Home > Storage > ObjectScale and ECS > Product Documentation > Dell ECS: Networking Best Practices > ECS network performance
Network performance is a major factor that can affect the ability of any cloud storage platform to serve data. When architecting or designing the customer network to connect with ECS data switches, there are some considerations to maintain optimal performance. Data, replication, internode traffic, and management traffic (ECS Portal, Rest APIs, and traffic to network services such as DNS, and AD), flows through the data switches. A reliable and highly available network is also important.
For production network, one uplink per switch to customer switch is required at the minimum. However, one per switch may not be enough to handle the performance necessary for all traffic specifically in multirack and single site deployment or when one switch fails. Internode traffic in a single site multirack deployment traverses through one rack, up to the customer network and down to the next rack of switches in addition to handling traffic associated with data, replication, and management. It is recommended at the minimum, four uplinks per rack (two links per switch) for performance and high availability. Since both the data switches are peers, if link to either switch is broken, one of the other switches is available to handle the traffic.
Network Latency is one of the considerations in multisite or geo-replicated environments. In a multisite configuration, recommended maximum latency between two sites is 1000 ms.
Understanding workload, deployment, current network infrastructure, requirements and expected performance is fundamental in architecting ECS network connections. Some additional areas to understand include:
Note: Network performance is only one aspect of overall ECS performance. The software and hardware stack both contribute as well.
Best practices: