How to choose a storage system: SAN, NAS, all-flash and hybrid
Servix technical team · Updated
In brief. Choosing a storage system comes down to four decisions, and they need to be made in exactly this order: access type (block for virtualization and databases, file for documents), interface (SAS is the cheapest, iSCSI is more flexible, Fibre Channel is more predictable under load), media (all-flash for response time, hybrid for capacity) and headroom for growth. In the Servix catalog, the price of the entry-level HPE MSA 2060 array with SAS is on request, the same array with iSCSI is from 93,084,000 UZS, and with Fibre Channel from 88,277,000 UZS. The difference between SAS and FC in the same chassis is the price not of the drives but of the storage network.
Jump to: access type: block or file · SAS, iSCSI or Fibre Channel · all-flash or hybrid · storage for virtualization · compression and deduplication · fault tolerance · growth planning · catalog prices · what to check · questions and answers
Start by deciding on the access type
This is the first decision, and it determines everything else. There are two options, and mixing them up is expensive.
- Block access (SAN). The array presents a “disk” to the server, and the server creates the file system itself. This is how hypervisors, databases and clusters work: they need a block device, not a folder. Protocols: SAS, iSCSI, Fibre Channel.
- File access (NAS). The array presents a ready-made network folder over SMB or NFS. This is how file archives, shared documents and backups work. Access permissions and the file system live on the storage side.
A practical rule: if the storage is needed by a hypervisor or a DBMS, choose block. If users need it for documents, choose file. Many arrays can do both, but paying for versatility makes sense only when both scenarios are needed at the same time.
A separate note on NAS: in a small office, the role of file storage is often covered by an ordinary server with drives and a properly built RAID — this is cheaper than a separate array and easier to maintain. A breakdown of the levels and the rebuild arithmetic is in the guide on RAID and the disk subsystem.
SAS, iSCSI or Fibre Channel
For block access, you choose from three interfaces, and the difference between them is not speed as such but how many servers can be connected and what happens under load.
Interface | How many servers | What else is needed | When it is chosen |
|---|---|---|---|
SAS | usually 2–4 directly | only cables and HBAs | One or two servers next to the rack; the cheapest entry point |
iSCSI | as many as the network can handle | 10/25G network adapters and dedicated switch ports | A virtualization cluster when there is no separate storage network |
Fibre Channel | dozens | FC HBAs and FC switches | Predictable latency under load, large clusters |
The main mistake with iSCSI is running storage traffic through the same switch and the same ports as the regular network. While the load is light, everything works; at backup time latency rises, and databases start “slowing down for no reason.” For iSCSI, dedicated ports are allocated, or better yet a separate switch, and jumbo frames are enabled along the entire path. Which adapters and modules this requires is covered in the article on network adapters and transceivers.
The cost is visible in the catalog: MSA 2060 SAS for 12 drives — price on request, the same chassis with iSCSI — 93,084,000 UZS, with Fibre Channel — 88,277,000 UZS. The drives and enclosure are the same; the difference is the controllers and what the array can be connected to.
All-flash or hybrid
All-flash is an array built only on SSDs. Hybrid is SSDs for active data and high-capacity HDDs for the rest, with automatic movement between tiers.
- All-flash is chosenwhen response time is critical: databases, terminal servers, virtual desktops. What matters here is not megabytes per second but latency and the number of operations.
- Hybrid is chosenwhen there is a lot of data but only a few percent of it is hot: file archives, backups, video surveillance. The cost per terabyte decides.
The trap with hybrid is that automatic movement of data between tiers is not instantaneous. If the load changes every day, the data does not have time to “warm up,” and the array keeps responding at the speed of the slow tier. For such profiles, it is more honest to go all-flash with a smaller capacity from the start.
A reference point for the cost per terabyte from the catalog: SSD Intel D3-S4520 7.68 TB — 10,782,352 UZS, drive WD Ultrastar 22 TB SAS — 8,625,882 UZS. The gap per terabyte is several-fold, and it is this gap, not the datasheet speeds, that determines the choice.
Storage for virtualization: what to calculate
The most common scenario and the most common source of mistakes. You need to calculate three things.
- Operations per second, not capacity. Twenty virtual machines with databases generate random load that high-capacity HDDs cannot handle, no matter how many of them there are. For virtualization, SSDs or a hybrid with a sufficient flash tier are used.
- The path to the array. Two paths from each server, two controllers in the array, multipath access on the hypervisor side. One cable or one controller is not a fault-tolerant cluster but an expensive single point of failure.
- Headroom for snapshots. Snapshots live on the same array and consume capacity faster the more actively the data changes. You must not plan for utilization above 80%.
Compression and deduplication
Both technologies save space, but in different ways and not equally useful everywhere.
Deduplication pays off where data repeats: a dozen identical virtual machines, backups, images. On unique data — photos, video, already compressed archives — it gives almost nothing but still consumes processor time.
The second thing people forget: the ratios claimed by vendors are calculated on profiles that suit them. You cannot plan capacity using a ratio from a presentation — calculate by raw capacity and treat the savings as a nice bonus. If the data does not compress, space will run out at the worst possible moment.
Fault tolerance: what to check
- Two controllers. Not “supported” but actually installed. An array with one controller is a server with external drives, and if the controller fails, everything stops.
- Two paths from each server and configured multipath access. This is checked by disconnecting one cable on a test bench, not on paper.
- Two power supplies and two different power lines, not two cables in one extension cord.
