The Sia Foundation Roadmap

This Sia roadmap provides mid to high level insight into core Sia development. It will be updated once a quarter at minimum, and will show an outline of what we’re currently working on, why we’re working on it, and what we have in mind after that’s done.

We’re focusing on updated versions of all the core components of the Sia network - renting, hosting, and hodling. These updates will bring significant improvements to scaling, performance, and modular internal architecture to meet the needs of individuals, enterprise users, and exchange partners. Furthermore, each component will be paired with a powerful new user interface, all with the overarching goal of making core Sia software more “legible” to end users. “Legible” is the antonym of “black-box:” in a legible system, the user can easily discern why any particular decision was made, and can override those decisions as they see fit. This goal follows naturally from the Foundation’s mission of empowering users.

Project Board

The new Sia renter, replacing the current siad renter module. Drawing on what we’ve learned from siad, skyd, and us, we designed renterd from the ground up to be modular and horizontally scalable. Although the average user likely won’t notice, a renterd deployment is actually a set of interconnected services. As such, it can distribute workloads in parallel across multiple machines, and can be configured to store its metadata in any SQL backend. This flexibilty makes renterd easier to integrate with other Sia ecosystem software, such as Sia Satellite and S5, and addresses the scalability barriers that have historically hampered enterprise solutions.

May 2024

June 2024

  • Add support for RHP4
  • Improve upload parallelism

The new Sia host, replacing the current siad host module. The host module has been chronically neglected for years, and suffers from poor upload performance, data integrity issues, and a general lack of user-friendliness. hostd is our greenfield reimagining of the Sia hosting experience, bringing a sorely-needed refresh to our host community. Aside from addressing performance bottlenecks, hostd also offers superior metrics and monitoring tools, which will allow users to make informed decisions about storage allocations, contract parameters, pricing, and quality of service.

May 2024

June 2024

  • Automate data integrity checks
  • Add support for RHP4

September 2024

  • Support Postgres as an alternative to SQLite
  • Support additional volume types to optimize performance for distributed file systems

The new Sia wallet, replacing the current siad wallet module. walletd aims to be the go-to option for exchanges, miners, or the average hodler who needs a secure place to store their SC. Accordingly, it supports both hot and cold setups, including multi-sig schemes and hardware wallet integration. And like renterd and hostd, walletd comes packaged with a sleek, yet powerful UI, which can be securely accessed from any device.

Note: you don’t need walletd to be a renter or a host; renterd and hostd include their own built-in hot wallets.

January 2024

  • Support SQLite - Done

May 2024

  • Add “Indexer” mode enabling easier integration for exchanges - Done
  • Beta Release - In Progress

The new Sia explorer, replacing and going far beyond the current siad explorer capabilities. explored will serve as both a standalone blockchain explorer with a web interface, and as a library providing powerful indexing and searching capabilities to third-party explorers (e.g. SiaStats) and “lite-client” systems like narwal. explored will be developed and launched alongside the Utreexo overhaul, making it one of the world’s first Utreexo-native block explorers.

June 2024

  • Beta release

“Utreexo” is an umbrella term that refers to an upcoming radical overhaul of Sia’s consensus code. The primary change is the transition from a large, unwieldy database of UTXOs to a compact cryptographic accumulator, bringing myriad benefits to performance, scalability, and functionality. This is a big deal: it requires changing the block and transaction formats, which have been untouched since Sia’s mainnet launch back in 2014. As such, we are taking this opportunity to clean up a few other warts in Sia’s consensus code, giving the project a solid foundation for many years to come. For a more technical dive on how Utreexo works, check out Luke’s blog post here and MIT Bitcoin talk here.

June 2024

  • Reset Anagami testnet

July 2024

  • Upgrade Zen testnet

RHP4, short for renter-host protocol 4, is the next iteration of the protocol hosts and renters use to communicate. Our goal for RHP4 is to increase the parallelism of data uploads, reduce protocol overhead, and improve download performance. A new protocol is required to enable some exciting new features to be available after the Utreexo hardfork, such as early contract termination, capacity reservation, and contract renewal fund rollover. RHP4 will also enable storage consumers to download and upload data directly in the browser without installing additional software.

February 2024

  • Finalize spec

June 2024

  • Implement RHP4 client in coreutils
  • Test RHP4 on Anagami

July 2024

  • Release experimental SiaJS SDK
  • Test RHP4 on Zen

fsd is our experimental implementation of an IPFS node with specific optimizations for storing and retrieving data from the Sia network. We plan to open source it in the near future once it has stabilized more and we’ve verified that pinned content is discoverable and accessible by other IPFS nodes.

