Once you have decided you need a SAN, the next question is how the servers reach it: iSCSI over ordinary Ethernet, Fibre Channel over a dedicated storage fabric, or — for one or two hosts — direct-attached SAS. The old framing that Fibre Channel is "enterprise" and iSCSI is "budget" is long out of date; on 10 or 25 GbE, iSCSI runs serious production storage, while Fibre Channel remains the choice where deterministic low latency and a fully isolated fabric matter most. This guide explains how each works, where each wins, and why the entry SANs we supply let you pick whichever fits — without locking you in.
What each protocol actually is
iSCSI carries SCSI block commands inside TCP/IP, so a server's storage traffic travels over Ethernet just like any other network data. The server uses a software initiator (or a hardware one) and the array presents block volumes (LUNs) it can format and own. Because it rides standard Ethernet, iSCSI reuses the switches, cabling and skills you already have, and it routes — you can reach storage across subnets and sites.
Fibre Channel is a purpose-built storage fabric. Servers carry FC host bus adapters (HBAs), connect through FC switches, and the fabric is zoned so each host sees only its own storage. It is lossless and deterministic by design — there is no TCP retransmission behaviour to reason about — which is why it has carried mission-critical block storage in large data centres for decades. The trade is that it is a second, separate network with its own hardware, switches and skill set.
Performance and latency in practice
On paper Fibre Channel tops out higher (16 and 32 Gb FC are common on current arrays) and is inherently lossless, giving very consistent latency under load — the reason latency-sensitive databases and the largest virtualisation estates still favour it. In practice, iSCSI on a properly designed 10 or 25 GbE network — isolated VLANs or dedicated NICs, jumbo frames, multipathing — delivers more than enough performance for the overwhelming majority of SMB and mid-market workloads, and 25 GbE iSCSI closes much of the headline gap.
The single biggest performance mistake with iSCSI is running it across a shared, congested, best-effort LAN. Give storage its own physical or logical network with redundant paths and the difference versus Fibre Channel narrows to something most workloads never notice. Where you genuinely cannot tolerate a latency spike — tier-one databases, very large consolidation — Fibre Channel's deterministic behaviour earns its keep.
- •iSCSI — block over Ethernet (10/25 GbE): reuses your network + skills, routes between sites, lowest cost to adopt.
- •Fibre Channel — dedicated lossless fabric (16/32 Gb): deterministic latency, fully isolated, the tier-one default.
- •SAS direct-attach (12 Gb): simplest and cheapest for one or two hosts cabled straight to the array — no fabric at all.
- •Design beats protocol: a well-isolated iSCSI network outperforms a badly-zoned FC one for most workloads.
Cost, skills and flexibility
iSCSI's advantage is that it adds little: Ethernet NICs, switches you may already own, and standard networking skills. That makes it the natural choice for organisations without a dedicated storage team and for multi-site setups where storage traffic needs to route. Fibre Channel adds HBAs, dedicated FC switches, optics and zoning — real capital and a specialist skill set — which is justified at scale and for the most demanding workloads, but is overhead a small estate rarely needs.
There is also a simpler option for the smallest deployments: direct-attached SAS. If only one or two servers need the array, 12 Gb SAS cables straight into the controllers remove the fabric entirely — no switches, no zoning, the lowest cost and complexity of all. It does not scale to many hosts, but for a two-node cluster or a single backup server it is often the right answer.
When to choose which
Choose iSCSI when you want flexibility and cost-efficiency, run on Ethernet, lack a dedicated FC fabric, or need storage to route between subnets or sites — which covers most SMB and mid-market SANs, especially on 25 GbE. Choose Fibre Channel when you have (or are building) a dedicated storage fabric, run latency-critical databases or a large virtualisation estate, or already have FC skills and switches and want the most deterministic option. Choose SAS direct-attach when only one or two hosts need the array and you want the simplest, cheapest path.
Crucially, this is rarely a permanent lock-in. The entry SANs we supply offer all three host types, so you can start on iSCSI and adopt Fibre Channel later, or run a mix. Our HPE MSA offers 16/32 Gb FC, 10/25 GbE iSCSI and 12 Gb SAS; the Dell PowerVault ME5 offers 32 Gb FC, 25 GbE iSCSI and 12 Gb SAS — and you can specify the connectivity in the MSA or PowerVault configurator.
On the entry SANs we supply
Because the MSA and PowerVault ME5 support FC, iSCSI and SAS on the same array, the protocol decision becomes a sizing and budget conversation rather than a platform commitment. For a typical SMB virtualisation cluster we most often land on 10/25 GbE iSCSI for its flexibility and lower total cost; for latency-sensitive or larger estates with existing fabrics, Fibre Channel; and for a two-node or single-host setup, direct SAS. If you are choosing between the two array families themselves, our HPE MSA vs Dell PowerVault ME5 comparison covers it, and the storage end-of-life checker helps if you are replacing an older SAN.