Modifying a QoS Replication Policy

Set the bandwidth upper limit for background asynchronous replication (including asynchronous replication upon a synchronization failure) of a cluster. The front-end synchronous replication traffic can be adjusted by controlling the front-end service traffic. For details, see Configuring and Managing SmartQoS.

Procedure

  1. Choose Data Protection > Configuration > Cross-Site DR > Topology.
  2. In the local cluster area on the right of the function pane, choose Operation > QoS Replication Policy.

    The QoS Replication Policy page is displayed on the right.

  3. Set the replication bandwidth, that is, the maximum value of the bandwidth upper limit for data replication. The value ranges from 1 MB/s to 655360 MB/s.

    • The bandwidth value is the maximum value of the bandwidth upper limit for data replication. In addition to data synchronization, metadata, heartbeat data, and system configuration data are also synchronized on replication links. Only that the latter is not restricted by the bandwidth value. Therefore, on the performance monitoring page, the link replication bandwidth displayed may be greater than the bandwidth value.
    • To accelerate replication, when the data volume is smaller than a certain threshold for the maximum data volume that can be carried in a replication metadata I/O interaction, metadata and data are replicated to the remote end in one I/O interaction. In this case, the replication bandwidth upper limit differs greatly from the actual link replication bandwidth. The threshold has three options: 8 KB, 64 KB, and 128 KB. You can run the fas set copy_with_meta_data_threshold command on the CLI as user admin to specify a threshold. For details, see the fas set copy_with_meta_data_threshold command in the command reference of the corresponding version.
    • The bandwidth value takes effect for the entire cluster instead of a single node. Therefore, you may need to adjust the bandwidth value when the number of nodes in the cluster changes. When sufficient physical bandwidth is available, you can adjust the bandwidth value in direct proportion to the number of nodes. For a node, the maximum value of the bandwidth upper limit is the bandwidth value divided by the number of nodes. By default, the system sets the minimum value of the bandwidth upper limit for the node to the smaller value between 100 MB and 1/10 x Maximum value of the bandwidth upper limit. When foreground services are processed, the system dynamically controls the replication bandwidth upper limit between the minimum and the maximum values for traffic control. (The higher the foreground service loads, the lower the replication bandwidth. Otherwise, the higher the replication bandwidth.) Setting the bandwidth value must take into account the customer's service models, networks, and hardware configurations. You are advised to contact R&D engineers for evaluation. The following describes some examples of typical configuration scenarios.
      • In the service scenario where objects are mostly small, you do not need to change the bandwidth value. The default value meets the bandwidth requirements.
      • When occupancy of the background physical link bandwidth is not considered, if the foreground service loads are low or there are no high requirements for foreground service performance, you can set the bandwidth value (that is, the bandwidth upper limit for data replication) to 10 x N x W. In this way, even if the foreground service loads are high or there are high performance requirements, the bandwidth upper limit for the background replication is not lower than the foreground service bandwidth.
      • If the foreground service loads are high or there are high requirements for foreground service performance, you can set the bandwidth value to N x M x W.
      • N: site amplification coefficient. For a replica replication group, N = n – 1. For an EC replication group, N = (n + m – 1)/n. In the preceding equations, "n" indicates the number of sites, and "m" indicates the number of parity fragments.
      • M: redundancy coefficient. Due to reasons such as retransmission upon failure and retransmission due to traffic control, the maximum value of the replication bandwidth upper limit must be multiplied by a proper redundancy coefficient. The empirical value of this coefficient ranges from 1.2 to 1.8.
      • W: average daily bandwidth of foreground services. That is, average bandwidth of foreground services calculated every calendar day.

  4. Click OK.