Write performance

[francisco_reyes]

Looking at ADATA SP 550 based on its listing in Best Value Drives. However, seeing many reviews on amazon that claim that write peformance goes down significantly after a certain amount of sustained writes to disk. Does performance test takes this into consideration? Look into drives for a database machine so writes are a very big component.

[Richard (PassMark)]

We don't (or at least very rarely) benchmark drives ourselves for the hard drive charts. Nearly all data in the charts comes from user submissions from our PerformanceTest software. The results are an average of all submissions. As such we do not know when in the lifetime of the drive that the benchmark took place when the user submitted the results.

[David (PassMark)]

Most databases have a lot more reads than writes done on them.

I think the slow down in performance you are referring only occurs after several hours of uninterrupted writing. If you are referring to the Futuremark Storage Consistency Test results, then these tests basically fill the entire drive with data 4 times take their 'degraded' and 'steady state' measurements. For most people this isn't a realistic test. Even for databases. Are you planning on filling the drive in the first hour of use?

[francisco_reyes]

Nearly all data in the charts comes from user submissions from our PerformanceTest software.

Yes. I am aware. Hence why I was asking if the software takes that into account.

[francisco_reyes]

I think the slow down in performance you are referring only occurs after several hours of uninterrupted writing.

Thanks for the clarification. Looking at this comment https://www.amazon.com/gp/customer-r...SIN=B013J7PP96 looks like degradation has been observed with less than one full write

Are you planning on filling the drive in the first hour of use

When loading the data (initial load) and also when fixing replication(1) may see a data close to capacity.

This other review claims that he noticed slow down after 6GB https://www.amazon.com/gp/customer-r...R2XUN0R4F57K7P

And another review https://www.amazon.com/gp/customer-r...SIN=B013J7PP96

None of the reviews are doing benchmarks. They claim they have noticed slowdowns during normal operations so I was trying to understand those reports vs the score in the site (from users running the benchmarks against their machines). There are several other comments claiming the same slowdowns so primarily trying to understand what the tests do, which you covered, 4 full passes of data, so thanks for the clarification.

Note 1: Doing postgresql and once replication is up.. writes are non continuous, but if replication drops to one of the slaves then fixing replication then a large continuous batch load will take place.

[David (PassMark)]

I still don't think you are going to have a problem. Initial load time shouldn't matter much (as it is a one off event). And even then populating a database isn't a linear continuous write operation. There are small pauses between the writes as the index is constructed and the SQL is processed. I think the drive will keep up without dropping off the performance edge.

More of a concern should be write endurance. If you are really writing data 24x7 non stop at near maximum throughput then you need to have a hard look at the endurance times.