Server Setup for a Medium-Sized Business in India
By Rohit, Founder ·
Quick answer: A medium-sized business server setup is usually not one bigger server. It is a plan: list every workload (directory, files, ERP or database, email or collaboration, backup), decide how many physical machines they should share, add redundancy where a failure would stop work, and size CPU, ECC memory, storage and network from your own measurements. For most Indian companies with a few hundred staff, the practical starting point is one well-built virtualization host (an Intel Xeon Scalable server with ECC DDR4 memory, NVMe storage and 10GbE networking) plus a separate, tested backup target. Move to two or more hosts when the cost of an hour of downtime is higher than the cost of the second machine. This guide explains how to make each of those decisions with real facts, and where ProStation Systems' custom-built tower servers fit.
This guide is for IT managers, founders and operations heads at companies roughly in the 100 to 500 employee range in India. That band is our planning shorthand, not a legal definition; India's MSME classification uses investment and turnover, not headcount. If you are smaller, read our small business server guide first. Nothing here is a ProStation benchmark: ProStation publishes no performance figures for specific headcounts, so every size recommendation below is a method, and the arithmetic examples are labelled as illustrative.
What Is a Medium Business Server Setup?
A medium business server setup is the set of servers, storage, network, power and backup that runs a company's shared systems for a few hundred people. It differs from a small-office setup in three ways:
- More roles at once. Directory and login, file shares, an ERP or accounting database, internal web apps, email or collaboration, CCTV or document archives and backups all compete for the same hardware.
- Higher cost of an outage. When 300 people cannot work, the loss is the whole payroll for that hour plus missed orders. Our guide to the real cost of server downtime shows how to calculate your own number from revenue, payroll and incident costs.
- More obligations. The Digital Personal Data Protection Act requires a data fiduciary to take reasonable security safeguards to prevent a personal data breach, wherever the data is hosted. A backup plan, access control and patching are part of that, not extras.
Step 1: Which Workloads Will the Server Actually Run?
Start with a list, not a model number. Each workload stresses different parts of a server, which is what drives the specification.
| Workload | What it stresses most | Planning note |
|---|---|---|
| Directory, DHCP, DNS | Little; needs to be always on | Light, but everything else depends on it, so protect it first |
| File shares and document archives | Storage capacity, network | Capacity grows steadily; HDD or SAS tiers can hold older data (see the NVMe, SAS and SATA guide) |
| ERP, accounting or database | Storage latency, RAM, CPU single-thread speed | NVMe storage and enough RAM to cache hot data; check the vendor's own requirements |
| Virtual machines for apps and test systems | RAM and CPU cores | Add up the RAM of every VM, then leave headroom for the hypervisor |
| Remote desktop or terminal sessions | RAM per session, CPU | Needs Remote Desktop licences if Windows Server hosts the sessions |
| Backup target | Capacity, sequential write speed, network | Best kept on a separate machine from the data it protects |
Ask each application vendor for its documented requirements and your own monitoring data for current CPU, memory, disk and network use. Where you have no data, say so in your consulting call rather than guessing; the number you invent now becomes the bottleneck later.
Step 2: One Server, Two Servers or a Cluster?
This is the main decision. Redundancy inside one server and redundancy across servers solve different problems.
What redundancy inside one server covers
- Disk failure: RAID keeps the server running when a drive dies. See RAID levels explained. RAID is not a backup.
- Power supply failure: dual power supplies keep it running if one fails. See the redundant PSU guide.
- Memory errors: ECC memory corrects single-bit errors. See ECC vs non-ECC RAM.
What it does not cover: a failed motherboard, a failed CPU, a corrupted operating system, or the room losing power or cooling. For those you need a second machine or a fast rebuild plan.
What a high-availability cluster needs
Clustering restarts virtual machines on another host when one host fails. The platforms document real prerequisites. Proxmox VE's HA documentation lists these requirements: at least three cluster nodes (to get reliable quorum), shared storage for VMs and containers, hardware redundancy everywhere, and reliable "server" components. Broadcom's vSphere documentation describes vSphere HA clusters as ESXi hosts working together so that virtual machines can be restarted on other hosts, and its HA pages list further requirements such as shared storage and network settings; read the current checklist for your version before you design. The point is the same on both platforms: HA is a system of several hosts, shared storage and network, not a feature you switch on a single box.
| Design | Protects against | Still exposed to | Fits when |
|---|---|---|---|
| One server with RAID, dual PSU, ECC, plus separate backup | Disk, PSU and memory faults; data loss | Board or CPU failure (restore time depends on spares and backup) | An hour or half-day outage is painful but survivable |
| Two servers: primary plus a standby you restore or replicate to | Single-server failure, with some manual work | Failover time and replication gaps | You want faster recovery without a full cluster |
| Three or more hosts with shared storage (HA cluster) | Host failure with automatic restart | Shared storage and network become the new single points of failure unless also duplicated | Downtime cost is high and you have staff to run the cluster |
ProStation builds new, custom tower servers. If your design needs several hosts, ProStation's free consulting call is the place to work out how many machines and what shared storage you need. Our tower vs rack server guide covers the form-factor trade-off once you are past two or three machines.
