SAN vs NAS for Small Business: A Bangalore Buyer's Guide

The ₹9 lakh mistake in a Peenya server room

In March 2024 we walked into a 90-person engineering firm off Tumkur Road. They had bought a two-controller SAN from a reseller in SP Road: 24 bays, dual 16Gb fibre channel switches, the works. Invoice value was ₹9.4 lakh for the array, plus ₹1.8 lakh for the switches and HBAs. The workload it served? A single Windows Server 2019 file share, Tally data, and a Synology RS1221+ that had been demoted to "backup".

Eighteen months later, the finance head asked why storage spend was three times the office's entire annual IT AMC. Nobody had an answer. The SAN worked fine. It was simply the wrong instrument — a scalpel used to open an envelope.

We see a version of this roughly every six weeks in Bangalore. The opposite also happens: a company tries to run a 40-VM SQL cluster off a ₹80,000 four-bay NAS, and then wonders why their ERP takes 14 seconds to post an invoice.

This piece is about choosing correctly. Not the marketing version — the version with block sizes, IOPS numbers, and GST-inclusive quotes.

Block versus file, without the consultant voice

Every storage conversation eventually lands on this distinction. Most explanations make it worse. So let's do it properly.

File storage: you ask for a document

A NAS is a file server. When your accountant opens \\NAS01\accounts\FY26\Ledger.xlsx, her laptop sends a request over SMB (Server Message Block, the protocol Windows uses for file sharing). The NAS's operating system, usually some Linux variant, translates that path into physical disk sectors and hands back the bytes.

Your computer never knows where the file physically lives. It does not need to. It asked for a named object and got one.

This is how 90% of Indian SMEs actually work. Shared folders, mapped drives, QuickBooks or Tally data directories, a design team's Adobe files, HR scanning salary slips to PDF. All file-level.

Block storage: you ask for sector 4,918,220

A SAN does not serve files. It serves raw numbered blocks. When a Windows server connects via iSCSI, it sees a disk that looks exactly like a local SSD or HDD. The server's own filesystem (NTFS, ReFS, VMFS) sits on top of it.

The SAN has no idea what is on those blocks. It does not know you stored invoices there. It does not care. It just guarantees that when you write to block 4,918,220, you get those exact bytes back, fast, every time.

That ignorance is the source of a SAN's speed. There is no filesystem overhead, no permission checks, no SMB negotiation. And it is also why a SAN cannot deduplicate meaningfully across different hosts the way a NAS can — it does not know what it is looking at.

Why the difference matters operationally

Read that table again. Notice that nothing in it says "SAN is for grown-ups." The two columns describe different jobs.

What iSCSI actually does on your network

iSCSI is the reason SMBs can afford SANs at all. It takes SCSI commands — the language of disks — and shoves them inside TCP/IP packets. No fibre channel switches, no HBAs costing ₹60,000 each. You use your existing 10GbE or even 1GbE switching.

That convenience has a price. Every iSCSI byte competes with your office traffic for switch bandwidth and CPU cycles on the initiator.

A real deployment we did in Whitefield

A logistics company, 140 staff, running Microsoft Dynamics 365 Business Central on-premises plus 22 Hyper-V VMs. Their old setup: two Dell PowerEdge R740xd servers with local RAID-10 SAS, replicating via Hyper-V Replica to a third box. Storage was 18TB usable per host. When the primary host's RAID controller failed in November 2023 (Intel RS3DC080, five years old, out of support), failover took 47 minutes because the replica was 40 minutes stale and they had to hand-verify data.

We redesigned it around a HPE MSA 2062 with 24 x 1.92TB SAS SSDs, connected via 10GbE iSCSI to both Hyper-V hosts through two Cisco Catalyst 9300 switches. Separate VLAN for storage. Jumbo frames enabled end-to-end (MTU 9000). MPIO (Multipath I/O) configured with two paths per host.

The result: live migration between hosts dropped from 90 seconds to 11 seconds. A full 2TB SQL database backup that previously took 4 hours 20 minutes now finishes in 38 minutes.

