How to achieve sequential IO in rocketmq and Kafka, because there may not be exactly 1g continuous address disk space in different operating systems.
Asked
Active
Viewed 74 times
0
-
You don't need continuous disk space... – OneCricketeer Dec 07 '21 at 17:35
-
Why don't need continuous disk space? Doesn't sequential IO in Kafka mean continuous disk space? please help me understand this。 – XHMao Dec 09 '21 at 07:25
-
Normal hard drives dont allocate disk space linearly. If they did, then you wouldn't have fragmentation. The question isn't really about Kafka since you're asking how the OS manages HDD drives. Besides that, your OS would start failing or swapping if you didn't have 1g of space available – OneCricketeer Dec 09 '21 at 19:46
-
I quite agree with this,however, what is the relationship and difference between the sequential IO in Kafka and what you said。I do not understand this. Is the sequential IO in Kafka not related to disk allocation? – XHMao Dec 10 '21 at 05:49
-
Well, for one, there's no guarantee at the OS filesystem partition scheme that'll force **linear** I/O reads and writes, so when you hear something like "sequential access", it means that SSDs and random reads are not useful, and therefore you can save yourself on costs by not buying those disks. Spindles in spinning disks move hundreds of times a second, and shouldn't really be something to concern yourself with – OneCricketeer Dec 12 '21 at 04:22