SomniaThe gaming industry is one of the largest sectors in entertainment, and recent trends suggest that the growth of the metaverse and blockchain industries—capable of functioning as the underlying laws of digital entertainment—has become inevitable.
However, existing blockchain-based games and metaverse projects have faced limitations due to low scalability and the inherent constraints of on-chain implementation. Overcoming these barriers requires infrastructure that offers exceptional scalability along with interoperability and composability to enable a true digital world.
Somnia aims to become a “dream computer” for building a fully on-chain world. To maximize scalability, it introduces technologies such as multistream consensus, accelerated sequential execution, IceDB, and advanced compression methods.
Based on this foundation, Somnia is constructing a vibrant application ecosystem and aspires to evolve beyond social, gaming, and metaverse use cases into a blockchain infrastructure for entertainment as a whole.
You are now entering a dream world through a dream computer. In this world of dreams, you are freed from the constraints of reality, and anything you imagine becomes possible.
Throughout human history, entertainment has gone beyond mere leisure activity to become a form of cultural expression deeply intertwined with human nature. From ancient Greek theater and gladiator games in the Roman Colosseum to modern films and video games, people across eras and civilizations have instinctively sought to share stories and emotions through various mediums. In this context, entertainment has served as a means to express and realize human dreams and imagination. In other words, entertainment acts as a bridge between dream and reality—stimulating creativity and imagination.
Somnia, the subject of this article, is a blockchain for entertainment. In other words, it is a dream computer. Built on a foundation of high scalability, Somnia enables the full on-chain implementation of various forms of entertainment. This article will explore 1) gaming, which is Somnia’s current focus within entertainment, and 2) the metaverse, which represents its long-term vision. It will also delve into the technical foundations that make Somnia a network optimized for a fully on-chain world.
Source: Gartner (Aug. 2022), Four Pillars
The metaverse refers to a virtual world where users—typically represented by avatars—can interact with one another. Because the term is loosely defined, some even include social media platforms like Twitter and Instagram under a broad definition of the metaverse. However, it generally refers to interaction within a 3D virtual space. This goes beyond simple social gatherings or gaming and encompasses a range of fields including work, education, manufacturing, healthcare, and commerce.
While the global spotlight is now on AI, especially since the release of ChatGPT 3.5 in November 2022, it was the metaverse that stood out as one of the hottest tech trends around 2021. Several examples illustrate how prominent the metaverse had become at that time:
Ready Player One, a film directed by Steven Spielberg and set in a metaverse, was released in 2018 with a budget of $155–175 million and grossed $607.9 million at the global box office, ranking 12th worldwide that year.
Fortnite hosted a virtual concert by Marshmello in 2019 that drew in 10.7 million players, and another in 2020 by Travis Scott that attracted 12.3 million players.
Numerous metaverse applications experienced explosive growth. Roblox, which had 5 million DAUs in 2019, surpassed 56 million DAUs by 2022. Zepeto, launched in 2018, reached 20 million MAUs in 2022.
Crypto-based metaverse projects such as The Sandbox and Decentraland saw massive price surges in their tokens and virtual land NFTs. For instance, SAND tokens, which were trading around $0.04 in January 2021, hit a local high of $0.80 in April, and soared to $7.5 in November following Facebook’s rebranding to Meta. At its peak, even the smallest plots of land in The Sandbox traded for around $15,000.
At the time, major institutions and experts released ambitious forecasts estimating the metaverse’s potential market size in the trillions of dollars. McKinsey predicted that global consumer and business spending on the metaverse could reach as much as $5 trillion by 2030. Similarly, Goldman Sachs, Morgan Stanley, and Citi Group projected the metaverse market would grow to between $8 and $13 trillion by that year.
While today public interest in the metaverse has cooled, one might ask, why was the metaverse once seen as a trend powerful enough to reshape the world?
COVID-19 Pandemic: The most critical catalyst for the metaverse boom was the outbreak of COVID-19 starting in 2019. Although responses varied by country, lockdowns and social distancing left many people unable to participate in offline activities. This accelerated the shift to a fully digital era. Remote meeting services like Zoom and Google Meet surged, and metaverse applications that allowed interaction in virtual worlds also rapidly gained traction.
Meta Rebranding: In October 2021, Facebook—one of the world’s largest social media companies—rebranded itself as Meta, pledging major investments in metaverse-related software and hardware development.
Crypto Bull Market: Following unprecedented monetary easing by the U.S. government during the pandemic, the crypto market experienced rapid growth. As blockchain allows peer-to-peer transactions without intermediaries in a digital environment, its synergy with the metaverse narrative sparked major interest.
