Data Center · HyperFlex
Size a HyperFlex cluster for VMs, VDI, or databases. Accounts for Replication Factor overhead, cache ratio, and the leaf fabric behind the cluster.
HyperFlex sizing has three levers: node model (hybrid vs all-flash vs all-NVMe), replication factor (RF2 or RF3), and workload mix. Undersized cache crushes IOPS. Undersized capacity runs you into compression-dependence. This tool avoids both.
Output shows required effective capacity, raw capacity after RF overhead, recommended node count, and cache ratio check. It'll flag under-cached all-flash and over-dense hybrid deployments.
Effective capacity = VMs × avg GB × headroom. Raw = effective × RF ÷ dedup_gain. Leave 25% free for snapshots, cloning, and failure recovery rebuild. Cache sizing: all-flash needs roughly 400 GB cache per 8 TB capacity (5%); hybrid needs SSD cache sized for working set — typically 10-15% of capacity.
RF3 is non-negotiable for production > 4 nodes. RF2 is fine for edge or ROBO 3-4 node clusters but you lose a node = cluster read-only.
Only if you need CPU/RAM beyond the storage-bound ratio. Most deployments don't benefit.
Adds RF4 overhead and a witness. Size each site as if it's full RF3; the "stretched" layer sits on top.
Effective 26 TB (with 30% headroom). Raw 39 TB after RF3 ÷ 2 dedup. 4 × HX240c AF nodes (10 TB usable each) comfortably handle it with RF3 and ~25% free space.
Effective 52 TB. Raw 52 TB after RF3 ÷ 3 dedup. 6 × HX240c AF nodes for dedup ratio and IOPS headroom.