Youssef Gueddari | Head of ICT Division, Ministry of Finance - General Directorate of Taxes

— Included in Data Dialogue, Issue 002

Guided by the Digital Morocco 2030 strategy, Morocco is committed to becoming a leading digital hub in Africa. As the steward of Morocco's economy, the Ministry of Economy and Finance (MEF), together with its subordinate Direction Générale des Impôts (DGI), is at the forefront of this digital transformation. With Morocco's GDP exceeding US$170 billion in 2025, the fiscal system would face increased processing pressure.

At this critical stage of digital transformation, transforming the underlying architecture seamlessly can be more difficult than building a new system from scratch. MEF must address long-standing technical iteration issues while ensuring zero interruption of government services, and reserve performance redundancy for the future peak workloads. This article explores how Morocco MEF has strategically upgraded its storage architecture from an external gateway–based pool model to a native active-active all-flash model through a fundamental transformation of its storage foundation. This not only resolves existing I/O bottlenecks, but provides a strong technical moat for security and continued evolution.

I.  The Intersection Between Technology Iteration and Explosive Business Growth

1.     Electronic invoicing: a technology surge in response to business demand

The mandatory implementation of electronic invoicing in Morocco is a crucial step toward modernizing fiscal and tax governance. According to the Loi de finances 2026, starting from early 2026, all enterprises must use standardized electronic invoice formats such as UBL 2.1 or CII XML and support real-time clearance by the DGI.

This policy shift creates substantial performance challenges for the IT architecture. Under Morocco's clearance model, each B2B invoice must be validated in real time by the tax authorities before it can become legally effective. This places the following requirements on the storage system:

High-concurrency processing: Millions of enterprises may issue invoices online simultaneously. This would generate I/O requests characterized by high concurrency, small file sizes, and low latency requirements.

Predictable response times: Any interruption, latency, or jitter at the storage layer directly affects the enterprise front end, which could hinder the operational efficiency of the broader economy.

Strict and compliant archiving: Regulations require invoices to be stored in an immutable format for up to 10 years. This places stringent demands on data immutability and effective capacity management.

2.     Gateway-based architecture: the intergenerational challenge of traditional design

For a long time, MEF relied on storage gateways to pool heterogeneous devices from multiple vendors. While this stitched architecture maximized investment value, it gradually became a technical bottleneck due to large volumes of transaction workloads.

Single point of failure and complexity: The gateway layer introduces a fault domain, and compatibility issues are likely to arise during firmware upgrades or configuration changes.

Performance bottlenecks: Data read and write operations must pass the gateway for data packing and unpacking, resulting in significant communication overhead. According to test results, when traditional HDD-based storage systems handle high-concurrency small-files workloads, physical seek time can cause response times to exceed 10 ms.

Protocol conversion overhead: Under cross-site data consistency protocols (such as traditional active-passive synchronous replication), the protocol forwarding logic at the gateway layer significantly increases end-to-end latency.

II. Architecture Selection: Embracing Native and Flattened Design

When formulating solutions, Huawei adheres to the principle of reducing intermediate layers and embracing native data storage. To address the architecture limitations and performance bottlenecks of the traditional gateway-based architecture while ensuring long-term service evolution, MEF ultimately chose an upgrade path based on OceanStor Dorado high-end all-flash storage.

1.     Native active-active architecture: from reactive DR to proactive load balancing

Unlike the traditional active-passive model, MEF uses native active-active technology to fundamentally transform performance and reliability at the storage layer.

Gateway-free design: External gateway forwarding is completely eliminated. Data reads and writes are directly synchronized between storage controllers, simplifying the architecture. According to financial and taxation system tests, the I/O response time for typical transactions remains stable within 0.5 ms.

Deep protocol-layer convergence: Integrated SAN and NAS active-active deployment is implemented at the storage controller kernel layer. This enables load balancing across two data centers for both structured database transactions and unstructured electronic document workloads.

Multi-level reliability assurance: Multiple levels of resilience and reliability are provided by the SmartMatrix full-mesh balanced architecture within a single device and active-active deployment across data centers.

2.     Effortless handling of massive amounts of small XML files

For small-file workloads like electronic invoicing (massive amounts of small XML files) that require high-frequency concurrency and fast response, OceanStor Dorado transforms scattered data into organized distribution, intermediate layers into direct access paths, and isolated operations into coordinated execution. This virtually eliminates the inefficiencies and chaos typically seen in traditional storage systems when handling large volumes of small-file workloads.

Solving high concurrency: To prevent long queues at any single "counter" (controller), millions of invoice files and their directories are evenly distributed across all storage controllers using a globally distributed file system and a directory balancing algorithm. This ensures that workload differences between counters remain below 5%, allowing all hardware to operate at full capacity and handle millions or even tens of millions of concurrent operations per second (OPS).

