The Riverbed Blog (testing)

A blog in search of a tagline

Shifting Models & Lumped Circuits

Posted by riverbedtest on August 22, 2011

Today's Guest Blogger is Mark Day, Riverbed's Chief Scientist.

Back when I was studying electrical engineering, I first learned about lumped circuit models.  In a lumped model, we considered the wires to be ideal and just focused on the behavior of the connected Lumpedcircuits components (resistors, capacitors, inductors).  After we knew what we were doing (more or less) in that simplified world, we learned about transmission-line models, where we modeled the behavior of the wires.  And we learned that for certain kinds of real-world problems like managing an electrical grid, a lumped model would give you hopelessly wrong answers.

A little later, when I was in graduate school for computer science, I read an entertaining rant about ideal wires vs. the reality of building a fast parallel computer. Today that item came back to mind as I was thinking about explaining WAN issues and cloud performance to people whose frame of reference (their model, if you will) might be mostly LANs.

I was happy to find that my memory had mostly served me correctly.  Here’s a relevant excerpt from Danny Hillis’s book The Connection Machine (MIT Press, 1985):

“Fundamental to our old conception of computation was the idealized connection, the wire.  A Connection Machine wire, as we once imagined it, was a marvelous thing. You put in data at one end and simultaneously it appears at any number of useful places throughout the machine.  Wires are cheap, take up little room, and do not dissipate any power.

“Lately, we have become less enamored of wires. As switching components become smaller and less expensive, we begin to notice that most of our costs are in wires, most of our space is filled with wires, and most of our time is spent transmitting from one end of the wire to the other.  We are discovering that it previously appeared as if we could connect a wire to as many places as we wanted, only because we did not yet want to connect to many places.”

I think you can see that some of the same reality-check critique applies to ideas about the performance of cloud computing and distributed computing across WANs, with the network taking the place of the wire.  It sure is simple when the network’s behavior doesn’t matter, but unfortunately that isn’t always true.

At Riverbed it’s nice to have a variety of technologies that can be brought to bear on those network-related issues. One subtle problem is that because we think about it all the time, it’s easy for us to take for granted that “switch of models” that is sometimes harder for our customers.  People sometimes have to shift from assuming that everything “just works” in ideal fashion, to actually thinking about the WAN as an element of the system.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: