OK, my tech friends, here's a puzzle in three questions. It concerns the following graph, created courtesy of pcp/grafana.

The questions:

  1. What should the units have been on the title/legend line, instead of "kbytes"?
  2. What kind of traffic might have caused this I/O throughput?
  3. What make/model disk drive could this be?
Please explain briefly.

update: answers below

  1. Units should have been rates, i.e, "kbytes/sec". The vertical scale numbers are thus in kilo*kilobytes/sec, i.e., megabytes/sec.
  2. This trapezoidal curve is typical of sequential I/O (in this case writing) to a traditional spinning hard disk drive (not SSD), from the outer track inward. Initial speed is higher because the linear speed of rotation is higher at an outer track (proportional to radius). It has to be sequential I/O because of the regularity.
  3. Estimating the area under the curve, one gets approximately 2 TB of traffic. The peak (initial) rate of apprx. 110 megabytes/sec is consistent with several drives, but most consistent with the Seagate Barracuda LP series, wherein 5900 rpm, 1417 kilobits/inch max. linear recording density, and vaguely estimated disk outer track length of 6" make the number match up nicely:
    % units
    Currency exchange rates from www.timegenie.com on 2014-04-02 
    2866 units, 109 prefixes, 79 nonlinear units
    
    You have: (5900/minute) * (1417 kilobits/inch) * (6 inch)
    You want: bytes/second
    	* 1.0450375e+08
    	/ 9.5690346e-09
    
    It is interesting to see that the final transfer rate was just around ½ of the initial rate. This means that approximately half the radius (thus ¾ of the surface area) of the hard disk is actually in use, and the inner rest must be taken up by the spindle axle / motor hardware.
Posted Sun Jan 18 17:46:00 2015 Tags: