Arcfra AECP 6.3 Breaks the 11M IOPS Barrier, Delivering Tier-1 All-Flash Performance and RPO=0 Resilience for Enterprise Cloud

Achieving a 4.6X performance leap on a 3-node standard hardware, redefining enterprise cloud economics by eliminating the need for expensive proprietary storage.

Singapore, March 20th, 2026 — Arcfra, a pioneer in high-performance cloud & AI infrastructure, today announced the release of Arcfra Enterprise Cloud Platform (AECP) 6.3. Powered by the latest hyperconvergence core Arcfra Cloud Operating System (ACOS) 6.3, the update delivers a landmark 4.6X performance increase and introduces native RPO=0 synchronous replication, offering enterprises a high-performance, cost-effective alternative to traditional VMware-based architectures.

As enterprises face mounting pressure to efficiently process massive data volumes and navigate soaring VMware licensing costs, AECP 6.3 balances extreme performance with economic sustainability through architectural enhancement and native integration, eliminating the need for extra hardware and complex third-party tooling.

“In 2026, the criteria for enterprises evaluating virtualization platforms has shifted from ‘can it run’ to ‘can it run efficiently and sustainably,’” according to Robert Li, Director of Alliance and Product Marketing at Arcfra. “The release of AECP 6.3 marks the official entry of hyperconverged infrastructure into a new era of native AI acceleration. Not only does it achieve a 4.6X leap in performance, but by building in complex features natively, it solves the complexity transfer issue that enterprises fear most when migrating from legacy architectures”.

Shattering the Performance Bottleneck

ACOS 6.3 breaks the I/O ceiling to deliver “Hardware-Limit” speeds through “Software-Only” optimization, far surpassing standard peer configurations and matching the capabilities of high-end proprietary All-Flash Arrays. This leap moves beyond general enterprise needs into the “Extreme Speed” domain, powering high-concurrency AI/VDI workloads and massive real-time databases.

Validated in a standard 3-node cluster environment, the performance breakthroughs include:

• 11 Million+ IOPS in 4K Random Read (4.6X higher than previous version).
• 130+ GiB/s in Sequential Read (3.6X higher than previous version).
• <100 μs latency in 4K Random Read.

These metrics are driven by deep architectural enhancements, including:

• Intel Data Streaming Accelerator (DSA) hardware acceleration to offload data movement tasks to dedicated hardware engines, reducing CPU contention and ensuring higher stability during peak operations.
• Modernized I/O stack with IO_uring enabled by default to significantly improve asynchronous I/O efficiency by reducing syscall overhead and lowering latency under high concurrency.
• Default 8-stripe virtual volumes to increase storage parallelism across distributed disks, improving workload distribution and scaling under high VM density while eliminating the need for manual performance tuning.
• Multi-channel bandwidth aggregation to allow multiple NICs to operate as a unified high-throughput pipeline for higher sustained throughput and reduced bottlenecks.

Zero-Risk Business Continuity for Mission-Criticals

To protect mission-critical AI and financial workloads, AECP 6.3 delivers enterprise-grade availability and disaster recovery without bolted-on third-party tooling. The introduction of real-time synchronous VM-level replication completes Arcfra’s industry-leading full-scenario disaster recovery solution, creating a unified matrix of Synchronous, Asynchronous, and Active-Active protection.

Enterprises can now architect flexible, cost-effective continuity plans by selecting the optimal replication technology at the individual VM level to meet specific SLA requirements. Key features include:

• Introduces real-time synchronous VM-level replication in Arcfra Backup & Disaster Recovery, achieving a Recovery Point Objective (RPO) of zero. This new capability ensures no data loss during site failures, providing the absolute data integrity required for high-frequency financial transactions and AI/VDI.
• Reduced minimum node requirements for active-active stretch clusters (from six to four nodes) to lower the financial threshold for mid-sized enterprises building active-active data centers.
• Extended VM high availability (HA) to support performance-accelerated workloads utilizing SR-IOV passthrough NICs and vGPUs, ​ensuring that even hardware-accelerated AI tasks remain resilient in the event of a host failure.

Native Security Architecture

Moving away from “patchwork” security, ACOS 6.3 provides security by design across data in transit, at rest, and network visibility. This release introduces built-in safeguards to simplify compliance and harden infrastructure without adding operational overhead. Key features include:

• Encrypted live migration to guarantee the confidentiality of memory and disk states during VM movement. This ensures alignment with strict regulatory data-in-transit requirements.
• Encryption at rest powered by a built-in Key Management Service (KMS). This native integration reduces operational complexity and overhead compared to external KMS solutions, simplifying the entire key lifecycle management process.
• Enhanced network visibility utilizing ERSPAN traffic mirroring tunnels for external threat detection systems. Additionally, the integration of IGMP/MLD snooping on the virtual distributed switch improves multicast efficiency and monitoring for compliance frameworks.

Operational Simplicity at Scale

ACOS 6.3 introduces several lifecycle management enhancements designed to liberate IT teams from manual labor and reduce hidden operational costs. These improvements focus on automating repetitive tasks and ensuring high reliability during infrastructure transitions. Key features include:

• Batch VMTools upgrades to allow multiple VMs to be selected and updated in a single action. This significantly reduces repetitive maintenance tasks and improves management efficiency in large-scale environments.
• Preserved virtual NIC PCI addresses during migrations to prevent post-migration network configuration failures and application binding issues, ensuring seamless service transitions.
• Optimized hot migration in non-Boost mode to transfer only valid data, reducing migration duration, lowering bandwidth consumption, and minimizing service disruption windows for business-critical applications.

Resources

For detailed features, please read What’s New in Arcfra Enterprise Cloud Platform 6.3.

Performance Disclaimer

Performance data is based on internal testing with specific hardware configurations (Intel platform, 3-node Arcfra cluster, ACOS 6.3 + 4 Instances); actual results may vary depending on system environment, workloads, and hardware specifications. Performance gains are calculated against the version 6.2 baseline. This configuration represents the maximum performance tier observed in the test. Please note that +4 Instances will result in higher CPU and memory consumption. This trade-off is necessary to achieve peak I/O density and throughput. Optimal performance is subject to system environment conditions and overall cluster health.