Step 3: How Do You Size CPU, RAM, Storage and Network?
Use this order, because each step changes the next.
- RAM first for virtualization. Add the memory every virtual machine needs, add the hypervisor's overhead, and keep spare capacity so a restart or growth does not force a purchase. ECC DDR4 is standard across ProStation's tiers.
- Cores from the workloads. Count the vCPUs you will assign and remember that Windows Server licensing counts physical cores, so core count is also a software cost (see the licensing step below).
- Storage by type. NVMe for databases and VM disks, larger and slower disks for archives. Separate the data you must back up from the capacity you merely store.
- Network last, but do not skip it. A 1 Gb/s link carries about 125 MB/s at best, shared by everyone using it. A 10 Gb/s link carries about 1,250 MB/s. Our 1GbE, 10GbE and 25GbE guide explains switch and cabling requirements.
How ProStation's listed tiers map to this, as guidance only:
| Tier | Listed specs | Typical role in a medium business |
|---|---|---|
| Starter | Intel Xeon E, 16-64GB ECC DDR4, 1TB NVMe SSD, 1GbE / 10GbE | A secondary or branch role: directory, print, light file server, small internal app |
| Pro | Intel Xeon Scalable, 64-256GB ECC DDR4, 2TB NVMe SSD plus HDD options, 10GbE / 25GbE, Ubuntu / Windows / VMware / Proxmox | The usual single virtualization host or database server for a growing team |
| Ultra | Dual Intel Xeon Scalable, 256-512GB ECC DDR4, 4TB+ NVMe in RAID, 25GbE | A heavily consolidated host or compute-heavy workloads; more than most office roles need |
These are starting points. ProStation builds to your configuration through the server customizer, and our office virtualization host guide and virtualization use case show how to turn a VM list into a specification.
Step 4: What Backup Plan Should a Medium Business Use?
Backup is the part that decides whether an incident is an inconvenience or a crisis. CISA's data backup guidance recommends the 3-2-1 rule: keep three copies of important data (one primary and two backups), on two different media types, with one copy stored offsite. Add two practices to it: keep at least one copy that ransomware on your network cannot reach, and test a restore on a schedule, because an untested backup is only a hope.
Illustrative arithmetic, not a measurement: a 2 TB full backup over a 1 Gb/s link at the theoretical best of 125 MB/s needs about 16,000 seconds, roughly 4.4 hours. Over a 10 Gb/s link at 1,250 MB/s the same copy takes about 27 minutes. Real speeds are lower because of disk speed, compression and protocol overhead, and incremental backups move far less data, but the gap shows why backup traffic is a network-sizing question. Our NAS storage server guide and storage and backup use case cover building the target.
Step 5: How Does Software Licensing Change the Bill?
The server and the operating system are separate purchases. Microsoft licenses Windows Server Standard and Datacenter per physical core with a minimum of 8 core licences per processor and 16 per server, and most deployments also need a Client Access Licence for every user or device. A company with 300 staff accounts needs 300 user CALs on that basis, plus Remote Desktop Services CALs if staff work in remote sessions on the server. Standard covers two virtual machines per fully licensed host; Datacenter covers unlimited. The details are in our Windows Server licensing guide. If your applications do not depend on Windows, Linux with Proxmox or another hypervisor carries no per-core fee, though any Windows VMs on it still need licences. Always get CALs and the OS licence written into the quote.
Step 6: Power, Cooling and Room
A server that is correctly specified still fails if the room is not. Size a UPS to the real load and the runtime you need, plan for cooling, and keep the machine off the floor and away from dust. See UPS sizing for servers and workstations. Dual power supplies only help if they are fed from separate, protected circuits.
Step 7: Support, Warranty and Spares
For a company that cannot stop, the question is how quickly a failed part is replaced. Hardware from ProStation comes with a one-year warranty, with extended warranty available at additional cost, and 24/7 support; the warranty page sets out the details, and on-site support is offered in major cities including Delhi NCR and the others listed on the cities page. Ask any vendor for the response commitment in writing, and decide whether to hold spare drives or a spare power supply on site. Each server is stress-tested before delivery; the process is described in our build and stress-test guide.
