Replacing several ageing servers with one newer system can reduce rack space, power consumption, cooling requirements, maintenance effort and hardware costs. Modern enterprise servers also provide far more CPU cores, memory capacity and I/O performance than systems purchased only a few generations ago.
But there is no universal answer to the question, “How many old servers can one new server replace?” The correct consolidation ratio might be 3:1, 5:1, 10:1 or something entirely different. It depends less on the number of physical machines you currently own and more on what those machines actually do.
Successful server consolidation therefore starts with workload measurement, not with selecting the largest processor available.
What Is Server Consolidation?
Server consolidation means reducing the number of physical servers required to operate an organisation’s workloads. In many environments, this is achieved through virtualisation: several virtual machines run independently on one or more powerful physical hosts.
A business that currently operates separate physical servers for file services, accounting software, databases, internal applications and web services may be able to migrate several of those workloads onto a smaller number of modern virtualisation hosts.
Microsoft identifies server consolidation as one of the benefits of Hyper-V, noting that it can reduce hardware acquisition and maintenance requirements as well as data-centre space, power and cooling costs.
Why Modern Servers Can Handle More Workloads
The capabilities of current enterprise processors are dramatically higher than those of many servers still operating in business environments.
For example, current AMD EPYC 9005 processors extend to as many as 192 CPU cores in a single socket. Intel also promotes consolidation as a major server-refresh use case for Xeon 6, citing an average 5:1 consolidation ratio against five-year-old servers under its tested scenarios, with higher ratios possible for some workloads.
These figures demonstrate the potential of modern hardware, but they should not be treated as a guaranteed consolidation ratio for every business. Real applications may be constrained by memory, storage performance, licensing, networking or availability long before CPU capacity is exhausted.
The Five Resources That Determine Your Consolidation Ratio
1. CPU Utilisation
Start by measuring processor utilisation across existing servers over a representative period. A server with 24 CPU cores does not necessarily require 24 cores in a replacement system.
Many older application servers spend much of their time at relatively low utilisation. Virtualisation allows several such workloads to share processing resources instead of maintaining dedicated physical processors that remain idle for much of the day.
However, average CPU utilisation alone is not sufficient. Capacity planning should also consider peak periods, sustained loads, application response times and future growth.
A system running at 15% most of the day but reaching 90% during month-end processing must be sized differently from one with a stable 15% workload.
2. Memory
Memory is frequently the practical limit in virtualised environments.
Imagine ten existing servers each using approximately 32 GB of RAM. Although their combined CPU demand might fit comfortably on a modern host, their workloads could require around 320 GB of active memory before considering the hypervisor, growth, caching and resilience requirements.
Fortunately, modern enterprise platforms can support very large memory configurations. When selecting a consolidation server, however, the important question is not simply its maximum supported RAM. Consider memory-channel population, DIMM capacities, future expansion and the cost of upgrading later.
3. Storage Performance
Storage capacity and storage performance are different requirements.
Ten servers containing 1 TB of data each may appear to require approximately 10 TB of replacement storage. But capacity alone says nothing about how quickly those applications need to read and write data.
Virtualisation concentrates the I/O generated by multiple workloads onto the same storage infrastructure. Applications that previously used separate disks may suddenly compete for the same SSDs, RAID controller, storage array or network connection.
When planning consolidation, assess:
- required usable storage capacity;
- read and write IOPS;
- sequential and random workloads;
- latency sensitivity;
- database write activity;
- RAID requirements;
- SSD endurance;
- backup traffic; and
- future storage growth.
Modern enterprise NVMe SSDs can dramatically increase storage performance, but the storage architecture must still be designed around the combined workloads.
4. Network Throughput
Consolidating servers also consolidates network traffic.
Several physical servers may currently have separate 1GbE connections. Moving all of their applications onto one virtualisation host can create significantly higher traffic through fewer physical network interfaces.
Depending on workload requirements, consolidation may therefore justify 10GbE, 25GbE or faster networking, multiple network interfaces, VLAN separation and redundant switching.
Storage networks, virtual-machine migration traffic, backup traffic and application traffic should also be considered separately where appropriate.
5. Availability and Redundancy
This is the factor that prevents a simple calculation such as “ten old servers equal one new server.”
If ten independent servers are replaced by one extremely powerful host, that host becomes a much larger failure domain. A hardware failure could affect every virtual machine at the same time.
For business-critical infrastructure, the better design may therefore be two or three virtualisation hosts rather than one large system.
A clustered environment can distribute workloads between hosts and provide spare capacity if one server becomes unavailable. The required reserve depends on the organisation’s recovery objectives and the importance of the hosted applications.
A Simple Consolidation Example
Consider a company operating eight older physical servers:
- two file and print servers;
- two internal application servers;
- one accounting server;
- one database server;
- one web application server; and
- one utility and monitoring server.
Simply counting these machines would suggest an 8:1 consolidation opportunity. That conclusion would be premature.
