Subject: Re: how to create 21,780 new free software jobs (2,530 in R&D)
From: Seth Gordon <seth@genome.wi.mit.edu>
Date: Tue, 05 Nov 2002 10:40:11 -0500

Tom Lord wrote:
> 
>         bfox:
> 
>         The nightmare of updating 5,000 user's desktops in a small
>         period of time
> 
>         [...]
> 
>         Oh, wait, then they won't be all running the same OS
>         version exactly, and you'll be fighting a
>         multiple-version-os nightmare in support.
> 
> Ah -- you seem to have ignored the premise of the plan or to be unware
> of the historical practice that validates that premise.  All those 60k
> users per community are getting their platform via a shared
> distributed filesystem.  They are all in sync.

Here's a traceroute from a server where I work, the Whitehead/MIT Center
for Genome Research, to MIT's Web server:

10:01am seth@iron
~/Working/sequence/pipeline/closure_tools/primer_picker/test-1096 >
traceroute web.mit.edu
traceroute to web.mit.edu (18.7.21.77), 30 hops max, 40 byte packets
 1  B24-RTR-2-WIBR.MIT.EDU (18.4.1.1)  1 ms  0 ms  1 ms
 2  W92-RTR-1-BACKBONE.MIT.EDU (18.168.0.25)  1 ms  1 ms  0 ms
 3  SICARIUS-SPATULATUS.MIT.EDU (18.7.21.77)  0 ms  0 ms  0 ms

Here's a traceroute from the server at my house, connected by DSL, to
the same IP address:

bash-2.05$ traceroute 18.7.21.77
traceroute to 18.7.21.77 (18.7.21.77), 64 hops max, 40 byte packets
 1  gw-081-076.dsl.speakeasy.net (66.92.76.1)  41.856 ms  33.215 ms 
27.403 ms
 2  border8.ge3-2.speakeasy-16.bsn.pnap.net (63.251.141.212)  23.74 ms 
22.156 ms  21.333 ms
 3  core4.ge1-0-0-bbnet2.bsn.pnap.net (63.251.128.70)  21.987 ms  21.652
ms  23.775 ms
 4  bos-edge-02.inet.qwest.net (63.145.1.93)  24.146 ms  21.790 ms 
23.156 ms
 5  bos-core-02.inet.qwest.net (205.171.28.29)  23.17 ms  24.319 ms 
23.195 ms
 6  bos-brdr-01.ip.qwest.net (205.171.28.34)  26.600 ms  21.342 ms 
21.454 ms
 7  * * *
 8  192.5.89.90 (192.5.89.90)  26.678 ms  26.268 ms  37.979 ms
 9  W92-RTR-1-BACKBONE.MIT.EDU (18.168.0.25)  24.273 ms  22.782 ms 
27.95 ms
10  SICARIUS-SPATULATUS.MIT.EDU (18.7.21.77)  27.617 ms  26.454 ms 
25.316 ms

My house is literally across the street from the local Verizon switch,
and about three miles from MIT, and I pay over $60/month for service
from Speakeasy, one of the most reputable ISPs in the Boston area.  Yet
the difference between my house and MIT's campus network costs me over
an order of magnitude in latency.

As a further data point, IDE hard drives seem to advertise a latency of
between 4.1 and 5.5 ms.

If I tried to use NFS/AFS/Andrew to run desktop applications over a link
with this kind of latency, what kind of performance would I see?

-- 
"The market can remain irrational longer than you
 can remain solvent."  --John Maynard Keynes
// seth gordon // wi/mit ctr for genome research //
// seth@genome.wi.mit.edu // standard disclaimer //