Problem Description
IBM FlashSystem provides advanced enterprise storage technologies, including hardware compression, encryption, data protection, and performance optimization.
However, I believe there is a significant architectural limitation in the current Distributed RAID (DRAID) expansion mechanism that can have serious financial and operational consequences for customers.
When an existing DRAID array is expanded, IBM Storage Virtualize preserves its original stripe width. Consequently, an expanded array may provide substantially less usable capacity than a newly created array with the same number of drives.
This limitation becomes particularly concerning when customers invest in additional IBM FlashCore Modules (FCMs), which represent a significant financial investment.
Real-World Example
During an IBM FlashSystem 7300 capacity expansion:
- Original configuration: 6 FCMs in DRAID6.
Expansion: 6 additional FCMs.
Additional raw physical capacity: 52.38 TiB.
Actual increase in usable physical capacity: 30.99 TiB.
Only 59.2% of the additional raw capacity became available for user data.
We fully understand that RAID6 requires capacity for parity, rebuild areas, and other internal reservations. These requirements are not being questioned.
The real concern is that the original RAID geometry can significantly limit the efficiency of future expansions, even when customers invest in additional storage hardware.
Financial and Operational Impact
Customers purchase storage capacity to meet specific business requirements.
FlashCore Modules are expensive enterprise components, and the financial impact becomes increasingly significant as organizations acquire higher-capacity drives.
IBM FlashSystem offers remarkable technologies for compression, encryption, and data protection. However, the value of these innovations can be undermined when customers cannot efficiently utilize the physical storage capacity they have purchased.
More importantly, the current architecture does not allow administrators to optimize stripe width during expansion.
To recover potentially significant usable capacity, customers may be forced to:
- Obtain additional temporary storage to accommodate existing production data.
Migrate all volumes away from the original array.
Delete and recreate the DRAID array with an optimized geometry.
Migrate all data back to the FlashSystem.
For enterprise customers, this represents significant additional costs, administrative effort, operational risks, and potential service disruptions.
Why should customers have to rebuild their entire storage configuration simply to achieve better capacity efficiency from hardware they already own?
Proposed Enhancement
I strongly encourage IBM to consider the following enhancements to Storage Virtualize:
1. Dynamic DRAID Geometry Optimization
Introduce the ability to modify or optimize stripe width during array expansion, ideally through a supported, nondisruptive process without requiring external data migration.
2. Accurate Capacity Forecasting
Before initiating an expansion, clearly display the expected additional usable capacity, RAID overhead, rebuild reservations, and final capacity efficiency.
3. Expansion Alternatives Comparison
Allow administrators to compare the expected usable capacity of expanding an existing array versus creating a new array with the same number and type of drives.
4. Explicit Architectural Warnings
Provide clear warnings when the existing RAID geometry will significantly reduce the capacity efficiency of an expansion.
Final Considerations
Enterprise storage systems must evolve to support predictable, efficient, and economically sustainable capacity growth.
Customers should not discover significant architectural limitations only after purchasing additional drives and completing a storage expansion.
Nor should they be forced to undertake complex data migrations and array reconstruction simply to optimize the capacity of their existing hardware.
An enterprise storage architecture must protect data, but it must also protect the customer's investment.
I believe addressing this limitation would represent a meaningful improvement to IBM Storage Virtualize and provide substantial value to the entire IBM FlashSystem customer community.
