跳到正文
VGroup.IT Solutions
数据中心中级7 分钟阅读

NAS、SAN 还是云:网速不稳时公司数据放在哪里

NAS、SAN 与云在协议、性能和成本上的区别,附真实上传时间计算,以及适合老挝网络条件的混合架构。

NAS、SAN 还是云:网速不稳时公司数据放在哪里
本页内容
  1. Three options for three jobs
  2. Comparison
  3. Which one for which workload
  4. NAS: where most SMEs start
  5. SAN: when servers need shared storage
  6. Cloud: good, but think about the line
  7. The number people forget: first upload time
  8. A design we use often
  9. Sizing the NAS
  10. RAID level and the rebuild window
  11. Misunderstandings that lose data
  12. Drives in 2025–2026: what changed
  13. What's different in Laos
  14. What we recommend

Three options for three jobs

NAS (Network Attached Storage) shares files over the network using SMB (Windows) or NFS (Linux). Staff see shared folders, like the S: drive everyone knows.

SAN (Storage Area Network) presents block storage to servers over iSCSI, Fibre Channel or NVMe-oF. Servers see it as their own disks. It's built for virtualization and databases.

Cloud storage means keeping data with a provider: file sync (OneDrive, Google Drive), S3-style object storage, or an offsite backup target.

Comparison

Topic NAS SAN Cloud
Access level File Block File / object
Protocols SMB, NFS iSCSI, FC, NVMe-oF HTTPS / API
Who uses it Staff, general apps Servers, hypervisors, databases Remote staff, offsite backup
Complexity Low High Low–medium
Needs internet No No Yes
Cost model One-off purchase (CapEx) High CapEx Monthly/annual (OpEx)

Which one for which workload

Workload Where we would put it
Shared office folders, scanned documents NAS, with cloud backup
Large CAD, GIS or survey files used every day NAS on the LAN; cloud only for the backup copy
Files that site teams and partners edit together Microsoft 365 or Google Workspace
Virtual machines on two or more hosts Dual-controller SAN, HCI or an iSCSI/NFS business NAS for small clusters
ERP or accounting database Fast local storage (SSD) on the server or SAN, never a file share
CCTV recording NVR or a NAS sized for surveillance, kept apart from office files
Archive you open a few times a year Cloud object storage or an older NAS
The offsite copy of everything Cloud backup or a second NAS in another building

Most organizations end up with two or three of these at once. That is normal: the goal is the right home for each kind of data, not one box for everything.

NAS: where most SMEs start

For an office of 10–200 people that needs to share files, a NAS is the simplest answer. Construction firms and architects with CAD drawings of several hundred MB notice the difference straight away when files open from a NAS on the LAN instead of downloading from the cloud. A NAS also makes a good on-site backup target, and many business models offer iSCSI for small virtualization setups.

SAN: when servers need shared storage

A virtualization cluster that needs live migration and high availability needs shared storage. So does an ERP or core banking database that needs high IOPS and low latency. The key point: a SAN must have dual controllers, or it becomes the single point of failure for every server attached to it.

An option we see more and more is hyperconverged infrastructure (HCI): compute and storage in the same set of servers instead of a separate SAN.

Cloud: good, but think about the line

Cloud suits staff spread across many sites, the offsite copy in a 3-2-1 backup, and old data that's rarely opened.

In Laos there are two things to weigh. Upload speed on most business connections is far lower than download, and many provincial sites have only one provider. If that line goes down for a day, nobody can open their files. Before putting sensitive data in the cloud, agree internally which data may leave the building and which must stay on hardware you own.

The number people forget: first upload time

2 TB of data over a 50 Mbps upload: 2,000,000 MB × 8 ÷ 50 = 320,000 seconds ≈ 89 hours (about 3.7 days), if every second of bandwidth goes to the upload. In reality it takes longer because staff are using the line during the day. After that only changes (incrementals) go up, which are much smaller. The scarier number is a full restore coming back down the same line on the day ransomware hits.

Check your link with our bandwidth calculator.

A design we use often

An 80-person company, head office in Vientiane plus three provincial sites:

  • A 4–6 bay NAS in RAID 6 for central files and backups
  • Microsoft 365 or Google Workspace for shared editing and the site teams
  • NAS backup to the cloud every night after 22:00, when the line is quiet
  • Hourly NAS snapshots for recovering from ransomware or accidental deletion

Sizing the NAS

  • 3 TB of data today, growing 25% a year
  • In 3 years: 3 × 1.25³ ≈ 5.9 TB
  • Allow for snapshots and previous versions (roughly × 2) → about 12 TB usable
  • RAID 6 on 6 × 4 TB = 16 TB usable (about 14.5 TiB in the system). Enough, with room to spare

RAID level and the rebuild window

With today's large drives, the dangerous moment is not the first failure. It's the rebuild that follows. Every remaining drive is read end to end while the array is still serving users.

