and then providing causal is ultimately the best you can do. That's awesome.

At the end of the day, I have to believe that there's a whole big mess of applications we'll build this, on systems in the portion of the design space Anna is choosing, one day. This is only recently not cutting edge research, but it's definitely still research. We don't know how to model these concepts at a high-level and in a composable way where the masses of software developers can engage with them.

It's interesting to think about how long ago we were graced with BAYOU [2], that thing was ahead of its time. I suspect it's going to take a little while longer before these sorts of techniques make their way into data storage primitives we think of as part of the standard vernacular, but I believe we'll get there eventually.

Databases are fun!

[1] https://github.com/apple/foundationdb/wiki/Multi-Region-Repl...

It's an interesting "time is a circle" problem given that a century only has 100 years and then we loop around again. 2-digit years is convenient for people in many situations but they are very lossy, and horrible for machines.

It reminds me of this breaking change to .Net from last year.[1][2] Maybe AA just needs to update .Net which would pad them out until the 2050's when someone born in the 1950s would be having...exactly the same problem in the article. (It is configurable now so you could just keep pushing it each decade, until it wraps again).

Or they could use 4-digit years.

[1] https://github.com/dotnet/runtime/issues/75148

These can also be passed as arguments to `dotnet publish` if necessary.

Reference:

- https://learn.microsoft.com/en-us/dotnet/core/deploying/nati...

- https://github.com/dotnet/runtime/blob/main/src/coreclr/nati...

- https://github.com/dotnet/runtime/blob/5b4e770daa190ce69f402... (full list of recognized keys for IlcInstructionSet)

Yes, that is true. I'm not sure about JVM implementation details but the reason the comment says "virtual and interface" calls is to outline the difference. Virtual calls in .NET are sufficiently close[0] to virtual calls in C++. Interface calls, however, are coded differently[1].

Also you are correct - virtual calls are not terribly expensive, but they encroach on ever limited* CPU resources like indirect jump and load predictors and, as noted in parent comments, block inlining, which is highly undesirable for small and frequently called methods, particularly when they are in a loop.

* through great effort of our industry to take back whatever performance wins each generation brings with even more abstractions that fail to improve our productivity

[0] https://github.com/dotnet/coreclr/blob/4895a06c/src/vm/amd64...

[1] https://github.com/dotnet/runtime/blob/main/docs/design/core... (mind you, the text was initially written 18 ago, wow)

If you care about portable SIMD and performance, you may want to save yourself trouble and skip to C# instead, it also has an extensive guide to using it: https://github.com/dotnet/runtime/blob/69110bfdcf5590db1d32c...

CoreLib and many new libraries are using it heavily to match performance of manually intensified C++ code.

DEBUG: packageFiles with updates (repository=local) "config": { "nuget": [ { "deps": [ { "datasource": "nuget", "depType": "nuget", "depName": "Microsoft.Extensions.Hosting", "currentValue": "7.0.0", "updates": [ { "bucket": "non-major", "newVersion": "7.0.1", "newValue": "7.0.1", "releaseTimestamp": "2023-02-14T13:21:52.713Z", "newMajor": 7, "newMinor": 0, "updateType": "patch", "branchName": "renovate/dotnet-monorepo" }, { "bucket": "major", "newVersion": "8.0.0", "newValue": "8.0.0", "releaseTimestamp": "2023-11-14T13:23:17.653Z", "newMajor": 8, "newMinor": 0, "updateType": "major", "branchName": "renovate/major-dotnet-monorepo" } ], "packageName": "Microsoft.Extensions.Hosting", "versioning": "nuget", "warnings": [], "sourceUrl": "https://github.com/dotnet/runtime", "registryUrl": "https://api.nuget.org/v3/index.json", "homepage": "https://dot.net/", "currentVersion": "7.0.0", "isSingleVersion": true, "fixedVersion": "7.0.0" } ], "packageFile": "RenovateDemo.csproj" } ] }

https://github.com/dotnet/runtime/issues/59591

Support zstd Content-Encoding:

We might not be that far away already. There is this issue[1] on Github, where Microsoft and the community discuss some significant changes.

There is still a lot of questions unanswered, but initial tests look promising.

Ref: https://github.com/dotnet/runtime/issues/94620

https://github.com/dotnet/dotnet exists for source build that stitches together SDK, Roslyn, runtime and other dependencies. A lot of them can be built and used individually, which is what contributors usually do. For example, you can clone and build https://github.com/dotnet/runtime and use the produced artifacts to execute .NET assemblies or build .NET binaries.

Yeah, it kind of is. There are quite a few of experiments that are conducted to see if they show promise in the prototype form and then are taken further for proper integration if they do.

Unfortunately, object stack allocation was not one of them even though DOTNET_JitObjectStackAllocation configuration knob exists today, enabling it makes zero impact as it almost never kicks in. By the end of the experiment[0], it was concluded that before investing effort in this kind of feature becomes profitable given how a lot of C# code is written, there are many other lower hanging fruits.

To contrast this, in continuation to green threads experiment, a runtime handled tasks experiment[1] which moves async state machine handling from IL emitted by Roslyn to special-cased methods and then handling purely in runtime code has been a massive success and is now being worked on to be integrated in one of the future version of .NET (hopefully 10?)

[0] https://github.com/dotnet/runtime/issues/11192

[1] https://github.com/dotnet/runtimelab/blob/feature/async2-exp...