28 reputation
5
bio website
location
age
visits member for 10 months
seen Jan 10 at 9:10

Jun
30
awarded  Popular Question
Dec
2
awarded  Scholar
Dec
2
accepted Relation between disk IOPS and sar tps
Dec
2
comment Relation between disk IOPS and sar tps
Ok, you're right IOPS does not make any sense for writes as there are lot of caches, reordering and merges involved. Closing... Thanks.
Dec
2
awarded  Supporter
Dec
2
comment Relation between disk IOPS and sar tps
Yeah, you're right. But in my case there's almost 100% of writes, reads are almost none.
Dec
1
comment Relation between disk IOPS and sar tps
The reason is I need to describe a minimum HW requirements. I have a server that receives data over network (we can assume constant bitrate here) and writes received data to disk. Data is written to files sequentially but there could be hundreds (e.g. 800) of them in parallel. I have figured out that when number of clients reach some point I start to get large iowaits. Actual disk throughput I can achieve is about 25MB/s which is quite low, less clients with higher bitrate can do 35MB/s, pure sequential about 130MB/s. So I guess IOPS is what matters here...
Dec
1
comment Relation between disk IOPS and sar tps
Ok, I tried a 10GB file through dd with 256kB block to actually fill the cache and after ~90 seconds tps drops to ~200, so maybe you're right. But still 80 and 200 is quite a difference... Is it possible that read and write IOPS differ? And is there any way to figure out required IOPS from this value?
Dec
1
awarded  Editor
Dec
1
revised Relation between disk IOPS and sar tps
Added OS info.
Dec
1
awarded  Student
Dec
1
asked Relation between disk IOPS and sar tps