Facts About Latency
NAS Performance Pitfalls: Inside the Box
NAS performance problems frustrate and plague storage administrators. It’s an endless time-consuming struggle to isolate and fix performance bottlenecks. Latency or delay is at the heart of the matter. Latency is measured as round-trip-time for a data packet to transit between two specific points but is experienced by users as slow application response time and manifests as muttering and explicatives.
NAS performance problems have 3 primary root causes:
1. Conventional hard disk drives (HDDs) high latency, low IOPS,
and poor throughput.
- HDDs are a bottleneck, delivering prohibitive read and write latencies.
- Large numbers of HDDs are required to help fill IOPS requirements.
2. Excessive NAS controller processor utilization.
- NAS cycles used for snapshots, deduplication, replication, storage tiering, thin provisioning, file system, virtualization integration, backup software integration, layer 7 NFS processing, networking, reads, writes, and NFS metadata processing can add up and lead to processor contention and increased latency.
- NFS and CIFS are chatty protocols that consume NAS controller cycles and system bandwidth.
3. Meager network throughput.
- NAS systems can’t process data IO fast enough.
- NFS Metadata is very chatty and consumes NAS controller cycles before Reads and Writes are executed.
- NFS Metadata can be as much as 90% of the total NFS operations.
- NAS controller cycles are rapidly consumed by processing data IO and data copies that amplify the latency chokepoint.
- TCP/IP inherent latency reduces IOPS and effective data throughput across the network.
These 3 root causes of NAS lackluster performance are enormously problematic to overcome inside the NAS system or “box”.
