Author: Haotian

Let's talk about the recently strong performance of the secondary market. In particular, why is the InfiniSVM technology roadmap attracting much attention? What are the features of hardware-accelerated SVM expansion solutions? And how will the Solana expansion ecosystem industry structure break through after hardware acceleration? Next, let’s talk about my outlook-based observations:

1) Unlike the previous horizontal expansion route dominated by Ethereum, the Solayer team showed a completely different expansion idea in the infiniSVM white paper: to accelerate deep optimization of SVM through hardware acceleration and build a million-level TPS blockchain network, which will essentially be expanded through a deep integration of hardware + software.

If you look back at the entire blockchain expansion history, the earliest on-chain expansion idea was achieved through parameter adjustment (larger blocks, shorter block time), but this idea is easy to touch the impossible triangular dilemma of blockchain; the layer2 expansion idea that emerged later is a horizontal expansion, and the core purpose is to divert transactions through layer2 (state channels, side chains, Rollup, etc.), which will undoubtedly sacrifice part of the global atomicity; and the hardware acceleration route explored by InfiniSVM is a new route that upgrades the expansion concept, while maintaining a single global state, it uses professional hardware to break through performance bottlenecks;

Simplely put: InfiniSVM The expansion method does not lie in simply optimizing algorithms, but reconstructing the SVM execution environment through microservice architecture and hardware acceleration, and resolving some critical tasks based on dedicated hardware, thereby achieving atomicity and consistency in the global state under high load state.

2) Following this idea, many friends must be puzzled, why does Solana's SVM execution environment require hardware acceleration? As can be seen from the data provided by the Solayer white paper, the current Solana verification node already requires CPUs above 3.1GHz, 500GB+ high-speed memory, and 2.5TB+ high-throughput NVMe storage. Even with such a high configuration, the CPU utilization is only about 30% at high loads, and P2P communication is close to the upper limit of 1Gbps bandwidth for consumer-grade networks.

The problem is coming. Since the CPU is not satisfied, whyMore powerful hardware? This actually exposes Solana's current performance bottleneck lies in other links beyond CPU computing capabilities, such as: the microservice processing architecture, which can isolate different processing links and match more suitable hardware resources; dedicated accelerators can assign specific tasks such as partial signatures to dedicated hardware, etc.

You see, InfiniSVM is not simply upgrading hardware, but redesigning the entire execution environment and providing more special hardware optimization solutions for each bottleneck link. For example, to improve workshop production efficiency, it is necessary to reorganize the hardware and software of the entire production line, rather than simply increasing the number of workers;

3) So, what are the features of InfiniSVM's hardware acceleration solution worth sharing?

1. Distributed microservice processing architecture. The previous Solana single transaction processing process could be decomposed into multiple extended processing links such as signature verification, deduplication, scheduling, and storage. Each link in the InfiniSVM architecture can be independently processed, avoiding the big problem of "one link is stuck and waiting in full line";

2. Intelligent transaction scheduling system. Originally, when Solana reads and writes transactions, it had to queue up to process when it belongs to the same account, but InfiniSVM can make operations not interfere with each other even under the same account, thereby greatly improving its parallel processing capabilities. To put it bluntly, it further enhances the refined management capabilities; 3. RDMA low-latency communication technology, normal communication between nodes requires at least necessary steps such as packaging, delivery, and unpacking. RDMA can directly transmit the data of one node to the memory of another node, achieving a breakthrough in communication technology from milliseconds to microseconds, which can significantly reduce state access conflicts; 4. Distributed intelligent storage network. If a single account is allowed to undertake its account data, Solana was previously limited by 10MB, but InfiniSVM adopted a distributed cloud storage solution to distribute the data to different nodes and marked as fast lanes, slow lanes, etc., which not only breaks through the capacity limit, but also optimizes the data access speed.

4) After explaining the problem of the technology upgrade route, many "what's use" sounds must have come out. Overall, with the support of hardware acceleration, Solana's competitive advantage in layer1 competition can be further enhanced. And compared to Ethereumayer2 needs to be supported by application data in the ecosystem to show expansion achievements. This performance breakthrough of using hardware to achieve a million-level TPS may only be directly supported by the access to a very small number of vertical scenarios, and the implementation path is shorter;

Another way to think about it, take @jito_sol as an example, as the Solana MEV infrastructure, its niche value in transaction sorting optimization, MEV extraction, validator returns, etc. If it cannot prove itself before the Solana MEME boom, Jito's method of optimizing transactions to conduct systems has become an indispensable part of the following MEME booms in the past year.

In fact, Solaer's current technical card is similar. Its upgrade method that allows transactions to perform better in the system is not obvious by relying on simple financial transaction scenarios. However, if we think of the large-scale implementation of PayFi in the future, if we want Solana to perfectly carry the payment and settlement infrastructure functions with high throughput and low latency, the performance of TPS will be clearly felt. In addition, there are DePIN ecosystems, complex chain games, AI Agent application scenarios, etc. Anyway, looking one step ahead to define the value of a technical infra project is easier to see clearly than just from the perspective of whether it is Utility at the moment.