About data store performance

Determines how data is replaced in the data store. The default policy works well for most appliances, but in some cases, changing the policy can improve performance. We recommend that the segment replacement policy is the same on both the client-side and server-side SteelHeads.

Replaces the least recently used data in the data store, which improves hit rates when the data aren’t equally used. This is the default value.

Replaces data in the order received (first in, first out). This can be useful for optimizing data stores for high-throughput workloads like data replication (DR) or data center tasks over high-bandwidth WAN connections. For DR workloads, we recommend using separate appliances from those used for other application traffic to maintain consistent performance. This approach ensures data reduction benefits while improving the overall data store performance.

These modes help the appliance manage and adjust resource use to boost LAN throughput. The default setting typically gives the best data reduction, and changes should only be made with guidance from Riverbed Support.

Use caution with the SDR-Adaptive Legacy setting, especially if you’re optimizing NFS with prepopulation.

SDR-M mode is not compatible with data store synchronization.

If you change the streamlining mode, restart the service on both the client-side and server-side appliances, and check the Optimized Throughput report to see if the change improved performance.

Is enabled by default and works for most implementations.

Includes the default features and adds balancing between read and write operations. It also monitors both disk I/O response times and CPU load. By analyzing performance trends, it intelligently combines disk-based and non-disk-based data reduction methods to maintain steady throughput, even during heavy disk or CPU usage.

Automatically adjusts to different data workloads to maximize LAN-side throughput. It uses a smart switching system that aims to reduce bandwidth usage while maintaining high performance, based on the available WAN bandwidth.

Performs data reduction entirely in memory, avoiding disk read/write operations. This can lead to higher LAN-side throughput by eliminating disk latency. It’s ideal for SAN replication environments and works best between two identical high-end SteelHead models. If used with different models, the smaller one may limit performance. After enabling SDR-M on both the client and server appliances, restart both to prevent performance issues. If SDR-M is selected as the adaptive data streamlining mode, the Clear the Data Store option will not be available when restarting the optimization service, as SDR-M doesn’t affect the data store disk.

Allows you to adjust the compression level, enable adaptive compression, and balance connection loads across multiple cores. These features help handle high-traffic loads by reducing compression, boosting throughput, and making better use of Long Fat Network (LFN) bandwidth.

Controls the balance between data compression and LAN throughput speed. Lower levels offer faster throughput with slightly less compression. Level 1 uses the least compression and the least CPU, while Level 9 offers maximum compression but uses more CPU. The default is Level 6. For high-throughput environments, like data center-to-data center replication, we recommend setting it to Level 1.

Dynamically monitors LZ data compression performance for each connection. If it’s not performing well, it temporarily disables compression (sets it to level 0). It helps improve overall LAN throughput by optimizing WAN throughput. By default, this feature is disabled.

Ensures even distribution of workload across all CPUs, thereby maximizing throughput by keeping all CPUs busy. Core balancing is useful when handling a small number of high-throughput connections (approximately 25 or fewer). By default, this setting is disabled and should be enabled only after careful consideration.