Capacity is the easy part. Latency is what you feel
SAN, NAS and direct-attached storage designed against your real IO pattern — tiering, IOPS and the growth curve worked out before anyone quotes you a shelf of disks.
Storage, end to end
SAN
Block storage over fibre channel or iSCSI for virtualisation and databases, with multipathing configured properly rather than left at default.
NAS and file services
Shared file storage with quotas, snapshots and permissions that reflect how the business is actually organised.
Direct-attached and hyperconverged
Where a shared array is not justified, local or hyperconverged storage sized to the node and the failure domain.
Tiering and caching
Flash where latency matters, capacity disk where it does not, and an honest answer about which of your workloads is which.
Snapshots and replication
Array-level protection that complements your backup rather than being mistaken for it.
Monitoring and growth
Capacity and latency trended, so you expand on a plan instead of on a Sunday night.
IOPS, latency, growth
Storage bought on price per terabyte is the most common expensive mistake in infrastructure.
1. Measure the IO
Read and write mix, block size, queue depth and peak latency from the live estate. A workload profile, not a capacity number.
2. Design the tiers
What genuinely needs flash, what can sit on capacity disk, and where caching changes the answer.
3. Plan the growth
Sized for the growth curve and for a drive failure and rebuild, which is when arrays are slowest and most vulnerable.
4. Build and verify
Installed, multipathed, tested under load, and documented with the expected performance written down so it can be held to.
Common questions
Do snapshots count as backup?
No. A snapshot on the same array does not survive the array, ransomware that reaches the storage layer, or the site. It is fast recovery, not protection. See backup infrastructure.
Our storage is slow. Is it full?
Often it is not capacity but queue depth, block size mismatch, a rebuild in progress, or a single misconfigured path. We measure before we recommend anything.
Can you support an array somebody else installed?
Yes, starting with documenting what is actually configured, including the multipathing and the firmware position.
SAN or hyperconverged?
It depends on node count, growth pattern and how your team operates. Both are right somewhere. We will show you the working.
Send us your array model and a busy hour.
Performance counters from a genuinely busy period tell us more than any capacity report.
02 3537 5791