Until now, the focus has mostly been on consumer-facing metaverse applications centered on interaction and entertainment. However, the metaverse also holds immense potential in B2B industries such as work, education, healthcare, and manufacturing. McKinsey identified e-commerce as the primary driver of value creation in the metaverse, with additional value expected in areas like remote education and virtual advertising. According to PwC, industries such as manufacturing and construction could use digital twins in the metaverse to improve design and optimize operations. In retail, virtual stores could enhance product placement and customer flows. Many companies are already using the metaverse for employee training, collaboration, and improving customer experiences.
Source: Reddit (ventas15)
Unfortunately, the rosy future of the metaverse that everyone anticipated around 2021 has yet to materialize. The cooling of overheated interest in the metaverse stemmed from a combination of factors.
From a market perspective, rising macroeconomic uncertainty and interest rates since 2022 have dampened investment in risk assets. Many companies that had once made large bets on the metaverse began reconsidering their strategies due to profitability concerns. In early 2023, both Disney and Microsoft either shut down or scaled back their metaverse divisions, and even Meta—arguably the most committed to the metaverse—reportedly began slowing its investment in the space.
The downturn in the crypto market also played a role. As blockchain, cryptocurrency, and NFTs were often associated with the metaverse due to their technical synergies, the shift toward monetary tightening led to a rapid contraction in the crypto market, dampening investor sentiment toward the metaverse. Furthermore, starting in 2023, explosive success in the AI industry—spurred by OpenAI's ChatGPT—shifted investor attention toward AI rather than the metaverse.
On the infrastructure side, the metaverse also faced limitations. Among its many components, the widespread adoption of VR/AR devices—which play a central role—progressed more slowly than expected, preventing the user base from growing significantly. In 2022, global shipments of AR/VR headsets declined by 20.9% year-over-year, largely due to a lack of killer hardware and waning post-pandemic consumer interest.
From a technological perspective, metaverse platforms and applications have failed to meet public expectations in terms of graphics quality and immersion. Additionally, technical limitations—such as poor interoperability across platforms—led to fragmented user experiences. For example, users could not carry digital assets purchased on one platform into another, making for a less cohesive and less appealing experience.
Source: newzoo
While the metaverse has yet to realize its full potential due to several practical challenges, there is one similar form of entertainment that continues to thrive—gaming. Contrary to popular perception, the gaming industry is the largest among all entertainment sectors.
In 2023, the gaming industry generated a staggering $183.9 billion in revenue, far surpassing the $29.6 billion and $32.3 billion earned by the music and film industries respectively in 2024. With an average annual growth rate of 3.1% between 2022 and 2027, gaming continues to demonstrate strong potential as society moves further into the digital age.
Interestingly, a wave of metaverse-style games has recently emerged and is growing rapidly. Key examples include Roblox, inZOI, the GTA series, and Minecraft. The metaverse gaming market was estimated at $47.7 billion in 2024 and is projected to reach $67.9 billion in 2025, with a CAGR of approximately 42%.
This growth has been enabled by advances in hardware and network technology, which have significantly improved graphics quality and made it possible to manage mass user interactions on a single server. Given these trends, it seems clear that the metaverse was not just a passing fad. Rather, it has the potential to evolve quickly alongside technological progress in both hardware and software, especially when seen as an extension of the gaming industry.
Source: Coinbase
When the metaverse concept was receiving significant attention, Coinbase and its CEO Brian Armstrong published an article on their perspective, arguing that blockchain technology would become the core protocol of the metaverse. Unlike the physical world, which exists as a single reality, the digital world will consist of many different metaverses. There are already multiple metaverse-like games such as Roblox, Zepeto, the GTA series, and more recently inZOI by Krafton. This trend is expected to accelerate.
Just as the physical world is governed by the laws of physics, gaming and metaverse environments will also need a universal protocol that can act as their governing layer. Blockchain functions as a decentralized database, enabling secure transactions in the digital realm without requiring trust between participants. This foundational technology allows users to prove ownership and transact safely in a fragmented digital world regardless of nationality, wealth, or the platform they use. It plays a critical role in enabling sovereignty in digital environments.
So what kind of blockchain-based gaming or metaverse projects have we seen so far? Are they still active? If not, why did they fail to remain relevant?
The most prominent projects around 2021 and 2022 included:
The Sandbox: A blockchain-based metaverse project emphasizing user-generated content and play-to-earn mechanics. Users could buy virtual land and create content for others to play. The content creation process resembles Minecraft, while the gameplay experience is similar to Roblox.