The jumbo frame trap

Every iSCSI deployment eventually gets bitten by MTU mismatches. A switch port configured for MTU 9000, a Windows NIC left at 1500, and suddenly you have a 3% packet loss rate that only shows up under load. Latency seems fine during testing. Then a Hyper-V host tries to live-migrate during a peak hour and the whole cluster hiccups.

We test MTU end-to-end with ping -f -l 8972 from every initiator to every target in the design phase. Every time. Even when the customer says it's already configured. Especially then.

iSCSI on 1GbE is a lie you will eventually pay for

Some resellers will sell you a 1GbE iSCSI SAN because the price is lower. At 1GbE you top out around 110MB/s of raw throughput. A SAN serving four VMs is fine. A SAN serving 20 VMs backing up nightly is a bottleneck you will find at 2 AM when the backup windows overlap.

Rule of thumb: if you are buying a SAN in 2026, budget for 10GbE or 25GbE iSCSI minimum. Fibre Channel is not worth the complexity for SMBs anymore, unless you inherit an existing SAN fabric you want to extend.

When a small business genuinely needs a SAN

I have said no to more SAN projects than I have said yes to. Here is the honest test.

You actually need a SAN if all of these are true

1. You run 10 or more production VMs on two or more hypervisor hosts, and those hosts need shared storage to enable live migration and HA. Local RAID cannot do this. A NAS on NFS can, but performance starts to bite past about 15 VMs.
2. You have a database workload (SQL Server, Oracle, PostgreSQL) that needs consistent low-latency random I/O. If your ERP posts a transaction in under 200ms because your LUN sits on NVMe or SAS SSD, the SAN is the reason.
3. You need to survive a single controller failure with no downtime. Dual-controller SANs fail over silently. NAS boxes at the SMB price point usually do not.
4. You have some regulatory or contractual reason to isolate storage from the general office LAN. Some BFSI clients in Bangalore insist on it.
5. You are spending more than ₹6 lakh on storage across two or three failed NAS attempts. At that point, just buy the right tool.

You probably do not need a SAN if any of these are true

• You have one host. One host cannot exploit a SAN's HA benefit.
• Your biggest storage consumer is a file share full of Word, PDF, and Excel files.
• Your virtualisation estate is under 10 VMs, or under 5 TB usable.
• Your IT person is a single generalist who also handles printers and CCTV.
• Your business closes at 7 PM and nobody works weekends.

The last point is underrated. A NAS with nightly backups and a 4-hour RTO is fine for a business that does not operate 24/7. A SAN makes sense for businesses where those 4 hours cost ₹5 lakh in lost orders.

What a SAN actually costs in 2026

All prices are GST-inclusive estimates for Bangalore, based on quotes from HPE, Dell, NetApp partners we work with. Range reflects enterprise-discount level, whether you buy on a new fiscal cycle (March), and whether you bundle support.

Entry SAN for 10-20 VMs

A NAS that does 70% of the same job

The gap is real but narrower than people expect. A SAN costs roughly 40-60% more than an equivalent-performance NAS once you are honest about switches, cabling, and support. That difference buys you dual controllers, better IOPS consistency under load, and vendor support SLAs that matter during a Monday morning outage.

Where SAN pricing gets ugly is at scale. A NetApp AFF A250 with 30TB usable, dual-controller, will start around ₹22-28 lakh in 2026. That is a different conversation and not one most SMBs should be having.

A worked cost comparison at two company sizes

Numbers get abstract. Let's ground them.

Scenario A: 35-person CA firm in Jayanagar

Workloads: 8 VMs (Tally server, two application servers, AD, file server, backup, two test VMs). Total usable storage needed: 6TB, growing 15% annually. Staff use mapped drives and shared folders all day.

The CA firm gets nothing from a SAN. Their VMs are not HA-critical (they can tolerate a 2-hour morning recovery), their I/O profile is mostly sequential reads (tax software loading PDFs), and a four-hour window for planned restores is fine. We deployed the Synology with 4 x 16TB in SHR-2. It has run without incident for two years.

