ERPC Brings XDP and zero-copy to Production for RPC and Geyser gRPC Nodes in Every Region

ELSOUL LABO B.V. (Headquarters: Amsterdam, Netherlands; CEO: Fumitake Kawasaki) and Validators DAO, which operate ERPC, are pleased to announce that ERPC has brought the XDP fast path and AF_XDP zero-copy of Solana v4 (Agave 4.x) to production for the RPC nodes and Solana Geyser gRPC nodes across all regions. This expands the optimization first proven on the source of the New York (NY) region Geyser gRPC to production infrastructure in every region.

The XDP fast path and AF_XDP zero-copy are Turbine-oriented optimizations that became available in Solana v4 (Agave 4.x). As for launch flags, the Agave 4.1 line deprecated the --experimental-retransmit-xdp-* family and organized them into --xdp-interface / --xdp-cpu-cores / --xdp-zero-copy. Building on this Turbine-oriented XDP fast path and zero-copy, ERPC has applied it in production not only to the source nodes that back the Geyser gRPC, but also to the RPC nodes. Both RPC nodes and Geyser gRPC nodes receive the shreds that make up blocks via Turbine. By using XDP and zero-copy to reduce kernel overhead on that shred propagation and ingest path, the Geyser gRPC sees shorter stream delivery lag, while RPC sees improved state freshness and a better ability to keep up with the latest state. This change is already running in production across all regions. Customers who prioritize first-arrival performance can try the all-region RPC and Geyser gRPC right away — via hourly billing (per hour) or Crypto Pay (SOL / USDC / EURC).

ERPC Official Site: https://erpc.global/en ERPC Dashboard: https://dashboard.erpc.global/en

What All-Region Rollout Changes — Making the Path That Grabs Shreds Faster, Everywhere

On Solana, the leader responsible for block production rotates on a short cycle, so the origin of communication is always moving. In this structure, what matters in practice is not being close to a single fixed point, but having a high probability of being close to the network where the major nodes and validators are concentrated — and this directly affects latency, retransmission rates, and failure rates in real operation. That is why ERPC believes it is meaningful not to make just one machine fast, but to raise the production nodes in every region to the same level.

The key to this rollout is that the target of the optimization is the "shred propagation and ingest path on Turbine." For both RPC and Geyser gRPC, the ultimate speed is supported by "how quickly a node can grab a block." XDP and zero-copy are optimizations that reduce the overhead of exactly that grabbing stage — the receiving, retransmit, and propagation of shreds over Turbine. By applying this to production nodes in every region, no matter which region's connection point a customer uses, they can receive data over an optimized path.

What the XDP Fast Path and zero-copy of Solana v4 Are

XDP (eXpress Data Path) is a Linux kernel technology that lets high-performance network code bypass much of the kernel's usual packet-processing path. By reducing data copies and context switches, it processes packets with far less overhead than the standard networking stack.

In Agave (the Solana validator client), XDP is applied to Turbine, the protocol that propagates blocks between validator nodes. Received shreds are processed by an eBPF program attached close to the network interface card (NIC) and mapped into user-space buffers via AF_XDP. When zero-copy mode is used, received data is handed off directly from kernel to user space without copying. Outbound shreds, too, leverage the AF_XDP send path to reduce copies and system-call overhead on the hot path.

Anza introduced XDP for Turbine in the Agave 3.x line (from v3.0.9) and carried it into the foundation of Solana v4 (Agave 4.x). The launch flags have been organized through successive releases: the Agave 4.1 line deprecated the --experimental-retransmit-xdp-* family and organized them into --xdp-interface / --xdp-cpu-cores / --xdp-zero-copy. According to Anza's setup guide, with XDP, large validators can approach 150,000 outbound packets per second through Turbine's fanout.

Anza Agave XDP Setup Guide: https://www.anza.xyz/blog/agave-xdp-setup-guide

Applied to Geyser gRPC Nodes — Speeding Up Source-Side Ingest

Geyser gRPC is the path for receiving account, slot, block, and transaction updates as a stream rather than by polling. Here, a one-millisecond difference connects directly to capturing execution opportunities and to perceived front-end speed. Geyser lag is ultimately supported by "how quickly the source can grab a block."

