Enterprise Architecture: From Incite comes Insight...

Why haven't venture backed startups figured out that the CTO should be the only one blogging and not the folks in marketing...



I was looking at the blog of Ram Appalaraju of Azul Systems and wonder if he realizes that blogs aren't about thinly-veiled marketing material. At no time does he provide any opportunity for customers to engage in a conversation. In his own words:Let me highlight some of the problems and pitfalls that come up with the vendors when touting their product performance, so I figured I would do the same.

In the case of Vega 3 it’s not a bad story: 70% higher performance than Vega 2, and a whopping 5X faster than Vega 1.

I love the way that no timeline was indicated. I bet it doesn't come as a surprise to anyone that an Intel Core Duo is at least 5x faster than a 80286 chip! The key metric for smart customers to understand is to see performance increases compared against timelines.

It is also a hard point to prove outside of our own headquarters, although some customers may choose to compare their existing Vega 2 deployment to Vega 3.

I am not sure why compute power and benchmarking outside of one's own headquarters is that difficult. Has anyone ever heard of grid computing? If I can access the Amazon compute cloud or the Sun Grid, why can't the community help Azul benchmark?

Typically customer use cases validate the benchmark or initernal company specifications (the two listed above), but then go much further in telling the whole story. It may be one thing that it’s 5X faster, but if it takes 10 weeks to get up-and-running it’s something different.

Many of us enterprisey types write horrific code and use horsepower as a way to cover up our incompetencies. The funny thing is that we are even better at covering up mistakes by simply overprovisioning servers. Has anyone ever noticed that the vast majority of servers in a corporate data center at peak never seem to utilize on average more than 20%? Reality says that performance sucks but the answer to it isn't either faster nor more CPUs as the vast majority of our applications block due to I/O.

However, these are usually the hardest pieces of product data to obtain because if the customers have achieved great things they want to hide it from their competitors, and often companies have strict policies on what information can be shared externally.

This is absolutely the biggest lie ever told! 99.9% of all vendors and industry analysts get the enterprise perspective twisted in two ways. First, there is a difference in an employee telling a story vs a vendor telling the story on behalf of a customer. Folks know that I speak at a lot of industry conferences on a variety of topics as do many of my peers, but we also don't allow vendors very often to publish on our behalf. If you want communication, then put us in front of an audience.

More importantly, there is not a single policy that prevents any of my peers from publishing a benchmark! However, there are hundreds if not thousands of contract clauses from many vendors that prevent me from sharing them. Of course, vendors would like for them to be published only if they are positive, but otherwise attempt to handcuff the enterprise if they aren't so shiny. I wonder if any customers of Azul would be shutdown if they published their own unfavorable information?

I guess the point I'm getting at is that we marketers can throw a lot of "evidence" and supporting numbers at you (such as the above 3) to show how great our product(s) is, but unless a product meets your exact needs, our numbers and evidence don't provide much value. That’s the other fun part of my job: making sure the products meet your needs!

Marketing is more than just talking about products. How about talking about the people who work in your organization and how they are benefiting open source and the community as a whole? How about providing best practices for writing Java applications that can efficiently utilize such capacity? How about helping the community brainstorm ideas for new enterprise applications that weren't previously possible due to capacity...

Am I missing something?