Oracle TimesTen Scaleout is a distributed database that has its own terminology. This blog discusses those terms and common variations.
A TimesTen Scaleout host is a physical machine, a virtual machine or a container. As TimesTen Scaleout has a shared nothing architecture, each host logically has its own CPU, memory, persistent storage and networking. For optimal performance and availability, physical [bare metal] machines should be used as hosts. Virtual machines can have significant network and disk IO overhead compared to bare metal machines. Also virtual machines and containers need to share their resources (CPU, memory, persistent storage and networking).
A database is a set of users, tables and other schema objects that applications can connect to. Every database is distributed across all the data instances.
A TimesTen Scaleout instance is a running copy of the TimesTen software. The instance includes the processes and files that exist on a host. A host may contain one or more instances. Each instance contains a single element of the database.
For optimal high availability, you should configure one instance per host.
For development and test environments, it can be convenient and cheaper to configure multiple instances per host. If you configure multiple instances per host, you need to make sure that there is sufficient memory and disk space for each instance. Having multiple instances per hosts means that each instance needs to share the hosts CPU, network, disks and memory.
Each instance contains a single element. Elements correspond to the traditional TimesTen In-Memory Database components (shared memory, daemon and sub-daemon processes, checkpoint files and transaction log files). Each element can persistently store its own data and be recovered independently.
A TimesTen Scaleout database will usually consist of many hosts.To enable high availability and avoid single points of failure, there can be multiple copies of an element. The copies of an element are called relica sets.
TimesTen Scaleout uses the two phase commit protocol which means that the copies of an element are relicas of each other. The number of copies of an instance corresponds to the number of data space groups. A TimesTen Scaleout database with a single data space means that there is no data redundancy and that each instance is a single point of failure. A TimesTen Scaleout database with two data spaces means that there is data redundancy and that each element is not single point of failure.
The picture on the left shows that the intersection of a replica set and the data space group is an instance. In the picture on the right, if a host contains a single instance then the intersection of replica set and a data space group is a host. Having a single instance per host provides the best availability.
The above picture shows how SQL Developer displays data space groups and replica sets.
TimesTen Scaleout uses the term K-Safety to define the number of copies of an element. K copies of an element corresponds to the number of data space groups. A database with K=2 will have two data spaces. A database with K=1 will have one data space.
To simply describe a configuration, the NxK term to describes the number of replica sets (N) and the K Factor (K) for a database.
The above is an 8x1 grid as there are 8 replica sets and only a single data space (K=1).
The above is an 8x2 grid as there are 8 replica sets and two data spaces (K=2).
The above shows two different representations of a 3x2 grid as there are 3 replica sets and two data spaces (K=2).
TimesTen Scaleout defines an installation as the read only copy of the files in the TimesTen software distribution.
Disclaimer: These are my personal thoughts and do not represent Oracle's official viewpoint in any way, shape, or form.