Scenario B: 180-person manufacturer in Peenya

Workloads: 32 VMs (12 SQL Server, 4 Tally, 6 app servers, AD, print, 6 desktops, backup), all on a 3-node Hyper-V cluster. 24TB usable needed. Nightly ERP dispatch runs, 3 shifts, operations 20 hours a day.

The manufacturer went with the MSA. Why not the cheaper NAS? Two reasons. First, dual controllers. If a controller fails at 3 AM during dispatch, the NAS bricks its volumes until someone physically reseats or replaces hardware — the SAN fails over in under 60 seconds with zero admin intervention. Second, we ran a 96-hour I/O trace and the SQL Server workloads were pushing 14,000 random write IOPS during end-of-month close. The QNAP NAS we tested held that for 40 minutes before latency drifted from 1.8ms to 22ms. The MSA held it flat.

The failure story that changed how we quote

November 2022. A 60-person logistics firm on Hosur Road. Existing setup: a QNAP TS-873 with 8 x 8TB WD Red in RAID-6, serving a single Hyper-V host running 11 VMs including SQL Server Express and a bespoke TMS application. Backups: nightly QNAP snapshot to a second identical unit.

What broke: the primary QNAP's backplane. Not the drives. Not the RAID. The physical connector board that the drive trays plug into. One day the array simply stopped seeing disks 3, 5, and 7. QNAP support took 11 days to ship a replacement unit from Taiwan, and even then, the RAID metadata was on the degraded unit.

Cost during the outage: 9 working days at an estimated ₹62,000 average daily gross margin contribution from dispatch operations, roughly ₹5.6 lakh. Plus 22 hours of engineer time flying in from SynergyScape and QNAP technicians at ₹4,500/hour, another ₹99,000. Plus a panic purchase of a replacement unit because they could not wait — ₹3.7 lakh for a TS-h1277AXU and drives.

What fixed it: we moved the workload to a dual-controller HPE MSA 2062 with SAS SSDs (12 units), kept the QNAP as a secondary target for Veeam Backup & Replication v12, and added an offsite copy to a Wasabi bucket via a Veeam SOBR (Scale-Out Backup Repository). Total spend to rebuild: ₹11.4 lakh. Total avoided in future single-point failures: hard to quantify, but the CFO signed off in one meeting.

The lesson was not "never buy a NAS." The lesson was: NAS units at the SMB price point have single points of mechanical failure, and a business whose daily revenue depends on the data being live cannot tolerate an 11-day parts wait. RTO of 8 hours is fine for a CA firm. It is not fine for a logistics firm.

Where NAS wins outright — and refuses to be beaten

I do not want the previous section to read as a NAS indictment. Four things a NAS does better than a SAN, no argument.

1. SMB/CIFS sharing with permissions

Try building a shared folder structure with per-user ACLs on a SAN. You cannot, because there is no filesystem. You will end up putting a Windows file server in front of the SAN, which means you paid SAN prices for a file server anyway. A NAS does this natively, correctly, with Active Directory integration that works on the first try in most cases.

2. Backup targets

Every Veeam user in India eventually buys a Synology, QNAP, or TrueNAS box as a hardened repository. Immutable snapshots, WORM support on the Synology RS3621RPxs and QNAP TS-h1290FX, S3-compatible APIs. These are the parts of a NAS that make ransomware recovery possible. SANs are typically bad here because block snapshots are crash-consistent only — you get the block state, not the file state.

3. Snapshot granularity and self-service restore

Synology's Snapshot Replication and QNAP's Snapshot Manager let end users restore yesterday's version of a single Excel file without an IT ticket. On a SAN, snapshots are LUN-level. Restoring needs the storage admin, a mount, and usually a copy operation. Different tools for different problems.

