Technical Specifications and Limitations

Performance

Runs on SSD Standard in shared mode. All metrics scale linearly with provisioned capacity.

Benchmark: iozone, 4 TiB SSD Standard, 5 clients, 10 GB/host, 20 runs, caches cleared between runs, outliers excluded.

Throughput

Metric	Per TiB
Peak sequential read	~370 MB/s
Peak sequential write	~319 MB/s

A 4 TiB cluster delivers ~1,480 MB/s sequential read / ~1,275 MB/s sequential write.

Read throughput is buffer-sensitive: when the active dataset exceeds the block storage read buffer, sustained throughput drops to ~850 MB/s (1 MiB block size, 4 TiB). Write throughput is not buffer-sensitive.

IOPS (4 KiB block size)

Metric	Per Client	Aggregate per 4 TiB
Read IOPS	~4,000	24,000
Write IOPS	~470	18,000

Reaching aggregate write IOPS requires ~372 concurrent operations at ~20 ms latency — above the default NFS RPC slot table limit (64–128 per mount). Per-client figures assume a single mount.

Latency

Operation	Mean	Median	P99+ (worst observed)
Read	2.8 ms	2.8 ms	292 ms
Write	20.7 ms	21.9 ms	2,187 ms

Write latency is synchronous to an active-passive replicated backend. P99 spikes correlate with block storage flush and GC events.

Resilience

Storage runs across two servers with automatic failover. Each server uses hardware or software RAID.

Data security

Network File Storage is based on Block Storage and inherits its data security properties, including encryption at rest and secure deletion. For details, see Data Security.

Known limitations

The following constraints apply:

• Cluster size cannot be decreased after provisioning

• Minimum cluster size is 2 TiB

• Write IOPS are constrained by synchronous write latency (~20 ms); achieving aggregate write IOPS requires concurrency levels that may exceed default NFS client RPC slot table limits

• Not suitable for synchronous write workloads with tight timeout requirements