Source: The Sandbox
Decentraland: Like The Sandbox, Decentraland emphasizes user-generated content but opts for smooth 3D graphics instead of voxel-based design. It focuses on avatar-driven interactions in immersive 3D environments.
Source: Daily genius
CryptoVoxels: CryptoVoxels also allows users to build environments based on land NFTs, similar to The Sandbox and Decentraland. However, it emphasizes easy browser-based construction and multimedia integration.
Source: supplied
Star Atlas: One of the most anticipated gaming projects in the early Solana ecosystem. It aims to deliver a wide range of experiences within a vast sci-fi universe.
Source: Star Atlas
These projects once led the charge for blockchain-based metaverse and play-to-earn gaming. However, each quickly lost momentum in the market for different reasons. The price of a single plot of land in The Sandbox, once worth around $15,000, now trades for around $150 to $200. Decentraland land NFTs, which averaged $10,000 in 2021, are now worth around $100. The governance and utility tokens of Star Atlas, POLIS and ATLAS, have both fallen over 99.5 percent from their all-time highs, even though the game itself has yet to launch.
What went wrong? The issues can be divided into two categories: gameplay and blockchain architecture. From a gameplay perspective, these projects had slower update cycles and lower quality compared to Web2 games. The graphics in The Sandbox, Decentraland, and CryptoVoxels were poor by modern standards. High-spec games like Star Atlas suffered from painfully slow development. Technical issues like poor optimization and undifferentiated gameplay made it hard to attract and retain users.
On the blockchain side, these projects mostly used blockchain for settling token and NFT trades, rather than leveraging it more broadly. This was due to scalability limitations that prevented full on-chain execution of game logic. Even though all of these were supposedly blockchain-based metaverse games, they failed to fully exploit composability and interoperability. In the end, they lacked both the quality of Web2 games and the unique strengths of blockchain. When the crypto market turned bearish, they quickly faded from attention.
In the past, many in the blockchain space believed in the “Fat Protocol Thesis” which argued that most value would accrue to the infrastructure layer due to cryptoeconomic security. But as many infrastructure projects have failed to capture value and popular applications have succeeded in doing so, the “Fat App Thesis” has gained traction.
Source: USV
In reality, the question of which comes first—apps or infrastructure—is like the chicken or the egg. Historically, they have always evolved together in a complementary way. From my perspective, the blockchain industry is still lagging far behind Web2 in both areas. For the industry to reach the next stage, we need a breakthrough in at least one of them—apps or infrastructure.
Returning to the topic of gaming and the metaverse, these areas have already achieved some degree of product-market fit in Web2, as seen in Roblox, Fortnite, Zepeto, and inZOI. However, Web3 gaming and metaverse projects have already been dismissed once by the market. To regain attention, they must address the shortcomings discussed earlier. That means Web3 games must offer gameplay and graphics on par with Web2. In addition, they must deliver experiences that only blockchain can enable, such as composability and interoperability. In short, we still need better infrastructure.
Source: Improbable
Founded in 2012 in London by students from the University of Cambridge, Improbable set out with the goal of developing large-scale simulation and metaverse platforms that enable massive numbers of users to interact simultaneously within shared virtual spaces.
In 2015, the company gained attention after raising about 20 million dollars in seed funding from well-known venture capital firms including a16z and Temasek. In 2017, Improbable raised a massive 502 million dollars in funding from SoftBank and a16z, entering unicorn status. In 2020, Improbable expanded its business toward the metaverse using its existing technology, and by 2023, it had repositioned itself as a venture builder.
A venture builder is a model in which a company moves beyond a single product and incubates various new businesses in next-generation fields such as the metaverse, Web3, and AI. These ventures are either spun out as independent legal entities or developed in partnership with external startups and enterprises. The approach includes value-adds across strategy, technology, brand, public relations, human resources, finance, operations, and legal support. In 2023 and 2024, Improbable successfully diversified its portfolio by launching or supporting entities like MSquared, Somnia, the Virtual Society Foundation (VSF), as well as projects like Jitter, Chamber, Kallikor, and Kosmopop.
A notable example is MSquared’s collaboration with Yuga Labs on Otherside, the metaverse project associated with the Bored Ape Yacht Club. A test play video released by Yuga Labs demonstrated thousands of users interacting in real time within a shared virtual environment.
Source: Bogdanux Youtube
Here are MSquared’s core products:
Morpheus Platform: Enables developers to build interactive virtual worlds that support tens of thousands of concurrent users using Unreal Engine 5. The platform allows real-time updates to virtual environments and includes an event management UI for easily controlling various elements of the metaverse.
Web Worlds: An open-source project that lets users create browser-accessible 3D worlds quickly. These can be built using MML Objects, which are introduced below. Users can enter the metaverse directly from their browser without installing any additional programs.
MML Objects: MML Objects are a markup language designed for the metaverse, allowing developers to easily describe and program objects and interactions within the virtual world.
Source: Somnia
While MSquared aims to help developers easily build metaverses, Somnia focuses on building a highly scalable blockchain that enables not just gaming and metaverse applications but the full implementation of a fully onchain world on blockchain infrastructure.
At this point, after discussing gaming and the metaverse, a new concept has appeared—the fully onchain world, or FOW. What exactly does this term mean?
FOW refers to an environment where all logic, data, state changes, and in some cases even media and interactions are recorded and executed entirely on the blockchain. While traditional Web2 or partial Web3 applications may record only essential logic on-chain and handle the rest off-chain, FOW seeks to handle as much as possible directly on the blockchain. This enables complete decentralization, immutability, and transparency.
Because all the rules governing the virtual world are implemented through blockchain and smart contracts, anyone can verify them. Users can interact securely without needing to trust any central party. This is a key enabler of the next-level digital entertainment economy.
(A meme from the community imagining a game where every time you shoot, you have to sign a blockchain transaction)
There have already been many blockchain projects aimed at metaverse, NFTs, and gaming. In the beginning, general-purpose chains like BNB, Polygon, and Solana targeted such applications. Later, dedicated networks such as Flow, WAX, Immutable X, and Ronin emerged, focusing specifically on metaverse, NFT, or gaming use cases.
However, all of these fell short of making FOW a reality. The biggest obstacle was scalability. Imagine implementing a first-person shooter like CS:GO as a fully onchain game. Not only would basic data such as player profiles and kill/death logs need to be recorded, but also rapidly changing data such as player coordinates and health status—all written to the blockchain in real time.
Moreover, the game logic itself would need to be recorded on-chain. This includes logic for bullet firing and collision detection, damage calculation, basic movements like jumping and dodging, player interaction with the environment and map, and many other functions. All of these would have to be computed in real time through smart contracts. In Web2, the term "tick rate" refers to how often a server updates its state. For CS:GO, the server tick rate is 64 times per second, meaning all of the above must be recalculated 64 times per second. This exceeds what current blockchain scalability can handle.
Even if the application is not a real-time FPS, implementing a metaverse or RPG on blockchain would still require enormous scalability. That is why Somnia aims to push scalability to an unprecedented level to serve as the foundation for the fully onchain world.
Somnia is an EVM-compatible layer 1 network that targets over one million transactions per second and sub-one-second finality. Its goal is to enable fully onchain applications across social, gaming, and metaverse domains. Somnia is developed by the Virtual Society Foundation (VSF), which initially received funding from MSquared.
The four core technologies that allow Somnia to achieve such scalability are:
MultiStream Consensus: A consensus algorithm based on Autobahn BFT that improves efficiency by separating data generation and propagation from consensus.
Accelerated Sequential Execution: By combining a custom EVM compiler with hardware-level parallel processing, Somnia maximizes the speed of sequential transaction execution.
IceDB: Somnia uses a custom database called IceDB to maximize read and write performance.
Advanced Compression Techniques: Using streaming compression and BLS signatures, Somnia reduces communication overhead between nodes.
Let’s take a closer look at each of these technologies.
4.1.1 Autobahn BFT
Somnia uses a MultiStream Consensus model inspired by Autobahn BFT. Autobahn BFT aims to overcome the limitations of both traditional BFT protocols and DAG-based protocols.
Traditional BFT protocols such as PBFT and HotStuff offer low latency under stable network conditions. However, when temporary issues like node failures or network attacks occur, incoming requests pile up. As a result, even after the system returns to normal, it continues to suffer from high latency for a prolonged period.
On the other hand, DAG-based BFT protocols like Bullshark and Narwhal are designed for asynchronous environments where node-to-node communication may not be fully reliable. These protocols are optimized to ensure that messages are eventually delivered even under network stress, making them more resilient than traditional BFT models. The trade-off is that DAG-based protocols tend to have higher average latency in normal conditions.
Autobahn BFT seeks to combine the advantages of both models by maintaining low latency in stable conditions while offering fast recovery under network failures. The key feature of Autobahn BFT is its separation of the Data Dissemination Layer and the Consensus Layer.
