Disadvantages of using a 4-server and 1 external storage architecture

As far as I understand, Sangfor HCI operates with a VSAN-like architecture and does not require external storage. However, I am planning to deploy it using a 4-server and 1 external storage architecture. What are the disadvantages of this setup? Would you recommend deploying it with the standard architecture? What would be the disadvantages? Is it reasonable to use it in a production environment?

You are correct in your understanding that Sangfor HCI, like many other Hyperconverged Infrastructure (HCI) solutions based on software-defined storage (SDS) similar to VMware vSAN, is architected to pool the direct-attached storage (DAS) of the HCI nodes to create a distributed storage layer. This design fundamentally aims to eliminate the dependency on traditional external shared storage arrays (like SAN or NAS) for the primary storage pool used by the virtual machines (VMs).

Now, regarding your plan to deploy with a 4-server (node) architecture plus 1 external storage unit:

If the intention is to use that external storage unit as the primary storage pool managed by Sangfor's SDS layer for running your VMs, this represents a non-standard and generally unsupported configuration for the core HCI storage function.

Disadvantages of This Non-Standard Setup (Using External Storage for Primary HCI Storage):

Undermines Core HCI Value Proposition: The primary benefit of HCI is the convergence of compute and storage management into a single, simplified layer. Reintroducing a separate external storage array creates a distinct storage silo, negating a key advantage of HCI.
Single Point of Failure (SPOF): The external storage unit becomes a critical SPOF for data availability. If the external array fails or connectivity is lost, all VMs whose data resides on that external pool will become unavailable, irrespective of the health of your 4 HCI nodes. Standard HCI distributes data across multiple internal disks within multiple nodes, providing resilience against individual component or node failures through data redundancy (e.g., replication).
Increased Operational Complexity: You will have two separate infrastructure stacks to manage: the Sangfor HCI cluster for compute and virtualization, and the traditional management interface/processes for the external storage array. HCI aims for unified management.
Potential Performance Bottlenecks: Performance in standard HCI benefits from data locality (VMs accessing data on their local node or nearby nodes) and distributed I/O across many disks and nodes. Accessing data from a single external array over a network (FC, iSCSI, NFS) can introduce latency and become a performance bottleneck, limiting the aggregate I/O capabilities compared to the distributed internal storage pool.
Limited Utilization of Sangfor SDS Features: Sangfor's advanced data services (such as inline deduplication, compression, snapshots, clones, distributed caching, and potentially specific replication features) are designed and optimized to operate on its distributed storage layer built upon the nodes' internal disks. These features may not function correctly or at all when data resides on a foreign, externally managed storage array.
Vendor Support Implications: Deploying Sangfor HCI with the primary VM storage residing on an external array not managed by their SDS layer is highly likely to fall outside of Sangfor's standard supported configurations. Troubleshooting issues related to storage performance or availability in this setup could be problematic, with limited or no vendor support.
Recommendation:

I would strongly recommend deploying with the standard Sangfor HCI architecture, utilizing the internal direct-attached storage within the 4 servers to form the distributed storage pool managed by Sangfor's software.

Advantages of the Standard Architecture (4 Nodes with Internal Storage Pooling):

True HCI Benefits: Simplified infrastructure procurement, deployment, and lifecycle management via a unified platform.
Built-in Resiliency and High Availability: Data is automatically replicated and distributed across the nodes' internal storage, providing fault tolerance against disk or node failures without relying on a separate hardware appliance.
Linear and Predictable Scalability: You can scale compute, memory, and storage capacity/performance together simply by adding more HCI nodes to the cluster.
Optimized Performance: Leveraging high-speed internal buses and a distributed I/O path provides excellent performance characteristics for virtualized workloads.
Unified Management Plane: All compute, storage, and network virtualization resources are managed from a single interface.
Full Feature Set Utilization: You gain the full benefit of Sangfor's data efficiency (dedupe, compression), data protection (snapshots, replication), and other SDS features operating seamlessly.
Disadvantages of the Standard Architecture (Relative):

Coupled Scaling: While generally linear, you typically scale compute and storage together when adding nodes. This might be less granular than separate server and storage array purchases if your growth needs are highly asymmetrical (e.g., needing significantly more CPU but very little storage, or vice versa).
Initial Node Configuration: Each node must be initially configured with sufficient internal disk capacity to meet current and near-future storage needs, which can impact the per-node cost upfront.
Platform Dependency: The storage layer is intrinsically tied to the Sangfor HCI platform. While data migration within the cluster is easy, migrating large amounts of data off the Sangfor platform to a completely different infrastructure vendor would require a specific migration process.
Is it Reasonable to Use it in a Production Environment?

Using the proposed architecture (4 nodes + external storage for primary HCI storage) is generally not reasonable or recommended for critical production environments due to the significant disadvantages listed above, particularly the SPOF, lack of resilience at the storage layer, and potential support issues.
Using the standard Sangfor HCI architecture (4 nodes pooling internal storage) is highly reasonable and precisely what the platform is designed for in production environments. Its built-in resiliency, performance, scalability, and unified management make it a robust and reliable platform for hosting business-critical applications.

Using Sangfor HCI with external storage deviates from its intended architecture, which is designed to leverage local storage on each node for optimal performance and simplicity. Here are some potential disadvantages to consider:

Disadvantages of Using External Storage
- *Performance Overhead*: Adding external storage might introduce latency and reduce overall system performance due to network bottlenecks or storage access times.
- *Complexity*: Managing an external storage system adds complexity to your infrastructure, potentially increasing the risk of configuration errors or compatibility issues.
- *Cost*: Depending on the external storage solution, this could increase the overall cost of your HCI deployment.

Standard Architecture Advantages
- *Optimized Performance*: Sangfor HCI is optimized for local storage, providing high performance and low latency.
- *Simplified Management*: The standard architecture is designed to be easy to manage, with streamlined configuration and monitoring.
- *Cost-Effectiveness*: Utilizing local storage can be more cost-effective than relying on external storage solutions.

Production Environment Considerations
- *Supported Configuration*: Verify that Sangfor supports your proposed architecture. Using an unsupported configuration might void warranties or affect support.
- *Scalability and Flexibility*: Consider whether your chosen architecture will meet your future scalability and flexibility needs.

Recommendation
Given the potential disadvantages, it's worth considering the standard Sangfor HCI architecture, which leverages local storage on each node. This approach is likely to provide optimal performance, simplicity, and cost-effectiveness.

If you have specific requirements that necessitate external storage, carefully evaluate the trade-offs and ensure that your chosen solution meets your needs. It's also recommended to consult with Sangfor support or experts familiar with your specific use case to determine the best approach.

Hi, just adding to the points already mentioned. Using a 4-server HCI cluster with an external storage array as the main storage backend isn't typically aligned with how Sangfor HCI is designed to work. The platform is built around pooling internal disks across nodes to form a distributed storage layer, so relying on an external array bypasses this core architecture. This setup can introduce performance issues due to increased latency and lack of data locality, especially under load. It also creates a potential single point of failure if the external storage goes down, VMs dependent on it will be affected regardless of the health of the HCI nodes. From a management perspective, it adds complexity by introducing a separate storage system to maintain alongside the HCI cluster. Additionally, such a configuration may not be fully supported by Sangfor, particularly for production workloads, which could limit access to vendor support and advanced features. Unless you have a very specific reason to use external storage, the standard HCI model is generally the more resilient and efficient approach.