- A spare drive in the chassis — a hot spare, so that the rebuild starts automatically. Why this matters and how long a rebuild takes is covered in the guide on RAID.
- Protection level. On high-capacity drives, this is RAID 6 or its vendor equivalent: when a 16–22 TB drive is being rebuilt, the array runs without redundancy for almost two days.
Growth planning
An array is bought for five years, while data grows every year. Two rules that save money in the second year.
Look at the expansion shelf, not just the enclosure. On MSA and PowerVault, capacity is expanded with shelves, but the shelf must be of the matching generation, and the number of shelves is limited by firmware. The connection rules are covered in the reference guides on HPE MSA and on Dell PowerVault. In the catalog, the MSA 2050 SAS 12LFF shelf costs 24,800,000 UZS, Dell PowerVault MD1400 — 33,000,000 UZS.
Count more than terabytes. Growth in the number of servers runs into the controller ports, and growth in load runs into their performance. An array that has run out of host ports can only be expanded by replacing the controllers, which effectively means buying a new one.
Catalog prices as of November 2026
Model | Interface and chassis | Price |
|---|---|---|
SAS, 24 × 2,5″ | price on request | |
SAS, 12 × 3,5″ | price on request | |
iSCSI, 12 × 3,5″ | 93,084,000 UZS | |
Fibre Channel, 12 × 3,5″ | 88,277,000 UZS | |
Fibre Channel, 24 × 2,5″ | 112,711,000 UZS | |
SAS, two SSDs included | price on request | |
Fibre Channel, 12 × 3,5″ | price on request | |
expansion shelf | 24,800,000 UZS | |
SAS expansion shelf | 33,000,000 UZS |
MSA configurations by generation are in the selection HPE MSA 2060 and 2062, the Dell lineup is in the selection PowerVault ME5.
What to check before buying
- Block or file access — everything else depends on this.
- How many servers connect today and how many in three years: SAS will limit growth before capacity runs out.
- Whether dedicated ports are allocated for iSCSI and whether jumbo frames are enabled.
- Two controllers, two paths, two independent power inputs — actually present, not “supported.”
- Protection level on high-capacity drives — RAID 6 or equivalent, plus a hot spare.
- An expansion shelf of the same generation and the firmware limit on the number of shelves.
- Capacity is calculated as raw, without counting the promised compression, and planned for no more than 80% utilization.
- A backup outside the array: replication between controllers and snapshots are not backups.
We will size an array for your load — with the number of operations, connection paths, shelves and headroom for growth — and tell you honestly if the task can be covered more cheaply by a server with drives than by a separate storage system. Write to us.
Questions and answers
What is the difference between SAN and NAS in simple terms?
A SAN presents a block device to the server — a “disk” on which the server itself creates the file system; this is how hypervisors and databases work. A NAS presents a ready-made network folder over SMB or NFS, and the file system lives on the storage side; this is how shared documents and archives work.
Which storage system should I choose for virtualization?
Block storage, with two controllers and two paths from each host, on SSDs or a hybrid with a sufficient flash tier. You need to count operations per second, not terabytes: twenty virtual machines create random load that high-capacity HDDs cannot handle in any quantity.
What is an all-flash array and when is it needed?
An array built only on SSDs. It is needed where latency is critical: databases, terminal servers, virtual desktops. If there is a lot of data but only a few percent of it is active, a hybrid is cheaper — provided the load profile is stable and the data has time to reach the flash tier.
iSCSI or Fibre Channel — which to choose?
iSCSI is cheaper and does not require separate infrastructure, but it has to be placed on dedicated ports, otherwise storage traffic competes with regular traffic. Fibre Channel costs more up front and provides predictable latency under load — it is chosen for large clusters and databases where spikes in response time are unacceptable.
How much does a storage system cost in Tashkent?
In the Servix catalog, HPE MSA 2060 with SAS is priced on request, with iSCSI from 93,084,000 UZS, with Fibre Channel from 88,277,000 UZS; an expansion shelf is from 24,800,000 UZS. The final amount depends on the interface, the number and type of drives, and the number of shelves.
Can a server with drives be used instead of a storage system?
For file storage in a small office, yes, and it is cheaper. A separate array is needed when several servers connect to the storage, controller-level fault tolerance is required, or the capacity no longer fits in the server chassis.
Related reading
SAN, NAS or all-flash: how to choose a storage system for your company
How block storage differs from file storage, when a SAS array is enough and when you need NVMe. A review of HPE MSA, Dell PowerVault ME5, PowerStore and Huawei OceanStor with prices in Tashkent.
Storage system or NAS: what to choose for an office and a company
NAS, file server or HPE MSA and Dell ME5 storage: what to choose for files, backup, a cluster, a video archive and databases. Price per terabyte from the catalog as of November 2026.
Dell PowerVault ME5 or HPE MSA 2060/2062: which storage system to choose for an office and virtualization
Dell ME5024 and ME5012 vs HPE MSA 2060/2062: 32G and 16G FC, iSCSI, SAS, ADAPT and MSA-DP+, licenses, choosing a storage system for virtualization with 2–4 hosts, and prices.
RAID 5, 6 or 10: what to choose for a storage system and why it depends on the drives
Why RAID 5 is dangerous on high-capacity drives, why RAID 10 is faster on writes, and how to choose the array level for HDD, SSD and NVMe. Practical recommendations with calculations.




