Skip to content

2008

  • 1 min read

Cloud Computing case studies

I've written about some of the considerations that go into the choice between physical infrastructure and new SAAS services such as Amazon's S3 and EC2. I also covered why the fact that you are comfortable relying on a hosting provider for rack mounted servers (so called ping & power) doesn't mean necessarily that you would come to the same conclusion about SAAS services.

Here's what the Wall Street Journal had to say about that tradeoff just a few days ago:

"Today was a bad day for a new computing model that could one day be the norm. Amazon’s S3 service –which companies can use to rent data storage on Amazon’s tech gear — crashed this morning, knocking many small businesses offline and highlighting one of the model’s drawbacks: You’re putting your operations in somebody else’s hands." (from Is Amazon’s Small Crash a Giant Crash for Cloud Computing?)

In researching these three posts, I came across this which reminded me that this wasn't the first time this happened, I guess not surprisingly:

"Cautionary tale indeed. It’s the other side of the wonderful world of mashups and web 2.0 and web services and all that jazz. If I build my product on the back of your service, then the quality of what I deliver depends on your carrying through on your promises. Not a very strong position to be in." (from A cautionary utility computing tale - or the dark side of Mashups)

  • 2 min read

5 things to keep in mind when relying on S3 and similar SAAS services

On a project I worked on recently, I asked myself whether I should make one or more of Amazon's awesome web services (AWS - Awesome Web Services) a mission critical part of the infrastructure of the product. I wrote yesterday about the considerations that go into deciding to rely on services like Amazon's web services as key infrastructure components.

Of the five I mentioned, these two considerations were especially confusing to me:

  • What is the comparable reliability?

  • If the whole scheme doesn't work out, how hard is it to switch?

On the one hand , the reliability of running a bunch of servers in a data center is fairly well understood, as well as the contingency plans to deal with hardware and software failures.

On the other hand(using S3, the storage service as an example,) the expected reliability is more or less unknown - although preliminary data is highly positive - and S3 is the only service of it's kind, so that there is a certain unavoidable amount of lock in.

I expressed that as the rhetorical question,

" Do I really want to entrust a business critical function to another company - if they decide to shut me off, my business is dead in the water."

Sounds ominous , doesn't it?

But wait , I entrust running my servers to a hosting service, don't I? If they shut me off I am kind of dead in the water too, aren't I? What's the difference?

Relying on your hosting provider for ping & power has a different risk

profile than relying on a SAAS provider for disk & cpu in the cloud

Here's why I say that:

  1. There are many many hosting providers (ISPs) to providing ping & power. It's a vibrant and competitive space.

  2. We understand them well. We understand their services, their pricing, their terms of service.

  3. They generally rely mostly on a well known and understood set of technologies, interfaces, software and hardware.

  4. It is feasible to switch from one to another.

  5. It is also feasible (and many do) to have parallel relationships with two totally different hosting services so if one shuts down, crashes, goes out of business, or whatever, you can seamlessly switch to the other.

  • 1 min read

5 Questions to ask when deciding whether to rely on S3 et al

Several products I've worked on recently have relied on Amazon's really impressive web services. I've written before about how much I admire services such as S3 and EC2, solid, commercial quality infrastructure services.

They address (and? or? solve) a big problem that application builders have:

"How many servers will I need, how many users will I have, how much storage will I need … and in the end, how much will all that cost?".

They solve this by offering key components of infrastructure as a "software as a service", SAAS. Starting with permanent storage, which is delivered by S3, followed by computes or servers, delivered by EC2, and further followed by a growing variety of other infrastuctural services.

Let's look at what questions to ask when deciding to use any of these services. For example, when deciding to depend on Amazon's S3 storage service (delivered over the internet, no hardware required) instead of buying a series of file and database servers (hardware) you need to consider the trade- offs:

  1. Given my expected growth, what is the comparable cost?

  2. What are my options in each case if my growth doesn't actually develop as I expect?

  3. What is the comparable performance?

  4. What is the comparable reliability?

  5. If the whole scheme doesn't work out, how hard is it to switch?

You see that you end up with a fairly tricky decision matrix. More tomorrow on how I solved this dilemma.

