Infinispan is a distributed in-memory data grid platform designed to offer fast access to large volumes of data across a cluster of computers. In enterprise computing, managing data effectively and efficiently is critical. Infinispan provides several operational modes to fit various use cases, ranging from simple local caching to complex, large-scale data grids.

Local Mode

In local mode, Infinispan functions as a standalone cache not sharing data with other nodes in a cluster. This mode is suitable for single-node applications where the primary goal is to speed up local data access without the complexity of clustering.

Example Usage:

Cache<String, String> cache = new DefaultCacheManager().getCache();
cache.put("key", "value");
System.out.println("Value for 'key': " + cache.get("key"));

Replicated Mode

In replicated mode, each node in the cluster contains a copy of all entries. This mode ensures data availability and fault tolerance. Changes to the cache are propagated synchronously or asynchronously to all other nodes, ensuring data consistency across the cluster.

Example Configuration:

<replicated-cache name="myCache">
  <locking isolation="REPEATABLE_READ"/>
</replicated-cache>

Distributed Mode

Distributed mode is where Infinispan shines in scalability. Unlike replicated mode, data is not stored on all nodes. Instead, each entry is stored on a configurable number of nodes (numOwners). This approach reduces memory usage while maintaining high availability and fault tolerance.

Example Configuration:

<distributed-cache name="distCache" mode="SYNC">
  <hash num-owners="2"/>
</distributed-cache>

Invalidation Mode

Invalidation mode doesn't actually share data across nodes. Instead, when a data change occurs, other nodes are notified to invalidate their local copies of that data. This mode is beneficial when it's more important to remove stale data than to distribute new data.

Example Usage:

<invalidation-cache name="invalidCache" mode="ASYNC">
  <transaction mode="NONE"/>
</invalidation-cache>

Caching Modes Comparison

Each mode serves different needs. Local is the simplest and fastest but offers no data resilience. Replicated provides complete data redundancy at the cost of higher memory usage and write latency. Distributed mode provides a balance with configurable options like numOwners to balance memory usage and data availability, suitable for large-scale deployments. Finally, invalidation mode is used primarily to maintain cache consistency rather than data sharing.

Advanced Features and Considerations

• Transaction Support: Infinispan supports transactional caches in which operations are done within the context of a transaction, which is crucial for maintaining data integrity.
• Querying: Infinispan offers querying capabilities so you can search through cache data using SQL-like queries, including full-text search.
• Persistence: Infinispan can be configured to persist data to external storage, ensuring that data is not lost even if all nodes in the cluster go down.
• Security: Infinispan includes features for securing data, such as authentication and authorization mechanisms and encryption capabilities both for data at rest and in transit.
• Cross-Site Replication: For disaster recovery, Infinispan supports cross-site replication allowing clusters to back each other up in different geolocations.

In conclusion, choosing the right operational mode in Infinispan depends significantly on specific application requirements concerning scaling, data consistency, and availability needs. With its versatile caching and data-grid capabilities, Infinispan caters to a broad spectrum of enterprise data management challenges.