In the world of VMware virtualization, it is important that you choose the right virtual disk type. Choosing the right virtual disk isn’t just a technical detail; instead, such details have far-reaching impacts on how your virtualized environment performs, consumes storage, and adapts to changes.
Be it raw, thick, and thick provisioning, each brings distinct strength and trade-offs. In this article, you will learn about the disk types that will allow you to make strategic choices that are aligned with your specific needs.
The Quick Guide: Choosing the Right VMware Disk Type
| Feature | Raw Device Mapping (RDM) | Thick Provision Lazy Zeroed | Thick Provision Eager Zeroed | Thin Provision |
| Definition | Direct access to physical storage | Pre-allocates storage on creation | Pre-allocates & zeroes space on creation | Allocates space ‘on-demand’ as data is written |
| Performance | Highest potential performance | Predictable, good performance | Predictable performance, initial creation slower | Can offer good performance, risk if storage overcommitted |
| Storage Efficiency | Least efficient | Medium efficiency | Least efficient | Most storage-efficient |
| Security | Depends on underlying storage | Data overwritten on deletion | Thorough data erasure on deletion | Data overwritten on deletion |
| Flexibility | Limited flexibility | Moderate flexibility | Limited flexibility | Highest flexibility |
| Management Overhead | Higher complexity | Low maintenance | Higher setup time, low maintenance | Requires monitoring to avoid space issues |
Demystifying Disk Types
1. Raw Device Mapping (RDM)
Raw Device Mapping provides a virtual machine (VM) with direct, almost bare-metal access to a physical storage unit (LUN) on your SAN. In essence, this will bypass the traditional VMware file system later.
a. When RDM Makes Sense
- Peak Performance Required: For workloads where even the slightest virtualization will impact the latency. So, the RDM will offer the closest experience to physical hardware.
b. Specialized Scenarios:
- Clustering Applications: Some clustering software will require direct control of shared disks. RDM can make this possible in virtual environments.
- Hardware Compatibility: Occasionally, legacy devices or specific applications mandate direct storage access that only RDM can provide.
c. The Trade-Offs of RDM
- Flexibility Takes a Hit: RDM-backed disks do not benefit from VMware features like snapshots, migrations, etc.
- Management Overhead: The Configuration and maintenance of RDMs is more complex than standard virtual disks.
- Potential Portability Issues: If you consider migrating VMs to different storage, the target will have to support the same LUNs that were used for the RDM.
All in all, Raw Device Mapping is a powerful but niche tool within VMware environments. Generally, it is reserved for exceptional use cases. For instance, in cases where you want the performance to outweigh the flexibility and management sacrifices. This is a key distinction when understanding the world of “VMware thick vs thin provisioning” – sometimes, raw performance is the deciding factor.
Thick Provisioning (Lazy & Eager Zeroed)
When it comes to Thick Provisioning, a virtual disk’s entire size will be allocated and reserved on the datastore from the start. This is what differentiates Thick Provisioning from Thin Provisioning’s — that has the dynamic growth model.
a. Lazy vs. Eager Zeroed – The Performance/Security Tradeoff
Thick Provision Lazy Zeroed: Here, the space is reserved upfront. However, the data blocks are zero-written only as the VM uses them. This offers:
- Performance Middle-Ground: Initial writes might be faster than Eager Zeroed (no upfront zeroing). Later performance will depend on usage patterns.
- Storage Efficiency Compromise: Better than Eager, but this will still consume the full VM’s defined size.
Thick Provision Eager Zeroed: The entire disk is pre-allocated AND zeroed immediately upon creation. This will provide:
- Security Benefit: Deleted data is overwritten in Thick Provisioning and makes the recovery difficult. This is important for highly sensitive workloads.
- Performance Cost: Initial creation is slowest. Some performance impacts always persist. This is especially true for random write workloads.
- Least Storage-Efficient: There is no space savings whatsoever. The full VM disk size in Thick Provisioning is used, even if empty.
When to Choose Thick Provisioning
- Performance-Sensitive VMs: If consistent, predictable I/O is essential (databases, critical apps), Thick Provisioning reduces the risk of latency spikes compared to Thin.
- Simplified Management: No need to monitor datastore capacity like you do with Thin Provisioning.
- Regulatory/Security Needs: When compliance mandates thorough data erasure for sensitive VMs, Eager Zeroed is often the path taken.
a. Situations Where You’d Use Code/Scripts:
- Automated VM Deployment: Scripts (PowerCLI, etc.) can define disk types during VM provisioning, ensuring consistency.
- Batch Conversion: If changing many existing VMs to Thick, scripts are more efficient than manual changes.
- Policy Enforcement: Scripts can scan for Thin Provisioned VMs in certain environments, converting them to Thick if they violate security guidelines.
Thin Provisioning
Thin Provisioning offers dynamic allocation — a virtual disk starts small and expands on-demand as the VM writes data. This will maximize the storage utilization at the cost of potential performance variability.
Advantages
- Storage Efficiency: This will have huge benefits in environments with limited space. Additionally, you can avoid paying for unused capacity upfront.
- Deployment Speed: Thin disks can be created extremely quickly, given there is minimal initial allocation.
- Flexibility: Well-suited for test/dev VMs or workloads will have unpredictable growth patterns.
Drawbacks
- Performance Risk: If the underlying storage is overcommitted, i.e., more promised than physically available, the Virtual Machines using thin disks can suffer latency spikes whenever the capacity will run low.
- Monitoring is Key: You must proactively monitor datastore free space to prevent thin disks from “outgrowing” from what is available. This can lead to outages.
The Decision Framework: When to Choose What
Prioritizing Performance
- Workloads that need guaranteed low latency (databases, critical applications) are likely to favor Thick Provisioning.
- Additionally, the performance-sensitive VMs generally avoid Thin Provisioning due to potential fluctuations in the system.
Optimizing Storage Efficiency
- Limited storage environments will benefit greatly from Thin Provisioning’s space-saving approach.
- Generous storage scenarios might trade extra space, given the Thick Provisioning’s performance advantages available and accessible.
Security Considerations
- Highly sensitive data (PII, healthcare info) might mandate Thick Eager Zeroed for its thorough erasure.
- Regulatory compliance can influence disk type choices to ensure data security.
Conclusion
The world of VMware virtual disks isn’t about a one-size-fits-all answer. The success lies in understanding the nuances between the Raw, Thick (Lazy/Eager Zeroed), and thin provisioning methods.
All that you have to do is carefully consider factors like performance demands, storage constraints, and security requirements and then you can empower your virtualized environment to run at its best.
It is also important to know that the disk that you consider right might evolve tomorrow, so it is important to value the ongoing assessment and make informed decisions.