June 2024

  • Release alpha
Document version date: May 14, 2024
May 16, 2024

This PR is a testing playground for a potential directories table. The goal is to find out whether a table like that can provide the index needed to significantly speed up fetching directory info on larger nodes with 500k+ objects.

  • Create dirs when objects are created
  • Delete empty dirs when objects are deleted
  • Create/Delete dirs when objects are renamed
  • SQLite migration code
  • MySQL migration code
  • Test performance before/after

UPDATE: I've experimented with multiple approaches so far. Directory table with full paths, with names only, recursive queries, non recursive ones, functional indices, generated columns with indices, partitioning, temporary tables and nested sets. The big problem seems to be our use-case of having 2 tables which don't just require a select but a group by on the sub-directories of the directory we want to query. According to some posts I have found about nested sets they were mostly used before MySQL supported recursive queries since recursive queries should perform the same for fetching immediate children and a lot faster for deep traversal. I confirmed that fetching the directories from the tree is quite fast at 10-20ms for a query that takes about 2 seconds in total using the recursive approach. However that part is fast for pretty much all approaches I have tested. The slow part is grouping the objects by parent dir which is pretty slow for every approach.

So I ended up choosing the fastest of all approaches I've tested which also happens to be the nicest query. A directory table that only contains the name of each directory instead of full path to avoid too much overhead. It takes about 800ms on my node with 500k objects and 150k directories in the higher level dirs compared to 2s for the fully recursive approach.

It is pretty straight forward and consists of 2 parts which are combined via UNION:

  1. We search for the directory that we want to query. Then we fetch all objects in that directory with a fast inner join.
  2. We search for all the child dirs with a quick inner join. Then we join all objects with these dirs using their paths before (and this is the slow part) grouping them by directory.
May 16, 2024
May 16, 2024

This PR adds errInvalidHandshake to the list of prune alert exception cases.

{
  "id": "h:010f0c2e5180fce43e054aab2d3479bcd860f5da960c1277ebfb97999e858ec5",
  "severity": "warning",
  "message": "Contract pruning failed",
  "data": {
    "contractID": "fcid:5f69293916b6a5ddecef0d60347b3ef198ca623d8453d14e949becb2b15f884d",
    "error": "failed to prune contract fcid:5f69293916b6a5ddecef0d60347b3ef198ca623d8453d14e949becb2b15f884d; NewRenterTransport: couldn't read host's handshake: EOF\n",
    "hostKey": "ed25519:0220dabd125931d000c48ada38fb39082e0c7e241a1a469f2527a0846f692d32",
    "hostVersion": "1.6.0",
    "origin": "autopilot.autopilot"
  },
  "timestamp": "2024-05-15T11:09:34.766139268Z"
}
May 16, 2024
May 15, 2024
May 15, 2024

This PR expands test coverage in merkle to 95.1% and overall consensus coverage to 76.2% Now that we have the coverage for DecodeFrom and EncodeTo I'm planning to add a roundtrip later, assuming you find such tests to be valuable.

May 15, 2024

Adds support for a yaml configuration file similar to hostd and renterd.

directory: /etc/walletd
autoOpenWebUI: true
http:
  address: :9980
  password: sia is cool
consensus:
  network: mainnet
  gatewayAddress: :9981
  bootstrap: false
  enableUPnP: false
index:
  mode: partial # full, partial, none (full index will index the entire blockchain, partial will only index addresses that are registered in the wallet, none will treat the database as read-only and not index any new data)
  batchSize: 64 # max number of blocks to index at a time (increasing this will increase scan speed, but also increase memory and cpu usage)
log:
  level: info # global log level
  stdout:
    enabled: true # enable logging to stdout
    level: debug # override the global log level for stdout
    enableANSI: false
    format: human # human or JSON
  file:
    enabled: true # enable logging to a file
    level: debug # override the global log level for the file
    path: /var/log/walletd.log
    format: json # human or JSON
May 15, 2024

Testing the migration on my node I found that we have to tweak the upload penalty in some cases.

May 15, 2024

This PR expands test coverage in policy.go to 86.5% and overall types coverage to 59.7%

May 14, 2024

Changes the read and write RPC to send internal errors and alert+log the actual error.

View full activity feed →