Score:-1

Is Software RAID1 faster on a larger SSD?

nc flag

I am evaluating 2 CentOS servers, in terms of performance of reading and writing to the database. The database sits on an SSD RAID1 pair, with MySQL 5.7 on each server. Server 2's specs are a bit better overall, except this:

Server 1: Hardware RAID1, two 2 TB SSD drives.

Server 2: Software RAID1, two 1 TB SSD drives.

Server 2 is faster when it comes to READS.

But Server 2 is slower when it comes to WRITES. A performance test on Server 1 is 33% faster than on Server 2 (e.g. 140 sec vs 210 sec). The test is the same on both servers: inserting 1000s of rows of data to the database, 64b per row.

Software RAID is slower than Hardware RAID, so this slower operation could be understandable. But it was also suggested that the SIZE of the drive is an additional factor, i.e. that a 2 TB SSD drive will be faster than a 1 TB SSD drive.

Does anyone know if this is the case? I have not been able to find anything on this online. Any help would be appreciated.

us flag
There is no direct causality between size and speed. It all depends on the details of the technology that is used to implement a particular drive.
janman05 avatar
nc flag
@TeroKilkanen, thanks. So then it could be that the larger drive is better/newer technology, and this is the reason why I was told that larger is faster.
Neppomuk avatar
am flag
When posting a question about performance issues, you MUST give details about the HW and SW you are using. Here it's good to know which RAID card you are using as there are some "minimal HW" solutions on the market, which leave most of the RAID logic to the driver and therefore to the host CPU.
Score:0
br flag

As Tero says this all depends on so many other factors - for a software RAID it's your CPU doing the work, if it's already very busy then you'll see storage performance degradation, if it's very lightly used it could be faster etc. Also when you use software RAID every write has to be written by the CPU twice, one to each disk, with a hardware RAID it gets written to the disk controller just once.

Ultimately the only way to know is to test both with the actual use-case you need and decide from that - anything else is guesswork.

janman05 avatar
nc flag
Thanks. I am aware that CPU is doing the writing in this case, and I'm not trying to guess, but test and observe. The test was done on a brand new Server 2, recently installed, and not used at all. The reason I've asked this here, because the provider's sysadmin said specifically that drive SIZE was the reason, and I wanted to get another opinion, not really being sure that he's right.
ua flag
As for the Hardware version -- it may take 3 steps: 1 to write to the controller. Then the controller has to do 2 writes, preferably _before_ ACKing the client to say that _both_ copies have been safely written.
br flag
If the HW disk controller has some form of battery-or-flash-backed cache then that 'ack' happens as soon as the controller has all of the data to be written.
I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.