Cache Clients are lightweight modules through which cache applications can connect to hosted cache servers.
TCP Based communication: Cache client connects to the cache server using TCP sockets. Cache client opens a persistent connection with the cache server so when it is disconnected due to network failure, it always reconnects with the cache server.
No Change in API: For both local and remote connections, API for cache operations does not change.
Data Serialization: Data to be added needs to be serializable. Serialization is performed in cache client.
Data Compression: Compression is configurable. If enabled, serialized data is compressed in cache client.
Data Encryption: Encryption is also configurable. If enabled, data is encrypted after compression in cache client. NCache supports AES and 3DES encryption.
Request Timeout: Timeout for synchronous cache operations like Add/Insert can be configured, if no response is received within configured time, exception will be thrown. Default operation timeout is 60 seconds. This request timeout should always be higher than cluster operation timeout.
Cache Server Information: Cache clients require server information in order
to establish connection. This cache server information can be provided in two
ways, either through
CacheInitParams or through client configurations
Add/Remove Node: If any node is added or removed in cache cluster, cache client will also add/ remove server information to its server list. In case of fail over, client will automatically connect to the newly joined node.
Fail Over: In case a server goes down, cache client seamlessly establishes connection with the next available server. if none of the server is available, e.g., due to network failure, cache client keeps on trying to establish connection with the server in background. When a server again becomes available, it immediately connects to it and application can resume its activity with the cache. In partition and partitioned-replica cache topology, cache client maintains connection with every cache server. In case cache client is unable to connect with some of the server(s), it executes cache requests through other available cache servers. Meanwhile, it keeps trying to establish connection with the server in background.
Connection Retry Interval and Retry Count: When an application executes any cache operation through cache client and no connection is available, then cache client tries to establish the connection. The connection establishment process tries to establish connection for a configurable number of retries. The interval between two consecutive retries can be configured. Meanwhile, application thread waits for connection establishment. When retries are over and cache client is unable to connect to any of the cache servers, then application request is considered as failed and an exception is thrown to the client application.
Load Balancing: Cache client is topology aware and it behaves according to topology. In replicated cache cluster, clients are load balanced among cache servers, if any node joins/leaves the existing replicated cluster, load is re-shuffled.
Awareness of Data Distribution: In partition and partitioned-replica cache, client is aware of data distribution and establishes connection with all cache servers. Data distribution approach has the following advantages:
Operation is one-hop, which results in a higher throughput.
Cache requests are truly balanced among cache servers and no single server is overwhelmed.