The core concepts in the Data Dissemination Layer are "lanes" and "cars." Unlike traditional blockchains where all nodes maintain a single shared chain, each node in Autobahn BFT maintains its own individual chain called a lane. Each node collects incoming transactions from users into batches and continuously propagates them. This means nodes do not have to wait for one another and can spread data in parallel at their own pace, significantly increasing overall throughput.
Source: Autobahn BFT
A "car" stands for Certification of Available Requests. It proves that a data batch has been successfully propagated to other nodes. After generating a batch, a node sends it to all other nodes. These receiving nodes then send back votes confirming receipt. Once at least f plus 1 votes are collected, where f is the number of Byzantine nodes, the batch is considered successfully propagated.
Source: Autobahn BFT
The Consensus Layer is responsible for aligning the views of different nodes to maintain a unified network state. As described above, the Data Dissemination Layer does not enforce a consistent global state by itself. Therefore, the Consensus Layer ensures that all nodes agree on the order and state of data across their individual lanes.
In Autobahn BFT, the Consensus Layer reaches agreement by taking a snapshot cut, which is a vector representing the latest state of each node's lane. It then determines the order of the most recent batches across nodes using the PBFT algorithm.
So why is Autobahn BFT more resilient than traditional BFT protocols in the face of network failures? Even if the system encounters issues, the Data Dissemination Layer continues to generate and propagate data without interruption. If one node experiences problems, the rest can still function independently, preventing widespread disruption.
Furthermore, the Data Dissemination Layer supports "instant referencing." Because each batch is connected through cars, verifying the most recent batch automatically confirms the presence of all prior batches. When the network stabilizes, processing a single latest batch enables recovery of all data that may have accumulated during the failure.
4.1.2 Somnia's MultiStream Consensus
Somnia adopts a consensus algorithm based on Autobahn BFT. Each validator in the Somnia network operates its own independent datachain. Only the validator itself has the authority to add blocks to its own datachain, which means these chains are not subject to a separate consensus process internally.
To reach consensus among these datachains, Somnia introduces a consensus chain that aggregates the latest blocks from each datachain. The blocks in this consensus chain are ordered and finalized using a modified version of the PBFT algorithm.
In summary, by adopting Autobahn BFT principles, Somnia separates data generation and propagation from consensus. This separation boosts the overall efficiency of the system and enhances its resilience to temporary network failures or disruptions.
Details about Somnia’s transaction processing can be found in the previously published article, “When Everyone Goes Parallel, Somnia Goes Sequential”. This section offers a brief overview.
4.2.1 Parallel Transaction Execution
To achieve high scalability, Somnia has optimized its transaction execution model. Recently, many blockchain projects such as MegaETH, Monad, Solana, and Aptos have introduced parallel transaction execution as a means of enhancing scalability.
Traditional EVM processes transactions one at a time in a sequential manner. In contrast, parallel execution refers to processing transactions concurrently using multiple cores, as long as those transactions do not access the same state.
There are two main types of parallel execution strategies:
State Access Based: This approach determines the state that each transaction touches before execution, and only executes non-conflicting transactions in parallel. For example, in Solana, the “Instructions” structure within a transaction contains relevant state data. Solana’s Sealevel engine uses this information to execute only non-overlapping transactions in parallel. This method is also used by projects like Sui.
Optimistic Execution: Transactions are executed in parallel first, and if conflicts are detected afterward, only the conflicting transactions are re-executed sequentially. This model is exemplified by Aptos’ Block-STM engine and is also used by projects like Monad and Polygon.
4.2.2 Does Parallel Execution Really Work?
Intuitively, parallel transaction execution seems more efficient than sequential processing, and it does improve scalability. However, according to MegaETH’s simulation results, parallelism alone does not produce a dramatic increase in throughput.
Source: Somnia
Also, consider when networks become most congested. One example is during a high-demand NFT mint event, when many users submit transactions to the same contract. Somnia’s analysis of Ethereum data from 2023 showed a clear concentration of calls to single contracts during peak activity. Because these transactions touch the same state, they cannot be executed in parallel. This means that during moments of highest demand, parallel execution becomes ineffective.
4.2.3 Somnia Goes Sequential
To solve this problem, Somnia maximizes the speed of sequential transaction execution. This improves scalability even in situations where parallel execution is not possible, such as when many transactions target the same contract.
Somnia achieves this by 1) compiling EVM bytecode and 2) utilizing hardware-level parallel processing. Let’s look at each in more detail.
