What system requirements are there for installing OceanBase Database?
The following table describes the minimum configuration requirements for the servers.
| Server type | Number | Minimum functional configuration | Minimum performance configuration |
|---|---|---|---|
| OCP server | 1 | 16 CPU cores, 32 GB of memory, 1.5 TB of storage | 32 CPU cores, 128 GB of memory, 1.5 TB of SSD storage, and a 10-GbE network card |
| OceanBase Database server | 3 | 4 CPU cores, 16 GB of memory,
NoteThe log disk requires more than three times the memory space, and the data disk requires sufficient space to store the target amount of data. |
32 CPU cores, 256 GB of memory,
NoteThe log disk requires more than three times the memory space, and the data disk requires sufficient space to store the target amount of data. |
If you want the OCP server to provide high availability, you need three OCP servers and a load balancing software or hardware solution, such as F5, Alibaba Cloud SLB, or OceanBase Database's ob_dns. In this case, deploy the OCP server in a three-node configuration.
You can install OceanBase Database on the Linux operating systems listed in the following table.
| Linux operating system | Version | Server architecture |
|---|---|---|
| Alibaba Cloud Linux | 2 | x86_64, ARM_64 |
| Anolis OS | 8.6 and later | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| KylinOS | V10 | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| Unity Operating System (UOS) | V20 | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| NFSChina | 4.0 and later | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| Inspur KOS | 5.8 | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| CentOS / Red Hat Enterprise Linux | 7.x, 8.x | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| SUSE Enterprise Linux | 12SP3 and later | x86_64 (including HaiGao) |
| Debian | 8.3 and later | x86_64 (including HaiGao) |
| openEuler | 20.03 LTS SP1/SP2 | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
| Nengsi LinxOS | V6.0.99, V6.0.100 | x86_64 (including HaiGao), ARM_64 (including KungPeng and Phytium) |
Note
The operating system must be configured with a network connection and a software management tool (yum or zypper).
How will OceanBase Database be deployed in a production environment?
The following deployment solutions are available:
| Solution | Characteristics | Infrastructure requirements | Scenarios |
|---|---|---|---|
| Single IDC with three replicas | RPO=0, RTO low, automatic switchover in case of failure. It protects against individual hardware failures but not IDC-level or city-level disasters. | Single IDC. | Scenarios where IDC-level and city-level disaster recovery are not required. |
| Three IDCs in the same region with three replicas | RPO=0, RTO low, automatic switchover in case of failure. It protects against individual hardware failures and IDC-level disasters but not city-level disasters. | Three IDCs in the same region. The network latency between IDCs in the same region is low. | Scenarios where IDC-level disaster recovery is required but city-level disaster recovery is not. |
| Five IDCs across three regions with five replicas | RPO=0, RTO low, automatic switchover in case of failure. It protects against individual hardware failures, IDC-level disasters, and city-level disasters. | Five IDCs across three regions. Two of the regions are nearby with low network latency. | Scenarios where IDC-level and city-level disaster recovery are both required. |
| Two IDCs in the same region with data replication between clusters | RPO>0, RTO high, manual switchover. It protects against individual hardware failures and IDC-level disasters but not city-level disasters. | Two IDCs in the same region. | There are two IDCs in the same region, and IDC-level disaster recovery is required. |
| Three IDCs across two regions with five replicas and data replication between clusters | IDC-level failures: RPO=0, RTO low, automatic switchover. City-level failures: RPO>0, RTO high, manual switchover. It protects against individual hardware failures, IDC-level disasters, and city-level disasters. | Three IDCs across two regions. | There are two regions with three IDCs, and both IDC-level and city-level disaster recovery are required. |
What is LSE?
LSE is a feature introduced in ARMv8.1, namely, "Large System Extensions" (large system extensions). LSE provides a set of atomic operations for supporting synchronization and mutually exclusive access in a multiprocessor environment. These atomic operations include Load-Exclusive (LDE), Store-Exclusive (STE), and Conditional Compare Exchange (CCXE) instructions. LSE also provides new instructions and memory barriers to maintain data consistency and sequentiality.
LSE instructions allow for efficient access to shared memory, enabling finer-grained locking mechanisms and reducing the overheads of locking. Compared with traditional synchronization instructions, LSE instructions can reduce lock competition, improve concurrent performance, and provide better scalability.
From which OCP version is the nolse package supported?
OCP supports the OceanBase RPM software package marked with nolse, starting from V4.3.0. This addresses the issue of identifying Large System Extensions (LSE) instructions in the ARM architecture.
Considerations for OBD deployment
You can upload both the nolse and the no-nolse OceanBase Database RPM packages. OBD will adaptively deploy the packages based on the system support. Alternatively, you can upload only the supported packages for installation.
Considerations for OCP deployment
Although you can upload both the nolse and the no-nolse packages to OCP, the packages cannot be installed adaptively. Therefore, when you deploy a cluster, you need to select whether to install the packages supporting LSE.