4. Cost-per-terabyte at scale

A Synology DS1823xs+ with 6 x 20TB WD Red Pro provides 80TB usable (RAID-6) for about ₹5.3 lakh in 2026. The cheapest SAN with comparable usable capacity lands around ₹18-22 lakh. For archive, cold data, media files, and document repositories, that NAS is the obvious answer. No SAN salesperson can argue with the math.

Bangalore-specific things nobody mentions until they bite

If you are deploying either a NAS or SAN in Bangalore, a few local realities deserve airtime.

ISP and last-mile upgrades take weeks

If your iSCSI design includes a second 10GbE uplink to a remote site via MPLS or SD-WAN, budget 4-8 weeks for the circuit. ACT Fibernet and Airtel have improved enterprise provisioning but do not expect same-week delivery on a new layout. Deploy storage first, WAN second, then migrate the remote replication traffic.

Monsoon and cabling

Every July-September we get called to at least two sites where water ingress into a rooftop conduit has corroded a CAT6A run that was powering a 10GbE link. Check your cable runs to the storage rack before monsoon. If your 10GbE iSCSI link between the SAN and the hosts runs through an exterior wall in a building older than fifteen years, inspect it. A single bad link does not fail loudly — it just drops to 100Mbps or flaps at 1Gbps, and your MPIO path goes to sleep.

UPS and dirty power

Storage arrays are more fragile to power quality than servers. Both NAS and SAN units sold for SMB environments expect clean 230V input and graceful shutdown signalling via USB or serial. Every deployment we do includes a networked UPS — typically an APC Smart-UPS SRT 3000 or an Eaton 9PX 3000 with a network management card — configured to trigger clean shutdown at 20% battery. Do not trust the array's own internal detection. It is always a step too late.

GST and capitalisation

Storage hardware attracts 18% GST. If your business is GST-registered and the spend is capitalised, the input tax credit is recoverable and effectively reduces your net outflow. The opex portion — vendor support contracts, backup software subscriptions — attracts 18% GST as services. If you are buying in a new fiscal year, timing matters: hardware purchased on or before 31 March can be depreciated in the outgoing year under the Income Tax Act's 40% block for computers and storage, depending on your auditor's view. Talk to your CA before you sign a March purchase order.

CERT-In and DPDP compliance

The DPDP Act's storage-location and log-retention requirements are still shaking out, but the present CERT-In directions (issued April 2022) require logs to be retained for 180 days and stored in India. For any NAS or SAN you deploy in a Bangalore office, this means: keep the storage itself in-country (no cross-border replication to a Singapore SAN without a compliance opinion), and configure your array logs plus any SMB audit logs to ship to a local syslog collector. Logs on the array alone will not survive an array failure.

Physical security

Bangalore server rooms sometimes double as stationery cupboards. If your NAS or SAN is visible from a common area or the door does not lock, plan for a rack with a lockable front panel. A 1U Synology on a shelf in a shared room is a theft risk. We have seen it happen twice.

The decision framework we actually use

When a client comes to us pre-sales, we run this triage.

Step 1: Count your IOPS genuinely

Before picking hardware, run Windows Performance Monitor or esxcli storage core device stats get on your existing setup for 7 days. Record peak IOPS, peak throughput, and the busiest 15-minute window. Do not trust averages. Storage failures happen at peaks.

Step 2: Define your RTO and RPO in minutes, not words

"Four hours" and "we cannot lose more than one hour of data" are real numbers. "High availability" is not. If your RTO is under 1 hour and your RPO is under 15 minutes, you are in SAN territory or you need a proper NAS HA pair like the Synology FS2500 or QNAP's dual-controller units. If RTO is 4 hours, nearly any modern NAS with good backups works.

Step 3: Price the outage

What does one hour of downtime cost your business? If the answer is under ₹10,000, buy the NAS. If it is over ₹50,000, the price gap between NAS and SAN starts to disappear in a single incident. If it is over ₹2 lakh, dual-controller SAN is not a question.

Step 4: Match the tool to the workload list

Cross-shopping the actual boxes on the shelf in 2026

Named products only. No "leading enterprise storage solutions."

