Home > Storage > ObjectScale and ECS > Product Documentation > ECS: Overview and Architecture > Failure tolerance
ECS is designed to tolerate a range of equipment failure situations using a number of fault domains. The range of failure conditions spans a varying scope including:
In either a single site, dual-site, or geo-replicated configuration, the impact of the failure depends on the quantity and type of components affected. However, at each level, ECS provides mechanisms to defend against the impact of component failures. Many of these mechanisms have already been discussed in this paper but are reviewed here and in the following figure to show how they are applied to the solution. These include:
Note: For the rack aware, when adding new rack to the exist cluster, some of the data will be moved to the new rack to balance the data across all the racks equally. However, the process could take a long period of time to avoid having a performance impact on the system. If the customer keeps writing aggressively and fills the first rack, then all the new writes will happen only on the new rack.
The following table defines the type and number of component failures that each EC scheme protects against per basic rack configuration. The table highlights the importance of considering the impact of protective failure domains on overall data and service availability in terms of number of nodes required at each EC scheme.
EC scheme | # nodes in VDC | # chunk fragments per node | EC data protected against… |
12+4 Default | 5 or less | 4 |
|
6 or 7 | 3 |
| |
8 or more | 2 |
| |
15 | one node with 2 fragments and other nodes with 1 fragment. |
| |
16 or more | 1 |
| |
10+2 Cold Storage | 11 or less | 2 |
|
12 or more | 1 |
|