4.2.4 EVM Compilation
Ethereum’s virtual machine, the EVM, is stack-based. A stack-based model means that commands manipulate values on a stack, similar to stacking and unstacking paper cups. For example, to add 2 and 3, you would execute PUSH1 0x02, PUSH1 0x03, and then ADD. This pushes 0x02 and 0x03 to the stack, pops them off, adds them, and then pushes the result 0x05 back onto the stack. The advantage of this structure is its compact and simple bytecode.
VMs generally use two execution models: interpreted and native. An interpreter reads and executes code line by line, while native execution involves running machine code directly on the CPU.
Stack-based VMs typically use interpreters because modern CPUs are designed to efficiently run instructions based on registers. A native stack execution model underutilizes CPU registers and becomes inefficient. EVM also uses an interpreter, which results in slower execution due to repeated instruction lookups.
Somnia addresses this by introducing its own EVM compiler, which converts EVM bytecode into native code that the CPU can execute directly. This allows for faster transaction processing compared to the traditional interpreter model.
4.2.5 Hardware-Level Parallel Processing
Somnia introduces hardware-level parallelism to maximize sequential transaction speed. It’s important not to confuse this with multi-threaded execution of multiple transactions. Instead, this refers to optimizing the processing of individual transactions using a CPU core more efficiently.
Although CPUs appear to execute instructions sequentially, they actually reorder and execute instructions in parallel at the hardware level. For example, if a command requires time to fetch data from RAM, the CPU doesn't idle but instead begins executing subsequent instructions in parallel.
Because EVM bytecode is not natively executable, it cannot take full advantage of this built-in CPU parallelism. By compiling EVM bytecode into native machine code, Somnia enables CPUs to process transactions much faster.
Take the example of processing an ERC-20 token transfer. In EVM, this involves: 1) hashing the sender’s address, 2) reading the sender’s balance from memory, 3) hashing the recipient’s address, and 4) reading the recipient’s balance from memory. With hardware-level parallelism, the CPU can simultaneously hash addresses and access memory, effectively halving the processing time for a single token transfer transaction.
Somnia introduces a new type of database called IceDB, designed to improve gas efficiency, enable fast read and write operations, and enhance the efficiency of state management.
4.3.1 Deterministic Performance
From a common-sense perspective, a database should ideally deliver consistent performance—such as speed and cost—regardless of when or where a data read or write occurs. This is called deterministic performance.
Most existing blockchains, including Ethereum, use embedded databases like LevelDB or RocksDB to store data. These databases are based on Log-Structured Merge Trees, or LSM trees, which are optimized for write-heavy workloads. To maximize write performance, LSM-tree-based databases temporarily store data in RAM and periodically flush it to disk.
This means the same data might be stored in either RAM or SSD at any given time, leading to huge differences in read speeds. Reading from SSD is typically hundreds of times slower than reading from RAM. As a result, traditional blockchains do not offer deterministic performance.
In blockchain systems, transaction fees are based on resource consumption during data access. If performance is inconsistent, how should gas fees be calculated? One approach would be to assume the worst-case scenario and assign high gas fees to all read operations. Another would be to assume average performance and set lower fees. However, neither method is ideal. In Ethereum, for example, the SLOAD operation initially charges 2100 gas and applies a reduced 100 gas fee for accessing the same storage slot again within the same transaction.
Somnia addresses this problem with IceDB. How does IceDB guarantee deterministic performance? Every time data is read or written, IceDB records the exact number of cache lines accessed in RAM and the number of disk pages read from SSD. This performance report is identical across all nodes and fully deterministic. This allows Somnia to apply precise gas fees based on exact resource usage.
4.3.2 Enhanced Read/Write Caching
In databases, a cache is used to store frequently accessed data in RAM rather than on SSD or HDD, enabling much faster access. Most caching systems are optimized for reads rather than writes.
IceDB introduces a balanced caching system optimized for both reading and writing. Instead of using key-value storage like LevelDB or RocksDB, IceDB stores data in columnar format using Parquet files. When reading, it uses DuckDB to efficiently access partitioned data based on specific criteria. When writing, it adopts an append-only model that adds new data without modifying existing entries, significantly improving write performance.
In summary, IceDB uses a fundamentally different storage and read/write architecture compared to LevelDB or RocksDB. This enables faster access to frequently used blockchain data and substantially boosts the overall performance and throughput of the network.
4.3.3 Built-in Snapshotting
Blockchains like Ethereum manage the entire network state using a Merkle Patricia Tree, which allows for simple verification of data integrity. However, in Merkle Patricia Trees, each node stores state data in key-value format, which is not particularly efficient for reading or writing.
