Server drives: HDD or SSD, SAS, SATA or NVMe — what to buy for the DL380 and R750
Servix technical team · Updated
In brief. For databases and virtual machines, choose SSDs; for files and backups, large-capacity SAS HDDs. In Tashkent, capacity on 8–22 TB drives costs 294,000–431,000 UZS per terabyte, and a terabyte on NVMe costs about 2,246,000 UZS (Servix catalog, November 2026). RAID 5 is not used on drives of 8 TB and up: a rebuild takes a full day. The controller is selected for the drive type — NVMe needs a tri-mode controller; a regular SAS controller will not see them.
What goes in the server bays: LFF or SFF
First, look not at the drive but at the drive cage. In HPE ProLiant and Dell PowerEdge servers, the cage is fixed when the chassis is ordered and can only be changed by replacing the backplane, so a mistake here is expensive.
- LFF, 3.5 inches. Eight or twelve bays in 2U. This is where 8–22 TB drives go. This is about terabytes: archive, file share, backup repository, video surveillance.
- SFF, 2.5 inches. Eight, sixteen or twenty-four bays in 2U, eight or ten in 1U. This is where SAS 10k, SSD and NVMe go. This is about operations per second: virtualization, 1C, databases.
A drive is installed in a drive tray. HPE and Dell drive trays differ, and they also differ between generations, so a drive without a tray simply will not lock into the bay. When ordering, this is a separate line item and it is easy to forget.
HDD or SSD, SAS, SATA or NVMe: what limits what
Type | Strengths | Use case | Catalog example | Price | Per terabyte |
|---|---|---|---|---|---|
HDD SAS 7,2k | price per terabyte, dual-port connection | files, backup, video | 3,234,705 UZS | 404,000 UZS | |
HDD SAS 7.2k, large | maximum capacity per bay | archive and backup repository | 8,625,882 UZS | 392,000 UZS | |
HDD SATA | cheaper than SAS at the same capacity | single server, non-critical data | 6,469,411 UZS | 294,000 UZS | |
SSD SATA | affordable flash capacity | warm data, file cache | 10,782,352 UZS | 1,404,000 UZS | |
NVMe U.3 | latency and operations per second | DBMS, 1C, virtualization | 4,312,941 UZS | 2,246,000 UZS |
Three things distinguish a server drive from a regular one. The first: SAS is dual-port — it is visible to both controllers, so it is used in fault-tolerant configurations and in storage systems, whereas SATA is single-port. The second: CMR versus SMR recording — SMR drives are cheaper, but their performance collapses under random writes, and they are not used in RAID. The third: server SSDs have power loss protection and a rated write endurance, while consumer ones do not.
A separate note on NAS drives such as the WD Red Plus 10 TB (price on request) and the WD Red Plus 4 TB for 1,868,941 UZS. They are designed for home and office storage: 5400–7200 rpm and a workload of about 180 TB written per year. For a server under constant load, choose the Ultrastar line — it is designed for 24/7 operation and for many times higher annual writes.
How many drives and which RAID for your task
- 1C and databases. Four or more SSDs in RAID 10. Transaction logs generate random writes in small blocks; what matters here is latency, not total terabytes.
- Virtualization. Six to eight SSDs or NVMe drives in RAID 10, a separate boot volume for the hypervisor. RAID 5 under a mixed workload carries a write penalty large enough to spoil virtual machine responsiveness.
- File server and archive. RAID 6 of four or more large drives. Double parity is not overcaution: a 16–22 TB drive takes more than a day to rebuild, and a second failure within that window is quite likely.
- Backups. RAID 6, maximum-capacity drives, a hot spare is a must. The workload is sequential; the speed of an individual drive is not the main thing here.
A detailed comparison of the levels is in the guides RAID and the server disk subsystem and RAID 5, 6 or 10 for storage systems. If several servers need the data at once, drives in a server no longer solve the task — see the breakdown of SAN, NAS and all-flash.
Which controller to choose
The controller determines which drives the server will see at all and how fast it will write. Three boundaries matter when choosing.
- Cache and battery. Models without cache (E208 from HPE, H345 from Dell) are suitable for RAID 1 and 10 or for HBA mode. RAID 5 and 6 need cache with battery protection; otherwise write performance collapses, and on power loss the data in the cache is lost.
- Tri-mode. Only controllers with tri-mode support work with NVMe U.2 and U.3 — these are the MR416i from HPE and the H755 from Dell. A classic SAS controller simply will not see an NVMe drive.
- Internal or external. The suffix i means ports facing inside the chassis, e means ports facing outside, for connecting a disk shelf. The P408e-p and E208e-p are external ones.
