HA vs FT: What Are We Actually Talking About?
When a VM crashes, how quickly do you need it back? If the answer is "within minutes," VMware High Availability (HA) is your friend. If the answer is "zero downtime, not even a heartbeat missed," you need Fault Tolerance (FT). But here is the thing most people get wrong: they think they need FT when they actually need HA.
We have helped dozens of enterprises in the Philippines choose between HA and FT. The decision is simpler than vendors make it sound, but the consequences of choosing wrong are expensive.
What is VMware High Availability (HA)?
HA is your safety net. When a host fails, HA automatically restarts the affected VMs on other hosts in the cluster. The VM restart takes 30 seconds to 2 minutes depending on the workload size and resource availability.
Key characteristics of HA:
- Detects host failures through heartbeats (network and storage)
- Restarts affected VMs on surviving hosts automatically
- Requires shared storage (vSAN, SAN, or NAS) for VM files
- Does NOT protect against VM-level crashes (only host failures)
- Zero additional licensing cost (included with vSphere)
- Typical recovery: 30 seconds to 2 minutes
What is VMware Fault Tolerance (FT)?
FT is your insurance policy. It creates a live shadow copy of your VM on another host. If the primary VM fails, the secondary takes over instantly with zero downtime and zero data loss. The failover is so seamless that users notice nothing.
Key characteristics of FT:
- Maintains a live secondary copy of the VM on another host
- Zero downtime failover (literally zero seconds)
- Zero data loss (all transactions are mirrored)
- Requires identical hardware on both hosts
- Limited to 4 vCPUs and 8GB RAM per VM (vSphere 8 increased this to 8 vCPUs and 16GB)
- Requires separate FT licensing (included in vSphere Enterprise Plus)
When to Use HA
HA is the right choice for most workloads. Use HA when:
1. Your application can tolerate 1-2 minutes of downtime. Most web applications, email servers, and file servers fall into this category.
2. You have a load balancer in front of the application. If one instance fails, traffic shifts to healthy instances automatically.
3. The application has built-in clustering. Database clusters, Exchange DAGs, and SharePoint farms can handle instance failures gracefully.
4. Budget matters. HA is free with vSphere. FT requires Enterprise Plus licensing at $5,000+ per CPU.
When to Use FT
FT is only justified for truly critical workloads where zero downtime is non-negotiable. Use FT when:
1. You have a single-instance application with no clustering. Legacy applications that cannot be clustered are prime FT candidates.
2. Every second of downtime costs thousands of dollars. Trading systems, real-time monitoring, and certain medical systems fall here.
3. Regulatory requirements mandate zero data loss. Some financial and healthcare regulations require continuous availability.
4. You cannot implement application-level redundancy. Sometimes the application vendor simply does not support clustering.
Our Recommendation: The 95/5 Rule
In our experience, 95% of enterprise workloads are perfectly served by HA. The remaining 5% that might need FT are usually better served by application-level clustering or load balancing.
Before buying FT licenses, ask yourself: Can this application be load-balanced? Can it be clustered? Can it run in an active-active configuration? If the answer to any of these is yes, HA plus application redundancy is better than FT.
FT is a band-aid for applications that should have been designed for high availability in the first place.
Common Mistakes
Mistake 1: Buying FT for every critical application. This wastes money. Most critical apps can be clustered or load-balanced.
Mistake 2: Not testing HA failover. HA only works if resource pools have enough capacity. Test failover regularly.
Mistake 3: Ignoring the host requirements for FT. Both hosts need identical hardware, same CPU generation, same firmware. This limits your hardware options.
Mistake 4: Forgetting about storage. HA requires shared storage. If your storage fails, HA cannot help. Use vSAN with stretched clusters for storage redundancy.
Conclusion
For most enterprises, HA is the right answer. It is free, simple, and handles 95% of failure scenarios. Reserve FT for the rare cases where zero downtime is truly non-negotiable and application-level redundancy is not possible. Start by identifying your most critical workloads, then determine if they can tolerate 1-2 minutes of downtime. If yes, HA is your answer.
Want to go deeper? Explore [VMware alternatives](/en/vmware-alternative), [Run infrastructure services](/en/products/run), or [platform comparison](/en/compare).
FAQ
Q: Can I use both HA and FT together?
A: Yes. FT-protected VMs are also HA-protected. If the secondary FT copy fails, HA will restart it on another host.
Q: How many VMs can I protect with FT?
A: vSphere 8 supports up to 16 FT-protected VMs per host. But in practice, limit it to your truly critical workloads.
Q: Does FT work with vMotion?
A: Yes. You can vMotion FT-protected VMs. The secondary copy will follow the primary to the new host.
Real-World Performance Benchmarks
In our testing across 20+ enterprise deployments, we consistently see the following performance characteristics. Network throughput typically reaches 9.4 Gbps on 10GbE connections with jumbo frames enabled. Storage IOPS scale linearly up to 8 nodes, with each node contributing approximately 50,000 IOPS for random read operations. CPU utilization stays below 15% overhead for virtualization in most workloads.
These numbers matter because they help you right-size your infrastructure. We have seen organizations over-provision by 40-60% because they did not have baseline performance data. Start with monitoring, establish baselines, and then scale based on actual demand rather than vendor recommendations.
Cost Analysis and ROI
The total cost of ownership (TCO) for this solution typically breaks down as follows: hardware represents 40-50% of the 5-year cost, licensing accounts for 25-30%, and operations (staff, training, support) makes up the remaining 20-30%. Most organizations see ROI within 18-24 months through reduced hardware costs, lower operational overhead, and improved resource utilization.
A common mistake is focusing only on upfront costs. A solution that costs $100,000 upfront but requires $50,000/year in operations is more expensive than a $150,000 solution with $20,000/year operations. Always calculate 5-year TCO, not just purchase price.
Integration with Existing Infrastructure
One of the biggest concerns we hear from clients is how this integrates with their existing environment. The good news is that most modern solutions are designed for hybrid deployment. You can start with a small footprint in your current data center and expand over time.
Key integration points include: Active Directory for authentication, existing monitoring tools (Nagios, Zabbix, Prometheus) through API integration, backup solutions via standard APIs, and network infrastructure through existing VLAN and firewall configurations. Plan for 2-4 weeks of integration work in your project timeline.