Reason why read and write performance is deteriorated

0

Reason why read and write performance is deteriorated:
Front-end links are faulty (error codes or the host) or the system has slow disks.
Error codes: Check the content of error codes.
Slow disk: Check the content of slow disks.
Host: Check whether the HBA card, network adapter, or upper-layer software is faulty.

Other related questions:
Method used to locate LUNs with poor read and write performance
Method used to locate LUNs with poor read and write performance: 1. The LUN write policy is write through. Locate the cause why the LUN write policy is write through. 2. If the host interface is iSCSI, check whether the network rate is 1 Gbit/s, whether the network is stable, and whether ping packets are lost. If the Fibre Channel or SAS network is used, log in to the storage's OSM page to check whether an alarm indicating that the front end encounters continuous error codes is reported. 3. Check whether slow disks exist in RAID groups. Specifically, run iostat -x 1 to check whether some disks differ from other disks (the util value of one or two disks is greater than that of other disks and reaches 100. If the util value of all disks in the RAID group reaches about 100, the configurations are incorrect and back-end disks are slow disks). If slow disks exist, replace them. 4. If a file system exists on the host, the file system produces enormous fragments when the disk space utilization rate exceeds 90%, causing the deterioration of read/write performance. You are advised to use disk space correctly.

Performance specifications of OceanStor 9000
Performance specifications of OceanStor 9000: - OPS: up to 25,000 (CIFS) or 20,000 (NFS) per node - CIFS stable read and ultimate read: up to 800 MB/s and 1280 MB/s respectively per node - CIFS stable write and ultimate write: up to 800 MB/s and 1300 MB/s respectively per node - NFS ultimate read and ultimate write: up to 1600 MB/s and 1100 MB/s respectively per node

Relationships between redundancy ratio, strip size, and read/write performance of OceanStor 9000
Relationships between redundancy ratio, strip size, and read/write performance of OceanStor 9000: The redundancy ratio and strip size of OceanStor 9000 directly impact on the storage space utilization. The strip size determines the data slice size. The data slice size is set based on the average size of files in actual services to prevent improper data slices due to a too large strip size, avoiding space waste. Particularly for small files, small strips of 16 KB are needed. The redundancy ratio determines the number of verification data copies and directly impacts on the storage space utilization. For the N+M redundancy ratio in the Erasure Code algorithm, M can be 1, 2, 3, or 4. A larger value of M indicates poorer algorithm performance and more complex computing. N can be 3 to 16. A larger value of N indicates higher algorithm performance and more disks concurrently read/written.

Why does the storage system performance deteriorate when the BBU or battery power is insufficient?
This symptom is caused by the security mechanism of the storage system. When the BBU has insufficient power or is faulty, the LUN write policy is automatically changed to write through. This compromises the performance of the storage system. If the LUN write policy needs not to be changed as the BBU power becomes insufficient, change the write policy to Write Back Mandatory.

Method used to plan LUN read/write policies
You can plan LUN read/write policies as follows: 1. Planning the write policy You can select the following write policies during LUN creation: write through, write back and mirroring, and write back without mirroring, mandatory write back and mirroring, and mandatory write back without mirroring. The following is the differences among write policies: Write through: writes data to disks. Each write operation must access the disk, contributing to low performance but high reliability. Write back: writes data to the cache and then write data to disks when there are idle host I/Os. Each write operation does not access the disk with high performance but low reliability. Write back and mirroring: writes data to both the local cache and peer cache simultaneously. Write back without mirroring: writes data to the local cache. Mandatory write back and mirroring: Storage devices must write data to the local cache and peer cache simultaneously. Mandatory write back without mirroring: Storage device must write data to the local cache. 2. Planning prefetch policies You can select the following prefetch polices: intelligent prefetch, constant prefetch, variable prefetch, and non-prefetch.

If you have more questions, you can seek help from following ways:
To iKnow To Live Chat
Scroll to top