Cloud computing Architecture Technology and Practice Series (8) 2.2.3 application-independent Reliability Assurance Technology

copyright, without the written permission of Huawei, do not reprint or forward.

2.2.3 Application-independent Reliability Assurance Technology1.Reliability Assurance Technology in the data center

Reliability Assurance Technologies in the data center mainly include HA (high availability) cold backup,ft (Fault tolerance) hot spare, lightweight ft.

HA (High availability) Cold Backup : A cold migration based on shared storage in the data center that triggers the application to start on the standby server in case of a primary vm/pm failure due to software or hardware reasons (see figure 2-13 ) for non-business 0 outages ）。

Figure 2-13 Cold Backup Principle

FT (Fault tolerance) Hot Spares: instructions, memory, and all state data are synchronized. The advantage of this method is that the state is fully synchronized and fully guaranteed . Supports SMP. The disadvantage is the high performance overhead, which can bring about 40% performance degradation (see figure 2-14).

Figure 2-14 principle of hot backup

lightweight FT : FT Hot-standby mechanism based on I/O synchronization. Advantages: cpu/ Network performance loss within 10% , support for single-core and multi-core. The disadvantage is the scenario that is appropriate for the network I/O primary service (see figure 2-15).

Figure 2-15 Lightweight FT principle

1) Reliability Assurance technology across data centers

Reliability assurance technology across data centers is based on both synchronous and asynchronous disaster recovery of I/O replication of storage virtualization layer.

Synchronous disaster recovery based on storage virtualization Layer I/o replication, deployment with the production and disaster Recovery Center (<100km) , delay 5ms, ample bandwidth betweenDCs , and RPO ( Recovery point Objective) is higher, and general RPO is close to or equal to 0 seconds. Distributed block Storage provides more efficient I/O synchronous replication efficiency (see figure 2-16).

Figure 2-16 the principle of disaster-tolerant replication based on application layer

Asynchronous disaster recovery based on storage virtualization Layer I/O replication uses production and disaster Recovery Center offsite (>100km) deployments, bandwidth constrained, latency >5ms, and RPO Have certain tolerance, such as rpo>5 minutes. The performance impact of I/O replication and snapshots is approaching 0 (see figure 2-17).

Figure 2-17 the principle of disaster-tolerant replication based on storage layer

Cloud computing Architecture Technology and Practice Series (8) 2.2.3 application-independent Reliability Assurance Technology