5
S
YSTEM DATABASE DEPLOYMENT EXAMPLES IN MX-ONE
C
HAPTER 2 OVERVIEW
sharing. For example, for a Data Center with four system database Nodes and a Replication Factor two,
each node will store half of the data. Each data is stored replicated to two nodes.
The total number of data copies that are replicated is referred to as the replication factor. A replication
factor of 1 means that there is only one copy of each row in a Data Center, whereas a replication factor
of 3 means three copies of the data are stored across the Data Center.
For ASP113, the data is always replicated to all database nodes in a Data Center.
Rack
A rack is the physical location of a server (or group of servers) within a Data Center. Servers are spread
between racks to get redundancy if rack specific hardware fails. Examples of such hardware are power
supplies, access switches and IP networks. system database uses racks to spread replicas to different
racks if they are available.
How Data is Written
When a write request is sent to one system database node, the write request is forwarded to all other
system database nodes that hold a copy of that data table row. The system database only waits for a
defined number of nodes to be ready with the write, before considering the write successful.
If some nodes are not reachable, the system database has methods to make data consistent when the
failed node has recovered. The system database stores a time-stamp for each write to know what the
latest write is.
When a write is forwarded to another Data Center, each Data Center distributes the write requests to
other nodes in the Data Center holding a copy of the data. This reduces the inter Data Center network
traffic.
The system database tries to write to the node that has the shortest response time.
How Data is Read
When a read request is sent to one system database node, the request is forwarded to one node that
holds a copy of the data. The system database also forwards digest requests to other Servers to verify
that the data read is consistent with other replicas in the database. If the digest is consistent with the data
read it is considered OK.
If the read data is not consistent with the digest, a repair action is started that will make this data table row
consistent.
The system database tries to read from the node that has the shortest response time.
System Database Deployment Examples in MX-ONE
The system database could when used in the ASP113 system be configured and deployed for example
according to the following figures, either as a single Data Center, or as multiple Data Centers, and either
co-located with the Service Node(s), or on separate server(s), standalone, and with one or multiple racks
within a Data Center:
78/1551-ANF 901 14 Uen E 2019-10-18