Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[STDS-802-3-BWA] Correction of Storage Growth and Ethernet Presentation



Dear BWA participants,

 

I was glad to make my presentation to you on “Storage Growth and Ethernet” in Chicago (http://www.ieee802.org/3/ad_hoc/bwa/public/sep11/index_11_0912.html).  I talked about the difficulty in grasping such large numbers as Petabytes (PB) and Zettabytes and managed to prove it by missing one calculation by 3 orders of magnitude. 

 

On slide 8, I tried to show how long it would take to transfer 15 PB of data over various Ethernet line rates from 1GbE to TbE.  I mentioned that the times seemed rather short - I should have followed my instinct.  I was calculating sending 15 TeraBytes instead of 15PB.  The transfer actually becomes much more interesting when the correct numbers are used because it shows how it will be impractical to transfer 15PB of data over 1GbE.  A 1GbE link would take 3.8 years to transfer the data instead of 33 hours like the presentation stated. 

 

The time that it would actually take to transfer 15PB of data at different Ethernet speeds are:

       3.8 Years at 1GbE

       137.5 days at 10GbE

       13.75 days at 100GbE

       33 Hours at TbE

 

Since it takes CERN about a year to produce the 15PB of data, it’s not a good idea to take several years to transfer the data.  10GbE even looks slow at several months of transfer time for this big data.  CERN visited Brocade last week and mentioned that they are generating 25 PB of data each year now too!

 

If these calculations don’t blow your mind, maybe one of the most recent discoveries at CERN will.  They recently reported that they think they’ve measured neutrinos that travel faster than the speed of light.  There are many skeptics of this, but maybe the impossible just happened: 

http://www.dailytech.com/CERN+Physicists+Observe+First+FasterThanLight+LongDistance+Travel/article22827.htm

 

John has uploaded a corrected version of the presentation to the website.

 

Thanks,

Scott