A virtual cluster is a group of virtual machines (VMs) distributed across multiple physical machines (nodes), interconnected via a virtual network. These clusters simulate the behavior of physical clusters but with greater flexibility, scalability, and manageability.
Virtual clusters are essential in cloud computing platforms like Amazon EC2, which dynamically provision VMs and scale applications based on demand.
Physical vs. Virtual Clusters
A physical cluster is a set of interconnected physical servers. In contrast, a virtual cluster is formed by VMs that can span across multiple physical servers or even across physical clusters.
Characteristics of Virtual Clusters:
- VMs can run different OSes on the same host.
- A VM runs on a guest OS, which is often different from the host OS.
- They consolidate multiple applications onto the same physical server.
- VMs can be replicated for parallelism, fault tolerance, and disaster recovery.
- VM count and placement can change dynamically, making the system scalable.
- VM failures do not bring down the physical host system.
Application Partitioning & Customization
Large numbers of VM images must be stored and managed efficiently.
- Common software (like OS and libraries) can be preinstalled in template VMs.
- Users can duplicate a template and install custom libraries/apps to create their VMs.
- VM clusters can be partitioned by application domain.
Fast Deployment and Effective Scheduling
Deployment involves:
- Constructing and distributing VM software stacks
- Switching environments between users’ virtual clusters
Key Points:
- Live migration of VMs enables moving workloads between hosts.
- Helps in green computing by reducing idle resources.
- Enables load balancing through migration based on usage.
- Cluster size can scale up/down automatically using user activity and load index.
Challenges:
- Live migration overhead may affect cluster performance, QoS, and throughput.
- Migration strategy must balance performance and power efficiency.
High-Performance Virtual Storage
Efficient storage of large numbers of VM images is crucial.
Techniques:
- Use template VMs with Copy-On-Write (COW) format for disk image duplication.
- Saves storage and deployment time.
- Hashing is used to detect duplicate data blocks.
- User profiles maintain the list of unique blocks.
VM Deployment Steps:
- Prepare disk image using a template.
- Configure VM (name, memory, CPU, network).
- Select destination node.
- Deploy VM on the selected host.
Optimization:
- Use pre-edited profiles for large-scale deployment.
- Assign UUID, name, IP address automatically.
- Strategy selects best host for VM to ensure load balancing and resource utilization.