NAS options worth evaluating

• Synology DS1823xs+ — 8 bays, 10GbE onboard, NVMe cache slots. Around ₹2.1 lakh chassis-only in Bengaluru. Rock-solid DSM 7.2. Best for general SMB file sharing and small VM use.
• Synology RS3621RPxs — 12 bays, redundant PSU, 10GbE, rack. Around ₹4.8 lakh chassis-only. The most common intermediate step before a SAN.
• QNAP TS-h1290FX — 12 bays, all-NVMe, 25GbE. Around ₹6.9 lakh chassis-only. Fast enough to make you question whether you need a SAN at all. QTS is less polished than DSM but the hardware is serious.
• TrueNAS M40 — 16 bays, dual controller option, built on ZFS. Around ₹9 lakh chassis-only when sourced through Indian partners. Excellent if you have an engineer comfortable with ZFS.
• Synology FS2500 — active-active dual-controller NAS. Around ₹11.5 lakh chassis-only. The most SAN-like NAS you can buy. If you want dual controllers and file protocols, this is the answer.

SAN options worth evaluating

• HPE MSA 2062 — entry dual-controller SAN, 12-24 bays, 16Gb FC or 10/25GbE iSCSI. Our most-deployed SMB SAN in Bangalore. Around ₹7.5 lakh chassis-only, 12-bay SSD config in the ₹14-17 lakh range fully loaded.
• Dell PowerVault ME5024 — comparable to MSA, sometimes cheaper on Dell-loyalty accounts. Sanjaynagar clients often have Dell anchors so this becomes the path of least resistance.
• NetApp AFF A250 — for larger SMBs who need real enterprise features. ₹22 lakh plus fully configured. Overkill for most, correct for a few.
• Dell PowerStore 500T — NVMe-native, unified block and file. ₹24 lakh plus. If budget allows and you want a single box that does both NAS and SAN correctly, this is the current sweet spot.

The honest take on each

The Synology DS1823xs+ is the most reliable SMB storage box at its price in 2026. QNAP's TS-h1290FX is faster but has a steeper operational learning curve. The MSA 2062 is boring in the best way — it just works, and HPE support in Bangalore responds within four hours on NBD contracts. Anything over ₹20 lakh should really justify itself in an SMB context, and often does not.

When SynergyScape is not the right call

Straight talk: if you can install and manage a Synology yourself, or your in-house IT person has done it before, you do not need us for the initial setup. Buy the box from a distributor in SP Road, watch the 40-minute YouTube video, and configure it in an evening. It will work.

We are worth calling when:

• You are moving from local server storage to shared storage for the first time and asking the wrong question ("Which SAN should I buy?") — you might need us to talk you into a NAS.
• You are running a hypervisor cluster and the HA design will cost real money when it goes wrong.
• You need to integrate storage with an existing Veeam deployment, WORM compliance, or DPDP/CERT-In logging requirements.
• You have three offices across Bangalore and need storage replicated in a way that survives a site down.
• Your previous storage project went badly and you want it audited before you spend again.

We are not worth calling if you need a ₹40,000 two-bay NAS for a 6-person office. Buy it online, we will help you set it up if you ask, but you do not need a managed-services relationship for one drive shelf. Full stop.

If you want to talk through a specific situation — the workloads, the numbers, the compliance angle — our storage team at SynergyScape's storage solutions practice does this every week. And if you want to just send us your current setup and ask us for a fit assessment, reach out via our contact page.

Three misunderstandings that cause most bad purchases

Misunderstanding 1: "SAN is faster than NAS"

Not inherently. A QNAP TS-h1290FX all-NVMe NAS at 25GbE will outrun an entry 10GbE iSCSI SAN every time. The performance difference between block and file at the protocol level is a few percent. What actually decides performance is the underlying media (SSD vs HDD), the network (10GbE vs 1GbE), and the array's controller and cache. Buy those right and either architecture flies.