  • 1 min read

The downside of free software

Yesterday the New York Times had an interesting article about our expectation that software (and web sites and web applications) all should be free. I've written about this before so I won't belabor it again. From the article:

A growing number of consumers are paying just that — nothing. This is the Internet’s latest phase: people using freely distributed applications, from e-mail and word processing programs to spreadsheets, games and financial management tools. They run on distant, massive and shared data centers, and users of the services pay with their attention to ads, not cash. (from New York Times)

I have more than a passing interest in this because as you may know I am the proprietor of BlogBridge, which is an open source application plus service which is available free. We modestly offer a few enhanced services for not- free, and the reality is that the majority of our customers opt for the free choice. You can understand why the topic of free and not free software is interesting to me.

Last month, a really great and popular web based application, after a few years of free service, decided it had to close it's doors. From the final post by Greg Linden, it's intrepid leader:

"Findory turned off its last webserver today. Sadness." (from Findory turns off the lights)

As it turns out, "We give it away free, and make it up in volume" doesn't always work.

  • 1 min read

Harper’s on torture

Came across this interesting article. You may agree or not, but it's definitely interesting:

"So now the process can be fully diagrammed, and the cast of characters is stunning. The torture system involves the operations division of the CIA on the implementation side. They rely heavily on contractors, it seems, in torturing people. And a special role is apparently played by a couple of psychologists. (from Harper's Magazine)

p.s. Happy Valentine's Day 🙂

  • 1 min read

From Demo 2008: Cool research site – SiloBreaker

Take a look at SiloBreaker (don't ask me why it's called that; is it supposed to break down silos?)

Try typing something into the search box, say "Pakistan". You will see that it identifies it as a country (it would also identify companies, people, and who knows what else.) Right off that's interesting, right?

Scroll down and in the right margin you see boxes with some rather unique perspectives: maps, article volumes, concept network, hotspots, trends, quotes, and lots of semi structured analysis.

It's a hard site to describe but looks like a unique and useful research tool.

  • 1 min read

Who’s your friend?

I just looked through my friend list and I see some people who's name I recognize but don't know exactly the context where I knew them from, and then there are mysteriously who I really don't think I've ever known, so I can't remember how they got there.

I was discussing with some people at the party how we are going to soon have a new word to distinguish a 'real' friend from a 'facebook' (or other social network) friend, as well as a term to distinguish a 'real' social network from one on the computer. See one of my favorite new terms, Retronym.

  • 1 min read

About Yahoo / Microsoft and an easier way to get $2B in revenue annually

You can read all about the potential acquisition of Yahoo by Microsoft, here for example, where Marc Andreessen writes:

"I think that a Microsoft/Yahoo merger would have practically no impact on any high-quality Silicon Valley startup." (from Silicon Valley after a Microsoft/Yahoo merger)

You should read his whole paper to get his 5 point argument. And I definitely agree with this conclusion:

"Build something of value -- something that people want, and something that will be profitable at the appropriate point -- and the world is yours. [snip…]

Or, reduced to a phrase: the best way to get bought is to not be for sale. " (from Silicon Valley after a Microsoft/Yahoo merger)

Good stuff , especially that last conclusion.

Personally, I am shorting both Microsoft AND Yahoo.

I predict that the 'integration' (a word used very loosely) will give both Microsoft and Yahoo a huge amount of indigestion because I think it's basically not possible.

The expected 'leverage' and 'reduction of duplication' will actually be an extended period of executive and strategic distraction , loss of the best people at Yahoo, and in the end, nothing.

I think Microsoft would do better putting those $40+ Billion in the bank and get a nice $2,000 million revenue stream annually, forever.