Somnia replaces this with the LSM-tree structure already built into IceDB, eliminating the need for Merkle Patricia Trees. This approach offers faster read and write performance, avoids extra overhead from additional processing, and allows for very fast and efficient creation and management of state snapshots.
Blockchain networks are maintained by decentralized nodes that constantly exchange large volumes of data. As a result, the more scalable a network becomes, the higher the hardware requirements to handle increased bandwidth. To address this, compressing the data exchanged between nodes becomes essential.
4.4.1 Streaming Compression
There are two major types of data compression: block compression and streaming compression. Block compression compresses data in independent chunks, where each block can be decompressed on its own. Common .zip files are a good example of this. The advantage of block compression is its simplicity, since the recipient does not need any previous data to decompress a block. However, its compression efficiency is low because it cannot eliminate redundancy across different blocks.
Streaming compression, on the other hand, assumes that both the sender and the receiver share the same history of transmitted data. Rather than resending redundant information, the sender references previously sent data to reduce the size of the message. For example, it might say "use the same address sent 3.4MB ago." This method offers much higher compression efficiency by removing redundancy, but it requires the sender and receiver to maintain a synchronized data flow. It also consumes more bandwidth to maintain this state.
Most blockchains follow a model in which each block is proposed by a different node. This means that the data for each block comes from a different machine, making streaming compression unfeasible. As a result, most blockchains rely on block compression for data propagation.
However, if we look at blockchain transaction patterns, we often find repeated calls to the same account or smart contract. This suggests that addresses and contract calls are not evenly distributed but heavily skewed. This creates significant potential for removing redundancy through compression.
As discussed earlier, Somnia uses a MultiStream Consensus model inspired by Autobahn BFT. In this model, each validator operates its own datachain, which functions as a dedicated data stream. Within a single datachain, the same process manages a consistent data flow, making it possible to reference prior data clearly. This environment is well-suited for streaming compression, which can drastically reduce redundant data and improve network performance.
4.4.2 Compression of Hashes and Signatures
Unlike transaction data that often contains repeated elements, hash and signature data is inherently more difficult to compress.
Hashes, by definition, change drastically even with the slightest modification to input data. This eliminates any predictable pattern, making compression nearly impossible. However, a simple solution is to transmit the transaction data instead of the hash itself. The recipient can then recompute the hash locally. This allows the transaction data to be compressed while bypassing the need to compress the hash directly.
Source: Inevitable Ethereum
Signatures present a different challenge. Since each transaction includes a nonce to prevent reuse, every signature ends up being completely unique. This makes it difficult to find repeated patterns for compression. Somnia addresses this issue using BLS signature aggregation. For example, instead of transmitting 100 individual signatures for 100 transactions, Somnia can aggregate them into a single BLS signature. This drastically reduces the amount of data transmitted and improves network efficiency.
Somnia’s exceptional scalability and architecture optimized for digital entertainment provide an ideal environment for gaming, metaverse, and social applications.
Chunked: A fully onchain MMO sandbox game developed by MSquared on the Somnia testnet. All game logic, user actions, and data are recorded on the blockchain. Chunked generated over ten million transactions on the testnet in less than a week.
Source: MSquared
MSquared: As described above, MSquared is a subsidiary of Improbable that provides technology, services, and consulting related to the metaverse.
Uprising Labs: Uprising Labs collaborates with Somnia to attract promising games into the ecosystem, offering developer mentoring and incubation programs for fully onchain gaming. Together, they operate the $10M Dream Catalyst program, providing funding, mentoring, industry connections, tools, and marketing support to help games onboard into the Somnia ecosystem.
Dark Table CCG: A four-player digital card game currently participating in the Dream Catalyst program, now being developed as an onchain game.
Adventure Gold DAO (AGLD DAO): With experience in building community-driven experiments and autonomous worlds in the Lootverse, AGLD DAO will provide development tools to support onchain game creation and help expand the Lootverse within the Somnia network.
Maelstrom: A fantasy-themed naval battle royale game that is expanding into the Somnia ecosystem with support from Dream Catalyst.
Kraft Labs: Offers various game services based on generative AI and blockchain.
EMERGE: A company specializing in game IP management, publishing, community operations, and marketing. EMERGE is the official gaming publishing partner of the Somnia ecosystem.
Masks of the Void: A roguelike action RPG developed by RolldBox Games and published by Uprising Labs. All gameplay and data are processed onchain, and the game is accessible directly via browser with no installation required.