Should a Medium Business Buy New, Custom-Built or Refurbished?
New custom builds suit companies that want exact specifications, current component warranty and a clean start. Refurbished enterprise servers suit tighter budgets and temporary or secondary roles, and often come with their own AMC options. Read new vs refurbished and custom-built vs off-the-shelf servers before you decide. If budget is the main constraint, our sister brand Serverwale sells tested refurbished tower servers and offers server AMC plans. For cost ranges see custom server cost in India.
Checklist: Questions to Settle Before You Order
| Question | Why it matters |
|---|---|
| Which workloads and how many users? | Drives RAM, cores, storage and CAL count |
| What does an hour of downtime cost? | Decides single server vs standby vs cluster |
| Which systems need Windows Server? | Sets licence cost and core-count trade-offs |
| Where is the second and third backup copy? | 3-2-1 and ransomware protection |
| Is the network 10GbE end to end? | A fast server on a slow switch is still slow |
| What power and cooling exist? | UPS runtime and room temperature |
| Who replaces a failed part, and how fast? | Warranty term and support commitment in writing |
| Has the vendor read your application list? | Compatibility and OS pre-installation |
Our custom server buyer checklist expands this into a pre-order list, and the enterprises and SMB page shows how ProStation approaches these deployments.
Where ProStation Systems Fits
ProStation Systems builds new, custom tower servers and workstations in India with Intel Xeon processors and ECC DDR4 memory, and pre-installs the operating system you choose. The practical help for a medium business is at the planning stage: bring your application list, headcount and backup needs to the free server consulting call, and get a configuration written against your real workloads instead of a generic model. Then specify it in the customizer or contact the team directly. Industry-specific examples are in our guides for hospitals and diagnostic centres, fintech teams and IT and SaaS companies.
Frequently Asked Questions
Q1. What server does a medium-sized business need?
It depends on workloads, user count and how costly downtime is. For many companies of a few hundred staff, one virtualization host with Intel Xeon Scalable processors, ECC DDR4 memory, NVMe storage and 10GbE networking, plus a separate backup target, is a sensible starting point. Size RAM and cores from your own workload list rather than headcount alone.
Q2. Do I need a server cluster for 200 to 500 employees?
Not necessarily. A cluster adds automatic recovery from host failure but also needs several hosts, shared storage and skilled administration. Proxmox VE's documentation lists at least three nodes and shared storage as HA requirements. If a few hours of downtime is tolerable, one well-built server with redundancy inside it and a tested backup may be enough.
Q3. How many CALs does a 300-person company need for Windows Server?
Windows Server needs a CAL for every user or device that accesses it, counted by user or device rather than simultaneous connections. If 300 staff accounts use the server, plan for 300 user CALs, plus Remote Desktop Services CALs if they connect through remote sessions. Confirm the count with an authorised Microsoft reseller.
Q4. Is RAID enough to protect company data?
No. RAID keeps a server running through a disk failure but does not protect against deletion, ransomware, fire or theft. Follow the 3-2-1 rule recommended by CISA: three copies, two media types, one copy offsite, and test your restores.
Q5. Should a medium business choose new or refurbished servers?
Choose new, custom-built hardware when you want exact specifications and a full component warranty; consider refurbished for secondary roles or tighter budgets. ProStation Systems builds new custom servers, and its sister brand Serverwale sells tested refurbished servers.
Q6. What warranty does ProStation provide on servers?
ProStation's hardware comes with a one-year warranty, with extended warranty available at additional cost, and 24/7 support. The warranty page has the full terms, so ask for any response-time commitment in writing before you order.
Final Recommendation
Write down the workloads, calculate what an hour of downtime costs, and choose the simplest design that keeps you inside that number: one redundant server with a tested 3-2-1 backup for most, a standby or cluster for those who cannot stop. Budget for licences, network and power as part of the server, not afterwards. When you are ready for a configuration, contact ProStation Systems or call +91 87962 44410 (WhatsApp: wa.me/918796244410) for free consulting, or browse the servers page.
Sources: Proxmox VE documentation, High Availability chapter (requirements: at least three cluster nodes, shared storage, hardware redundancy); Broadcom techdocs, vSphere Availability, Creating and Using vSphere HA Clusters; CISA, "Data Backup Options" (3-2-1 rule); Microsoft Windows Server licensing documentation (per-core minimums, CALs, RDS CALs, VM rights) as summarised in our licensing guide; Digital Personal Data Protection Act, 2023 (reasonable security safeguards). Check current vendor documentation before purchase.