Controller | Cache and battery | Drives | Price |
|---|---|---|---|
no | SAS, SATA, RAID 1 and 10 | 4,312,941 UZS | |
yes | SAS, SATA, all RAID levels | 5,750,588 UZS | |
yes | SAS 12G, SATA | 7,188,235 UZS | |
yes | SAS, SATA and NVMe (tri-mode) | 8,625,882 UZS | |
no | SAS, SATA, RAID 1 and 10 | 5,750,588 UZS | |
yes | SAS 12G, SATA, all levels | 7,188,235 UZS | |
yes | external ports for a shelf | 7,188,235 UZS | |
yes | NVMe and SAS 12G (tri-mode) | 9,344,705 UZS |
If the server has more bays than the controller has ports, an expander is installed between them — for example, the HPE 12G SAS Expander for 7,188,235 UZS. This is cheaper than a second controller, but the bandwidth is shared among all the drives behind it.
Where to put the hypervisor: the boot volume
Giving two cage bays to the system is wasteful, so separate modules with a pair of M.2 drives in a mirror are used for booting: Dell BOSS-S2 with two 480 GB M.2 SATA drives (for 15th generation PowerEdge) for 7,907,058 UZS and HPE NS204i-p with two 480 GB NVMe drives for 10,063,529 UZS. The mirror is built by the module itself, the hypervisor sees one disk, and all cage bays remain for data. For ESXi, Proxmox and Hyper-V this is the standard scheme.
Why you should buy drives in vendor drive trays
A server works with a third-party drive too, but with caveats. Vendor firmware adds health monitoring in iLO and iDRAC: failure prediction, error counters, automatic support case creation. With a third-party drive, this data is either incomplete or missing, and when you contact support the configuration is considered modified. Plus the mechanics: each generation has its own drive trays, and without them a drive does not lock into the bay and does not connect to the backplane.
Practical conclusion: for a production server, buy drives in trays for your generation and check compatibility before ordering. We do this free of charge — send us the server model and serial number.
Three ready-made sets with prices
- File server, 16 TB usable. Four WD Ultrastar 8 TB SAS in RAID 10 and an HPE Smart Array P408i-a — 20,127,055 UZS. Fits any 2U with an LFF cage.
- Virtualization, about 5.7 TB usable. Six Micron 7400 Pro 1.92 TB NVMe in RAID 10 and a tri-mode MR416i-p — 35,222,351 UZS. For thirty to forty virtual machines with an average load.
- Backup repository, about 96 TB usable. Eight WD Ultrastar 16 TB SAS in RAID 6 and a P408i-a — 47,442,347 UZS. Plus a hot spare if the bays allow.
The amounts are for the drive part: the server itself is calculated separately; ready-made configurations are on the pages HPE ProLiant Gen11 and Dell PowerEdge. Sizing for 1C is in the article server for 1C by number of users.
Warranty and lead times in Tashkent
All items in the drives and RAID controllers catalog come with a warranty of 12 months, serviced by Servix. Supply is made to order — 2–3 weeks, and the price is fixed in the invoice. We work by bank transfer with VAT, prepare specifications for tenders, and deliver throughout Uzbekistan. Selection of drives and a controller for your server model is free: write to us or call +998 92 182 11 44. Other components are in the Server componentssection.
Questions and answers
How much does a server drive cost in Tashkent?
An 8 TB SAS drive costs 3,234,705 UZS, 16 TB — 5,031,764 UZS, 22 TB — 8,625,882 UZS. A 1.92 TB NVMe U.3 drive costs 4,312,941 UZS, a 7.68 TB SATA SSD — 10,782,352 UZS (Servix catalog, November 2026).
Can a regular desktop drive be installed in a server?
Technically it will be detected, but it lacks dual-port SAS, endurance for 24/7 load and vendor firmware, and monitoring in iLO and iDRAC will be incomplete. For a production server, choose a server line in trays for your generation.
Which RAID to choose for 1C and virtualization?
RAID 10 of four or more SSDs. RAID 5 carries a write penalty, and on drives of 8 TB and up it is also dangerous because of the long rebuild — for such capacities RAID 6 is used.
Do you need a controller with cache and battery?
For RAID 5 and 6 — definitely: without cache, write performance collapses, and without a battery, the data in the cache is lost on power loss. For RAID 1 and 10 or HBA mode, a model without cache such as the PERC H345 or Smart Array E208i-a is enough.
Related reading
RAID and the server disk subsystem: levels, fault tolerance, controllers
RAID levels for a server: how RAID 4 differs from RAID 5, how many drive failures RAID 6, 10 and 50 survive, why RAID 5 is dangerous on drives of 4 TB and larger. With prices.
SFF or LFF: which drive bays to choose in a server
What 8SFF, 12LFF and 24SFF mean, which drives fit in 2.5″ and 3.5″ bays, when to choose LFF, when SFF and what NVMe requires. DL380 and R760 with 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.
A server for 1C: how to choose a configuration for the number of users
How to choose a server for 1C for 10, 30, 100 or more users: processor, memory, drives, ready-made HPE and Dell configurations with delivery in Tashkent.