Achieving low latency: The "address book" for data has been indexed to eliminate the time spent searching through pages. When handling massive amounts of small files, the biggest bottleneck is often not the data transfer, but locating where each file is stored (metadata retrieval). The Huawei solution improves cache efficiency by using metadata compaction and AI-driven intelligent prefetching to compress and organize scattered file location information. In addition, the built-in AI processor predicts which invoice files are likely to be requested next based on historical access patterns and preloads them from storage to the counter. This increases cache hit rates by 50% and ensures read and write latencies of just 0.5 ms.

Minimizing jitter: This is analogous to filling the delivery crate before it leaves the mailroom. When traditional storage systems handle a large number of small file writes, write amplification and frequent background garbage collection (GC) can occur in the storage media, causing occasional jitter. It's like a sorting clerk trying to sweep the floor while constantly stopping to pick up individual envelopes. The Huawei system addresses this by accumulating thousands of small invoices into a single data block in memory before writing it to disk in one operation. This reduces write amplification and minimizes GC activity, ensuring a stable, smooth response curve, even during the peak load of annual tax filing.

III. Best Practice: Heterogeneous Takeover and Non-degraded Migration

MEF deployed two OceanStor Dorado 8000 all flash arrays and one OceanStor Dorado 6000 all-flash array to build a cross-site disaster recovery (DR) data center. This upgrade was not completed overnight. Rather, it was executed through two clearly defined phases to ensure a smooth transition when upgrading the underlying capabilities.

Phase 1 (pilot breakthrough): Huawei introduced OceanStor Dorado 6000 to the core, integrated DGI taxation platform as a breakthrough point to validate the ability of all flash to support complex tax workloads.

Phase 2 (full-scale deployment): Following the successful pilot, the solution was rapidly scaled and deployed across the entire organization. The core production system uses OceanStor Dorado 8000 high-end all-flash arrays to support the active-active DR architecture, while the general service system uses OceanStor Dorado 6000. This forms a tiered but unified all-flash storage DR system.

For government systems, data migration is like changing tires on a moving vehicle. MEF must seamlessly transition from legacy architecture to an all-flash storage foundation with zero downtime for mission-critical tax applications. To achieve this, Huawei used a phased upgrade strategy of "takeover first, migration later", anchored by two technologies.

1.     Reshaping the value of existing assets

The SmartVirtualization technology enables the legacy storage devices of different brands to be logically integrated into the new storage arrays for streamlined management. This bypasses vendor lock-in, eliminates compatibility barriers, and extends the lifecycle of existing assets for intensive resource utilization.

2.     Zero-impact service switchover

In the data migration phase, the SmartMigration technology handles data transfers silently in the background.

• Zero service interruption: This is like changing tires while driving. Upper-layer tax applications stay online throughout the entire process, with no shutdowns or reboots.

• Response in milliseconds: The final path switchover is completed in milliseconds, remaining completely invisible to users.

This combination of logical takeover and seamless migration ensures zero migration risk, zero service interruption, and maximum utilization of existing assets.

IV.  Proven Performance: MEF's Digital Transformation

For Morocco's real-time clearance model, MEF used the OceanStor Dorado all-flash array's intelligent FTL algorithm to reduce the read latency to 80 μs, providing near-real-time physical response for the verification of massive amounts of electronic signatures. Post-deployment measured data revealed IOPS exceeding 1 million, far beyond the expected 350,000. Overall latency was reduced by over 60%. Tasks like quarterly tax reconciliation that once took hours can now be completed in minutes. This eliminates high concurrency bottlenecks and future-proofs performance for projected surges in 2026.

The solution ensures zero downtime with its all flash active-active architecture. In simulations, automatic failover was executed in seconds, ensuring zero RPO and near-zero RTO. Furthermore, DME intelligent management platform enhances O&M efficiency, enabling unified, visualized management of storage devices and achieving the optimal balance between service continuity and operating expenses.

V.  Conclusion

By moving beyond external gateways, embracing native all-flash architecture, and implementing seamless migration and intelligent management, MEF has done more than just solve performance bottlenecks. It has established a benchmark for reliability and infrastructure resilience.

The all-flash transition is just the first step. MEF is already looking toward the horizon, with plans to expand its all-flash backbone into a geo redundant 3DC DR framework. Through higher density storage and deep cloud integration, MEF is building an elastic, intelligent, and sustainable digital tax and finance brain. This will enable MEF's continued leadership in government service transformation in Morocco and the broader North Africa region.

Access Data Dialogue, Issue 002 →