Mullet Cop: Mall Sim: A simulation game based on a retro-style graphic novel where players design a mall and respond to various events. In-game assets are issued as NFTs on Somnia.
Netherak Demons: An action RPG set in a dark fantasy world where players control demons and battle various enemies. All equipment and items in the game will be registered on Somnia. This title is also part of the Dream Catalyst program.
Lucky Adventure Gambit: A turn-based RPG dungeon crawler built onchain.
YOM: A decentralized cloud game streaming platform that enables high-performance games to be played in real time across various devices.
Hyperlane: One of the leading cross-chain protocols that enables Somnia to achieve interoperability with over 150 connected chains.
Glacis Labs: A project focused on simplifying cross-chain messaging and enhancing security, helping developers efficiently build cross-chain applications.
Ankr: A blockchain infrastructure company providing RPC nodes and subgraph services for Somnia.
Ormi: A data infrastructure platform for blockchain developers and analysts that supports advanced querying using The Graph and AI-powered onchain data analysis.
DIA: One of the leading blockchain oracle projects providing price feeds for assets in the Somnia ecosystem.
Salt: A decentralized MPC infrastructure allowing users and institutions to manage assets without relying on centralized custody.
Dune: An onchain data analytics platform that helps developers and users analyze dApp activity and transaction patterns within the Somnia ecosystem.
Palmera DAO: A platform for managing Safe-based multisig wallets on blockchain networks. It helps Somnia users manage assets securely and efficiently.
Protofire: A blockchain infrastructure company providing services such as oracles and subgraphs. Developers in the Somnia ecosystem can use Protofire to easily build applications.
Privy: Provides simple APIs that allow applications to easily integrate wallets and enable user interaction. Many of Somnia’s games and dApps use Privy for seamless wallet integration.
Coin98: A multi-chain DeFi and wallet service that also supports trading tools, dApp browsers, and AI assistants to deliver a high-quality user experience.
thirdweb: A Web3 development platform offering various tools, templates, and SDKs to help developers build dApps and games more quickly and easily within the Somnia ecosystem.
Sequence: A Web3 game development platform that offers smart wallets, game engine integrations, gasless transactions, and multi-chain support, allowing developers to add Web3 features to games without deep blockchain knowledge.
Galeon: A development platform that automates the Web3 game creation process using multi-agent AI systems.
QuickSwap: An AMM DEX that provides users with services such as token swaps, liquidity provision, perpetual trading, and yield farming.
Standard Protocol: An all-in-one DeFi platform offering services including a CLOB, derivatives trading, stablecoins, and lending.
Nomis Protocol: A protocol that builds identity and reputation based on a user’s onchain activity in the Web3 ecosystem. dApps in the Somnia ecosystem can use this data to offer personalized experiences.
QSTN: A decentralized survey platform where users receive token rewards for participating in surveys.
Unstoppable Domains: Users in the Somnia ecosystem can easily link blockchain-based digital identity systems like .Dream domains to their wallet addresses.
Haifu.fun: A DeFAI project that combines DeFi with AI agents.
Otomato: A platform that allows users to create blockchain-based AI agents without writing code.
LootMogul: An AI-powered, fan-centric sports gaming and e-commerce platform that helps fans collaborate with athletes and teams to build games, play, and monetize products.
Sogni AI: A creative AI platform based on a decentralized GPU network.
ForU AI: A platform that allows users to manage and utilize their personal data and identity onchain using AI agents.
Grillz Gang: A collection of 5,555 Ethereum-based NFTs. Through a partnership with Somnia, Grillz Gang NFTs can be used as avatars in the Somnia ecosystem.
Quills Adventure: A collection of 3,333 PFP NFTs. Integrated with the Somnia ecosystem, holders can explore the world and complete various quests.
Somnia represents more than just a high-performance blockchain. It embodies a technical and philosophical evolution aimed at realizing a fully onchain digital world. While earlier blockchain-based gaming and metaverse projects struggled due to infrastructure limitations and lack of scalability, Somnia pursues scalability through foundational innovations like MultiStream Consensus, Accelerated Sequential Execution, IceDB, and streaming compression.
Ultimately, Somnia functions as the dream computer of the digital era. Here, dreams are not just fantasies but realities expressed in code, represented as objects, and shared through common protocols. Historically, entertainment has long been a central pursuit of humanity. Now, that dream is ready to be realized on a technological foundation. Somnia is both the stage for that dream and the beginning of a new order where dreams are shared.
