Peeling back the Onion on a Sales Call

It's interesting to see how "potential customer" interests have changed over the last 5 years.

In 2003, when I was first meeting with customers, talking about this product we were working on, everyone agreed that they had challenges in their network with remote users. The challenges varied — either performance of centralized applications/servers, or backing up their remote servers (ok, and a few who wanted to centralize).  But everyone blamed bandwidth as the culprit for poor performance.  Often they couldn't afford the larger links, but sometimes they had already upgraded and were scratching their heads as to why it didn't help as much as last time.

So, since they viewed it as a bandwidth problem, we'd explain our killer compression (SDR), and they'd spend 90% of the meeting focused on that, then graciously allow us to talk about what we wanted to discuss, which was how transaction prediction (and VWE) would ameliorate the effects of latency and really speed things up. Sometimes light bulbs would nearly visibly go off over people's heads, usually I got a "hm, I'm sure that'll help too, now back to SDR, what happens when..."

Later they'd test the product (as no one believed our speedup claims...remember WDS hadn't even been invented as a term and we were this dinky startup), and be amazed.  The vast amount of their speedup, of course was due to transaction prediction, the majority of the time.

So, how about 2008?  Last week I was meeting with a large financial enterprise, discussing their challenges.  They knew that latency was the root of application performance problems, and we spent most of the time talking about that (still did discuss SDR — as they'd rather not get more OC3s).  But they were puzzled with their test results — things were clearly faster (i.e. stopwatching user actions), and the reports showed dramatic bandwidth reduction on the Steelhead, but the routers reported the WAN links were still full — where were the "savings" coming from.  Anyone see the answer to this mystery?

Well, the data reduction enabled more traffic to flow, and the transaction prediction stopped applications being bottlenecked on the WAN, also enabling more traffic to flow, so...  People got more done — in the same amount of time — thus the WAN was "full again."  It's sort of Parkinson's Law experienced in a different domain (and with a more hopeful spin).

So, how far have we come?  It's mixed, the great news is that people are aware of latency and it's insidious effects.  We still have a ways to go with the second order implications of this though, i.e. that filling a WAN link can actually be a good sign.  And someone needs to come up with a way to measure ROI on knowledge workers — surely them getting more done in a day is valuable to the enterprise. 

So, what do you think...Latency*chattyness is now well understood, or still breaking news?

Comments

[Anthony Giannecchini]

Just a comment. When writing these articles, it is not correct to assume your audience understands all of your acronyms. These should be "spelled-out" when first used, and then referenced as an acronym.

[Steve Smoot]

Thanks for schooling me on my first post! Sorry for any and all confusion, here are the defs:

SDR - Scalable Data Referencing - what we do to "dedupe" data on the fly

VWE - Virtual Window Expansion - this is part of how generic TCP streams are accelerated (hard to describe in brief)

WDS - Wide area Data Services - the name for the space that "WAN Accelerators" play in - speeding up remote email/file share/web systems (etc). Basically making remote offices perform better or else enabling consolidation of IT resources.

OC3 - 3 OC-1s (Optical Carrier 1 - a SONET thing) basically 155 Mbps.

WAN - Wide Area Network (just for completeness).

Thanks again for the tip.