In my HDD performance analysis series, I would now like to move on and have a look how fast my ‘real world’ data on my backup hard disks can be read. At the moment, I have around 7 TB of data on my backup drives, which consists of a significant amount of very large virtual machine snapshot files with a size in the double digit gigabytes, many smaller sized image files of 2-3 MB and an even bigger number of very small document files of a few hundred kilobytes at most. So how fast can I read such a data mix from hard drives with moving heads?
Let’s start with my slowest 8 TB drive. The graph above shows the data read speeds over almost 7 TB of data. The read data rate is not uniform at all, which is probably due to the clustering of my files. The super large gigabyte sized files are in one directory tree, while the smaller files are in other directory trees. The overall data transfer rate was 121 MB/s and it took around 16 hours to read all of the data.
Next up is another 8 TB drive of mine from a different manufacturer. The overall read data rate was around 145 MB/s, i.e. pretty similar to the drive above. The graph looks very different, however, despite reading the same data from it:
From the graph, one could interpret that the data is much more distributed over the drive. But I admit, that is just speculation. As the average data rate is very similar to the other 8 TB drive of a different manufacture, it doesn’t matter much for my purposes.
While the previous two 8 TB drives were about 4 years old, let’s have a look how data is read from my 20 TB drive, that is about half a year old at the time I wrote this post:
Average read speed over the same mixed data as used on the drives above was 220 MB/s, i.e. significantly faster than the data read from the older drives. Also, while the read speed from second to second differed tremendously on the previous two drives, the graph for the 20 TB drive looks a lot more straight over all.
And finally, let’s come to my brand new 16 TB drive with the same amount of data.
Average read speed was 217 MB/s, i.e. only slightly slower than the 20 TB drive above. However, the graph looks again completely different. One potential explanation I have is that while my real world data on the 20 TB drive has been updated many times in recent months, the same data on the 16 TB drive was only written once. No way of knowing for sure, however. But in return this would mean that despite updating the data many times on the 20 TB drive, the read speed is still very good, despite the data perhaps being more distributed due to the overwriting over time.
Summary
While there are a number ‘maybes, coulds and potentiallies’ in the above text, it is quite clear that the newer 16 and 20 TB drives are performing a lot better faster than the older 8 TB drives. One additional reason for this: The 16 and 20 TB drives are nowhere near full, i.e. the fast areas of the drive could preferentially be used for the data, while the older 8 TB drives are almost full and hence, a lot of data is located in the inner and thus slower parts of the drive. As shown in part 2 of this series, write speeds on the newer and bigger drives also reduce to around 120 MB/s on the inner parts of the platters, so there will probably also be a noticeable slowdown effect on those drives as well as more data gets written onto them. Average write speeds of those bigger drives across the whole drive was around 177 MB/s for one 198 MB/s for the other. Based on this, I would not expect to overall average speed to go down below those values. But that’s of course just an assumption, time will tell.
And again so much for today. In the next post, I’ll have a look at read vs. write speed of the fastest drive.