XDP and zero-copy are precisely the optimizations that reduce the overhead of that source-side shred propagation and ingest path. As the source can receive and propagate shreds faster, it can observe and reconstruct blocks at an earlier stage, which shortens the lag for those updates to reach customers through the Geyser gRPC stream. In the New York (NY) region, where we applied this first, we confirmed through open-source measurement that this optimization is effective in the tail region of delivery lag. We have now extended the same optimization to the Geyser gRPC nodes in every region.

Applied to RPC Nodes — Improving State Freshness and Keeping Up With the Latest State

What is new in this rollout is that we applied this optimization to RPC nodes as well. RPC nodes, too, receive the shreds that make up blocks via Turbine and update their own ledger and state. When XDP and zero-copy reduce the overhead of that Turbine-based shred ingest and propagation, RPC nodes can ingest newer blocks sooner.

For customers using RPC, this is not a matter of all RPC methods getting uniformly faster, but surfaces as the freshness of the state the node holds. When you query the most recent slot or block, or the latest state of an account, the fact that the information the node has already ingested is newer connects directly to the freshness of the data in the response. In addition, lower overhead on the propagation and ingest path translates into headroom for the node to process updates without dropping them under heavy load. The same optimization as the Geyser gRPC stream delivery also works as the foundation that supports the freshness of the data RPC returns — that is the reason we extended the scope to RPC nodes this time.

Brought to Production Nodes in Every Region — What We Enabled

ERPC migrated the RPC nodes and Geyser gRPC nodes across all regions to Solana v4 (Agave 4.x) and put the Turbine-oriented XDP fast path and AF_XDP zero-copy into production. Building on the --xdp-interface / --xdp-cpu-cores / --xdp-zero-copy launch-flag scheme organized in the Agave 4.1 line, we enable it according to each region's configuration.

Enabling XDP requires advanced and error-prone tuning: a recent kernel, an XDP-capable NIC, the right systemd capabilities for the validator process, correct launch flags, and appropriate CPU-core pinning. Deploying this not to a single node but to production nodes in every region — while verifying each region's differing NIC, kernel, and network configuration — raises the operational difficulty further. ERPC applies the operational know-how cultivated from running validators at the top of the network directly to building and operating the source nodes and RPC nodes in each region.

And the operational know-how for this optimization is consolidated as a recipe in the open-source Solana operations tool SLV. SLV provides everything from enabling XDP (via configuration variables such as xdp_enabled / xdp_zero_copy) to measuring delivery lag (slv check geyserbench), in a form that anyone can reproduce through conversations with an AI agent or via CLI. The optimization ERPC achieved across all regions is not a one-off trick for a single machine, but stands on a reproducible operational recipe.

SLV GitHub: https://github.com/validatorsDAO/slv

Verify With Your Own Numbers, From Your Own Connection

How much difference this optimization makes varies with the connection origin, route, time of day, and leader distribution. That is exactly why ERPC places importance on showing delivery quality not through subjective claims or marketing copy, but through measurement that anyone can verify with the same method. What customers can verify is not a fixed number but the measurement method itself.

For the Geyser gRPC, ERPC's benchmark tooling is open source. For first-arrival comparisons you can use slv check geyserbench --kind grpc, and for connectivity and latency checks of an individual endpoint, slv check grpc, comparing under conditions close to your own workload. For RPC as well, the most reliable approach is to measure and observe the freshness of the data returned and the behavior of responses from your own connection point, using the requests your own bot or application actually sends.

Being able to make decisions based on numbers you measured yourself, rather than a vendor's claims, is the starting point for customers who prioritize first-arrival performance. The steps from installing SLV to running the measurement are published in the SLV Getting Started guide.

SLV Official Site: https://slv.dev/en SLV Getting Started: https://slv.dev/en/doc/general/getting-started/ Solana Geyser gRPC Speed Comparison: https://erpc.global/en/doc/geyser-grpc/speed-comparison/

Suppressing Distance-Derived Latency by Design — the AS200261 Solana-Specific Data Center

ERPC's latency advantage does not come from software optimization alone. By placing source nodes, receiving endpoints, and processing nodes inside premium data centers where Solana validators are densely concentrated, ERPC suppresses distance-derived latency at the design stage.