A rough sum for one 16 TB drive at an average 150 MB/s: 16,000,000 MB ÷ 150 ≈ 107,000 seconds, about 30 hours, and longer when people are working on the NAS. During that time:

RAID level Can survive during rebuild Our view
RAID 1 / RAID 10 Depends which drive fails next Good for 2-bay units and fast workloads
RAID 5 No further drive failure Fine for small drives; avoid with drives of 8 TB and up
RAID 6 / RAID-Z2 One more drive failure Our default for 4 bays and up with large drives

Two more habits help. Don't fill every drive from the same production batch if you can avoid it, because drives from one batch tend to age together. And keep a spare drive on the shelf, so a rebuild starts the same day instead of after an order from Thailand.

Misunderstandings that lose data

  1. Believing RAID or file sync is a backup. It isn't. A wrong delete or ransomware syncs everywhere.
  2. Exposing the NAS straight to the internet so people can reach it from home.
  3. Buying a single-controller SAN and calling it "high availability".
  4. Never working out how many days a full restore from the cloud would take over the line you actually have.

Drives in 2025–2026: what changed

  • Supply is tight. Since late 2025, demand from AI data centres has pushed hard drive and SSD prices up sharply and stretched lead times, especially for large-capacity drives. Order drives with the NAS, not afterwards, and treat spares as part of the purchase.
  • Vendor drive lists. Synology's 2025 Plus models at first accepted mainly Synology-branded drives. With DSM 7.3 in October 2025 it restored normal support for third-party SATA hard drives and SSDs; M.2 drives used for storage pools still need to be on its compatibility list. Check the current list for whatever brand you buy.
  • Use CMR, not SMR, drives. Some NAS-labelled drives use shingled recording (SMR), which rebuilds very slowly in RAID. Choose drives the maker specifies as CMR for NAS use.

What's different in Laos

  • Brownouts and shutdown. A NAS that loses power mid-write can damage its file system. Put it on a UPS and connect the UPS by USB or network so the NAS shuts down cleanly before the battery runs out.
  • Heat. Drives in a hot cupboard fail early. Keep the NAS where the air is under about 30 °C if you can, and watch drive temperatures in the NAS dashboard.
  • Offsite doesn't have to mean cloud. If upload is slow, replicate the head-office NAS to a second NAS at a branch or a director's office in another building. It still gives you a copy away from fire and flood, and a restore over a local drive is fast.
  • Test the restore. Once a quarter, restore a real folder from the offsite copy and time it. That number is your recovery time, whatever the brochure says.

What we recommend

For most SMEs, NAS plus cloud backup is the best value and the easiest to look after. Organisations running multi-server virtualization should look at a dual-controller SAN or HCI. Always keep critical copies on site, because restoring over the LAN is tens of times faster than over the internet. And turn on snapshots and immutable backups so ransomware can't delete your copies.

Tell us three things: how many TB you have today, how many people and sites, and your internet upload speed. We'll size the NAS and show you the backup and restore times.

常见问题

Is a NAS with RAID enough as our backup?

No. RAID keeps the NAS running when a drive fails, but it copies deletions, corruption and ransomware encryption instantly. You still need snapshots on the NAS and at least one copy somewhere else, either in the cloud or on a second device in another building.

Can we use ordinary desktop hard drives in a NAS?

It works for a while, but we don't recommend it. Desktop drives aren't designed for 24-hour operation or the vibration of several drives in one box, and many use SMR recording that makes RAID rebuilds painfully slow. Use NAS or enterprise drives rated for the number of bays you have.

Do we need a SAN for two or three virtualization hosts?

Not always. Small clusters can use a business NAS over iSCSI or NFS, or hyperconverged storage built into the hosts. A dual-controller SAN makes sense when downtime is expensive and the workload is heavy. If you are leaving VMware because of licensing, decide the storage design together with the new hypervisor, since each handles shared storage differently.

How long would a full restore from the cloud take?

Divide the data size by your real download speed. 2 TB over 100 Mbps is roughly 44 hours of continuous transfer, before any slowdown. That is why we keep a local copy for fast restores and use the cloud copy for disasters, and why some providers offer to ship large restores on a disk.

  • #storage
  • #nas
  • #san
  • #cloud
  • #iscsi
  • #backup

相关解决方案

相关指南

发给我们建筑平面图或几张现场照片

我们会先询问一些细节,再给出初步设计方案。首次咨询免费,可用老挝语、泰语或英语,通过 WhatsApp 或 Facebook 主页联系。