For nearly a decade, VMware vSAN relied on the Original Storage Architecture (OSA), which required dividing drives into distinct "Disk Groups" comprising one dedicated write cache SSD and up to seven capacity drives.
With the advent of uniform high-speed TLC/QLC NVMe drives and 25/100GbE host fabrics, the cache-tier became a severe bottleneck. VMware redesigned the entire storage engine in vSAN 8 with the Express Storage Architecture (ESA).
1. Death of Disk Groups: The Single-Tier Storage Pool
In vSAN ESA, all qualified NVMe devices participate in a unified, single-tier Storage Pool. Every device simultaneously serves reads, writes, and metadata. If a drive fails in ESA, you lose only that single drive's capacity—unlike OSA, where the failure of a single cache SSD took offline the entire disk group of 7 capacity drives.
2. The Log-Structured File System (LFS) Engine
At the heart of ESA is a new patented Log-Structured File System (vSAN LFS). Incoming writes are ingested in sequential log appends directly into non-volatile memory, bypassing random in-place updates. Writes are acknowledged immediately to the guest VM with sub-millisecond latencies.
3. RAID-5/6 Erasure Coding with RAID-1 Performance
In traditional storage systems, RAID-5 incurs a 4-to-1 write penalty (read data, read parity, write data, write parity). vSAN ESA changes the game:
- Writes are ingested as full stripe writes in the log without pre-reading old parity.
- RAID-5 (3+1) or RAID-6 (4+2) delivers the exact same throughput and IOPS as double-mirrored RAID-1, saving 50% of raw storage capacity without performance compromise.
4. Native High-Performance B-Tree Snapshots
In legacy vSphere (VMFS/vSAN OSA), taking VM snapshots created -delta.vmdk redo logs. Deep
snapshot chains caused severe I/O degradation and lengthy stun times during snapshot consolidation.
ESA replaces redo logs with a native copy-on-write B-Tree metadata pointer system. Creating, reverting, and deleting snapshots in ESA happens in milliseconds with zero guest stun, even under heavy transactional databases.
5. CLI Inspection & Verification
You can verify ESA storage pool health and component placement directly from the host shell: