No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search


To have a better experience, please upgrade your IE browser.


Inferior Switch Performance Leads to Low PoC

Publication Date:  2013-07-23 Views:  160 Downloads:  0

Issue Description

Product name and version: N8500 V2R1C00
When a CentOS client is used to perform a dd performance test for the N8000, the test is normal, and the bandwidth of the client is 100 MB/s. When two or more clients (mounted through different IP addresses) are used to perform a dd test concurrently for the N8000, the performance of a single client is impaired, and the total bandwidth of all the test clients is 100 MB/s.

Alarm Information


Handling Process

1. Ask the customer to replace CISCO switches with H3C switches.
2. After switches are replaced, the bandwidth for concurrent client access is normal. When a dd test is performed on four clients, the performance can reach 380 MB/s, which almost brings the N8000 performance to its fullest potential.

Root Cause

Fault location process:
1. A dd test is performed on the local file system of the N8000 at a rate of 350 MB/s. A test N8000 node contains 12 NL_SAS disks that are RAID10-protected. The performance of disk arrays and engine node is normal.
2. A dd test is performed on a single client at a rate of 100 MB/s, which is a normal rate and ensures that data can be normally accessed.
3. A dd test is performed concurrently on multiple clients. The rate cannot reach the normal value, and the N8000 performance is impaired.
4. Use the bandwidth between the iperf test client and the N8000. A single client can normally receive and send packets, but multiple clients cannot normally receive and send packets at the same time. As the number of clients increases, the bandwidth of a single client decreases. This indicates that the bottleneck lies in the service switch network. The test switch is CISCO x65.