[
https://issues.apache.org/jira/browse/TUBEMQ-120?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Guocheng Zhang updated TUBEMQ-120:
----------------------------------
Fix Version/s: (was: 0.5.0)
0.6.0
> Aligned disk data storage
> -------------------------
>
> Key: TUBEMQ-120
> URL: https://issues.apache.org/jira/browse/TUBEMQ-120
> Project: Apache TubeMQ
> Issue Type: Sub-task
> Reporter: Guocheng Zhang
> Assignee: Guocheng Zhang
> Priority: Major
> Fix For: 0.6.0
>
> Attachments: image-2020-05-15-11-28-20-118.png
>
>
> 1. Data read and write operations should consider the characteristics of the
> disk, for example, the disk is based on 512-byte sectors as its storage unit,
> and read data in batches of 64k; the file system will eliminate the cache
> according to certain rules Pages in memory etc. If the read and write
> operations take these contents into account, I believe that the current TPS
> can be higher;
> -----------------------------------
> I understand the reason for this problem, the problem I think is that the
> data alignment needs to be considered when storing data to disk, so that the
> head can read the data with as few accesses as possible. For example, the
> data b in the picture， when it unaligned stored， two sectors need to be
> accessed, but the head only needs to access and read one sector after aligned
> storage：
> !image-2020-05-15-11-28-20-118.png!
> The performance difference of this problem may not be obvious under a single
> access, but under the reading of massive messages, the performance will be
> essentially improved from quantitative to qualitative changes, especially
> since TubeMQ internally reads data randomly, it is more necessary to consider
> this .
> After the modification, a new problem introduced based on this modification
> needs our attention, that is, the storage space used after the system
> modification will increase under the same data amount. As shown in the
> picture, because the data is aligned, for small packet data (below the sector
> size ) a, will occupy more space. However, considering that the overall
> performance of the disk is unchanged, and the disk space can reach a very
> high capacity, this waste is considered acceptable to me
> For the optimization of this problem, do you have any suggest ? If not, I am
> going to claim this modification and optimize it according to this idea.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