After monitoring the workloads, the company might discover that most servers use very little CPU, but the database requires considerable memory and fast storage. The business may also decide that losing every application during one hardware failure is unacceptable.
Instead of replacing eight systems with one, the appropriate design could be two modern virtualisation hosts with shared or replicated storage. Normal workloads can be distributed across both machines while sufficient spare capacity is maintained for essential services during maintenance or a host failure.
The physical-server count has still been reduced from eight to two—a 4:1 consolidation ratio—but the resulting environment may be substantially more resilient than attempting to maximise the ratio.
Why the Highest Consolidation Ratio Is Not Always the Best
A high consolidation ratio looks attractive on a spreadsheet, but maximum density should not be the objective by itself.
Over-consolidation can create several problems:
- insufficient spare CPU or memory during workload spikes;
- storage bottlenecks;
- large numbers of applications affected by a single host failure;
- longer recovery times;
- complex maintenance windows;
- unexpected software licensing costs; and
- limited capacity for future growth.
The better target is the consolidation ratio that provides an appropriate balance of utilisation, performance, resilience and cost.
Do Not Forget Software Licensing
Hardware consolidation can reduce physical infrastructure while increasing the importance of software licensing decisions.
Some operating systems, databases, backup products and enterprise applications are licensed according to physical CPU sockets, processor cores, virtual CPUs or virtual machines. A high-core-count server can therefore have very different licensing economics depending on the software being deployed.
Licensing should be analysed before hardware is ordered—not after the migration has already been designed.
What Should You Measure Before Buying a Consolidation Server?
A useful capacity assessment should collect data from the existing environment over several weeks, including busy operational periods.
- CPU: average utilisation, peaks and sustained loads.
- Memory: active memory, maximum utilisation and growth.
- Storage: usable capacity, IOPS, throughput and latency.
- Network: average and peak traffic for each workload.
- Growth: expected workload requirements over the planned server lifecycle.
- Availability: applications that must continue operating during hardware maintenance or failure.
- Backup: backup windows, retention requirements and restore performance.
- Licensing: operating system, database, hypervisor and application licensing models.
VMware’s VMmark benchmark demonstrates why virtualisation performance should be evaluated as a complete environment rather than simply comparing CPU specifications: virtualised infrastructure combines compute, storage and infrastructure operations across multiple workloads.
When Should You Keep a Workload on Dedicated Hardware?
Virtualisation is suitable for a large range of business applications, but not every workload needs to be consolidated.
Dedicated physical infrastructure can remain appropriate for applications with specialised hardware requirements, extremely demanding or predictable performance needs, unusual licensing restrictions, very high storage I/O requirements or vendor support conditions that favour bare-metal deployment.
A consolidation project should therefore classify workloads rather than automatically virtualising everything.
Should You Consolidate onto One Server or Build a Small Cluster?
For non-critical systems, laboratories, development environments and smaller businesses, a single properly configured virtualisation server may be sufficient.
For production systems where downtime has a meaningful business impact, two or more hosts usually provide greater operational flexibility. Maintenance can be performed with less disruption, and workloads may be moved or restarted on another host depending on the virtualisation and storage architecture.
The decision should be based on business continuity requirements rather than simply the number of virtual machines.
Server Consolidation Can Also Be a Hardware Refresh Strategy
Consolidation does not require replacing every server with the newest and most expensive platform.
Depending on workload requirements and budget, organisations can evaluate current-generation systems, previous-generation enterprise servers or professionally sourced refurbished equipment. A previous-generation platform with sufficient CPU capacity, ECC memory, enterprise storage and redundant power may provide an attractive balance of performance and acquisition cost.
The key is to size the complete platform correctly rather than choosing hardware based only on processor specifications.
How ILA Express Can Help
ILA Express supplies enterprise servers, processors, ECC memory, SSDs, networking equipment and related infrastructure components for business and professional environments.
When planning a server consolidation or infrastructure refresh, start with the workload requirements: CPU, memory, storage, networking, redundancy, expansion and budget. These specifications can then be matched to an appropriate server platform rather than purchasing capacity that is either insufficient or unnecessarily expensive.
Explore our enterprise hardware products or learn more about enterprise hardware distribution. For projects involving dedicated infrastructure, cloud services or a combination of both, ILA Express can also help evaluate the most appropriate deployment approach.
Conclusion
One modern server may be capable of replacing several older physical systems, and current processor platforms demonstrate just how far compute density has advanced. But a useful consolidation ratio cannot be calculated from CPU core counts alone.
CPU utilisation, memory demand, storage I/O, network throughput, software licensing, growth and resilience all influence the final design.
The goal should not be to fit the greatest possible number of virtual machines onto the smallest possible number of servers. The goal is to reduce unnecessary infrastructure while maintaining enough performance, redundancy and expansion capacity for the workloads the business depends on.
Planning a server refresh or consolidation project? Contact ILA Express with your existing server specifications and workload requirements to discuss suitable enterprise hardware and infrastructure options.