ELSOUL LABO operates a Solana-specific data center under its own ASN (AS200261), assigned by RIPE NCC, as part of the ERPC platform. Software optimizations like today's XDP and zero-copy deliver their maximum effect only on top of this physical and network proximity design. With both design-level proximity and node-side software optimization in place, first-arrival performance, low-latency streaming quality, and fresh RPC responses are realized.

A Lineage of Continuous Infrastructure Strengthening

This all-region rollout sits within the lineage of infrastructure strengthening that ERPC has continuously pursued. It is the latest-generation optimization, following the all-region Geyser gRPC infrastructure upgrade in December 2025, the large-scale strengthening of the Frankfurt (FRA) region in January 2026, and the earlier application of XDP and zero-copy in the New York (NY) region in June 2026. The optimization proven in NY has now been extended — with RPC nodes added to its scope — to production infrastructure in every region.

Rather than responding to rising demand with limits or degradation, ERPC consistently absorbs it by strengthening the infrastructure itself. We continue to reflect the latest-generation optimizations into our production infrastructure while repeatedly verifying compatible NICs, kernels, and network configurations. ERPC's RPC and Geyser gRPC will continue to evolve.

Verify From a Single Hour With Hourly Billing

ERPC's RPC and Geyser gRPC can be tried from a single hour via the hourly billing plan. This makes a low-risk verification loop possible: contract for just one hour, confirm the actual behavior as seen from your own bot or application's connection point within that hour, and decide on moving to a monthly or annual plan based on the results. The slv check measurements described above can be run as-is within this one-hour trial.

Once your configuration and usage become clear, switching to a monthly or annual plan keeps you on the same dashboard and the same endpoint quality.

ERPC Dashboard: https://dashboard.erpc.global/en

Crypto Pay (SOL / USDC / EURC) Supported

ERPC offers Crypto Pay for purchasing ERPC credits and for paying for its plans, and it is supported for the hourly billing plan as well. You can choose SOL, or the stablecoins USDC / EURC, as the payment asset. EURC can be sent directly, while USDC or SOL is swapped to EURC via Orca, with the transfer completed within the same flow.

For teams building and operating on Solana, being able to handle infrastructure costs in a manner close to their existing wallet-based fund management flow is a practical improvement that lowers the friction to start verification. The hourly-billing verification described above can also be started directly from the assets in your Solana wallet.

Order, Pay, and Manage Solana-Specific Infrastructure on One Platform

ERPC lets you combine Solana RPC, WebSocket, Solana Geyser gRPC, Solana Shredstream, Direct UDP Stream (Raw Shreds), VPS, bare-metal servers, dedicated RPC, SWQoS, a Pyth-enabled Price API, and Jet Analytics & Indexed RPC on a single platform.

The ERPC Dashboard supports 16 languages, letting you handle plan selection, region selection, stock checks, adding to cart, credit top-ups, checkout, reviewing API keys and endpoints, checking usage, and creating support tickets — all from the same screen.

R&D and Continuous Improvement of Solana-Specific Infrastructure

Behind ERPC is the research and development of Solana-specific infrastructure that ELSOUL LABO continues to pursue. ELSOUL LABO has been approved for five consecutive years since 2022 under WBSO, the Netherlands' government R&D support program. It continues R&D on Solana RPC infrastructure, validator operations, real-time data delivery, and AI-agent-assisted operations and development, and those results are reflected across services including ERPC, SLV, SLV AI, and the AS200261 Solana-specific data center.

Today's Solana v4 / XDP / zero-copy support across all regions also took shape from operating validators at the top of the network. ERPC will continue to provide low-latency infrastructure close to the Solana network and to demonstrate its quality through measurement that anyone can verify with the same method.

Usage and Consultation

For optimal regional configurations including the all-region RPC and Geyser gRPC, selecting between standalone gRPC plans and gRPC Bundle plans, choosing between hourly, monthly, and annual billing, and migration design from an existing configuration, we provide individual consultation on the official Validators DAO Discord.

ERPC Dashboard: https://dashboard.erpc.global/en ERPC Official Site: https://erpc.global/en Validators DAO Official Discord: https://discord.gg/C7ZQSrCkYR

We sincerely thank all of our users for their continued use of ERPC.

ERPC Brings the Solana v4 XDP Fast Path and zero-copy to Production for RPC Nodes and Geyser gRPC Nodes Across All Regions — Extending the NY Proof to Every Region, Lifting Both Stream Delivery and RPC State Freshness