Misunderstanding 2: "A SAN is more reliable than a NAS"

A dual-controller SAN is more reliable than a single-controller NAS. A dual-controller NAS (Synology FS2500, QNAP TS-h1290FX with HA) is more reliable than a single-controller SAN (which mostly does not exist in the SMB segment but you find used units on the grey market). Compare reliability by controller count and support SLA, not by category label.

Misunderstanding 3: "We need a SAN to support VMware or Hyper-V"

Both hypervisors run perfectly well on NAS-backed iSCSI or NFS datastores. Every SMB VMware deployment we did in 2024-25 used either local SSD in a vSAN configuration or iSCSI to a NAS or MSA. VMware's own HCL includes many NAS units for NFS datastores. The hypervisor does not care. Your workload profile cares.

FAQ

Is a SAN overkill for a small business with under 50 staff?

Usually yes. If you have fewer than 10 production VMs, no HA requirement, and standard file-share workloads, a NAS on 10GbE handles it comfortably for a fraction of the cost. The exceptions are businesses with high-transaction databases or strict uptime windows — a 40-person trading firm or a 24/7 dispatch operation can justify a SAN where a 40-person consultancy cannot.

Can I use a NAS as a SAN?

Most mid-range and higher NAS units today offer iSCSI target functionality, which means any host can connect to a LUN on the NAS and treat it as a block device. You are effectively running a SAN on NAS hardware. Performance is limited by the NAS's CPU and the network path, but for 10-20 VMs it works. Synology, QNAP, TrueNAS and Netgear all support this. It is the single best cost-saver in SMB storage today.

What is iSCSI in plain English?

iSCSI carries disk commands over regular Ethernet instead of dedicated fibre cabling. Your server sees a virtual disk, sends and receives block data over TCP/IP, and the array responds. You need 10GbE for serious workloads and proper MPIO configuration to avoid single points of failure. Once configured, iSCSI is invisible to the operating system — it just thinks it has a fast local disk.

How much does a SAN cost for a small business in India in 2026?

An entry dual-controller SAN from HPE or Dell, with 12-24 SSD bays, 10GbE switching, and 3-year support, runs ₹13 lakh to ₹19 lakh GST-inclusive. Larger configurations with more drives or Fibre Channel start at ₹25 lakh and go up quickly. NAS alternatives at comparable performance cost ₹9 lakh to ₹14 lakh.

Which is better for backup storage — SAN or NAS?

NAS. Always NAS. Immutable snapshots, S3-compatible targets, deduplication that actually sees file content, and cost-per-terabyte that is two to three times better. Almost every Veeam deployment we do uses a NAS for the primary repository plus cloud or a second NAS for offsite. SANs are bad backup targets because block snapshots cannot deduplicate across servers meaningfully.

Can I run a NAS and SAN together?

Yes, and this is often the right answer. Put your high-transaction VMs on SAN LUNs, your file shares and backup repositories on a NAS, and use Veeam or Commvault to copy between them. This gives you the performance of block storage where it matters and the cost efficiency of file storage everywhere else. Most of our Peenya and Whitefield clients run exactly this hybrid pattern.

What to do this quarter

Stop reading and start measuring. The single best thing you can do before spending any money on storage is to run a 7-day monitoring pass on your current setup. Windows Performance Monitor, ESXi's esxtop, Hyper-V's perfmon counters — pick what you have. Record peak IOPS, peak throughput, average and 95th-percentile latency, and storage growth in GB per week.

If you get numbers that look like under 5,000 peak IOPS and under 4TB growth per year, buy a NAS and stop worrying about SANs. If you get numbers above 15,000 random write IOPS with sustained latency under 5ms, you have a legitimate SAN case, and the investment will pay for itself in one prevented outage.

If you want help interpreting the numbers before you sign a purchase order — the kind of review where the output is a page of concrete recommendations and a quote aligned to what you actually need — that is the work our storage solutions team does. Send us your monitoring data and your workload list. We will tell you which box you do not need.