pugovok
By:
pugovok

Why there is such a different in speed?

November 3, 2015 1k views
DigitalOcean

2013 year Amsterdam1:

dd if=/dev/zero of=tempfile count=512 bs=1024k conv=fdatasync
512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 1.68047 s, 319 MB/s

2015 year Amsterdam2:

dd if=/dev/zero of=tempfile count=512 bs=1024k conv=fdatasync
512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 18.5109 s, 29.0 MB/s

Why there is such a different in speed? Could you show your results, please?

2 Answers

Because you're not the only one with a droplet on that server.

@pugovok - Each central node (read: bare metal server) houses Droplets for multiple customers. Each customer, IIRC, is given what is commonly referred to as "fair share" usage of the resources available when it comes to I/O and CPU, with Disk Space and RAM being the guaranteed resources. If a node is new, or fairly empty, the I/O from the resulting command will be higher than it would be on a node that is 1/2, 3/4 or entirely full (by DigitalOcean's standards).

That said, unless you're consistently writing 1MB files (or files equal to the size passed to the command), the command isn't a true indication of speed. It can provide a baseline by which to gauge, but you should run the command and use more realistic file sizes (unless you truly are writing 1MB files all the time).

Have another answer? Share your knowledge.