Blogs

RedMonk

Skip to content

Power’s Great, Knowing How to Use it is Better

On Wednesday of last week I had the opportunity to sit down with Sun’s Aisling MacRunnels, senior director of utility computing at Sun and discuss the state of their utlity computing – grid, if you prefer – efforts over at network.com. One of the questions they posed to me during the session was whether or not I’d signed up for an account. The answer to that question was no, I had not. I’ve walked around the site, checked out some of the examples, but never signed up for an account. Why not? Because it’s not entirely clear to me what I would do with it if I had it.

If I was signing up for a plain old hosting account, I can think of a dozen or more things I could do with it off the top of my head: run a blog, run an email server, run a website, run a stats package, run a game server, run a wiki, run forums, and I think you get the idea. When I think about what I’d use a grid for, however, I’m distinctly uncreative. I can’t think of anything obvious that I’d employ it for. Oh, I could probably use it, along with distcc, to farm out my Gentoo compiles to a much larger and more capable network, but is the hassle of getting that set up more than its worth? I haven’t found out because I haven’t taken the time to figure out how and why I’d do that.

What is my point here – that grids serve no useful purpose? Hardly. What I’m trying to point out, instead, is that with new offerings comes additional responsibility with respect to education. It’s not enough to tell me how powerful, secure or flexible the grid is – I need to be educated on how it’s relevant to me specifically. Preferably with a wiki that would offer a HowTo style set of instructions that would let me get up and running. What are customers using it for? What can I try quickly that will demonstrate the value of the offering? [1] How, in other words, do you begin transitioning from a niche offering that appeals to climatologists or actuaries to an extension of a developers workspace, on demand hardware? Education, it seems to me.

I can say this with some authority because speaking candidly, we at RedMonk are faced with precisely the same problem. We work so differently than some of our competitors that half our challenge is explaining just what we can do for customers, current or potential. We have to spend a considerable amount of time getting customers in the mindset that it’s more about asking what we don’t do (i.e. commissioned whitepapers) than what we do. How many analyst firms are about driving you web traffic? But ultimately, we can’t ask customers to answer those questions themselves; at the end of the day, we need to do a much better job of educating customers on the myriad things we can do for them in the marketing, community building, networking, product strategy, etc areas. That’s one of the things I’m hoping to address in the long awaited RedMonk 3.0 site (I see an FAQ in our future).

So to the folks at Sun Grid, don’t feel bad – you’re in good company ;)

[1] The “create a spoken-word mp3 audio file from your text input” thing isn’t that appealing – I could already do that here – but speech to text? *that* would be pretty cool.

Categories: Trends & Observations.

  • Chris Rijk

    Yeah, it’s tricky to explain. For apps suitable to run on the current Sun Grid, they need to be batch processes – the more parallel and CPU intensive the better. If you don’t already have any such batch processing tasks, then it’s hard to understand.

    Good examples are essential to get real interest, though in Sun’s case, in the short term, I think it’s more a matter of converting existing grid users. They already have money after all.

    I wonder if we’ll ever see a tiny start-up specifically use the Sun Grid (or similar) to help accelerate their growth (and reduce risks). Imagine some people come up with a hot new for the stock market, or whatever. The computational analysis may require a lot of processing, but if it gives someone an edge, it can be worth a lot of money. But for the developers, they need customers (maybe they have some friends who can help kick-start things) but they also need a compute center to do the processing. Normally, you’d try to get VC funding, get some staff, buy a load of computers etc. But with a utility grid, you could simply offload the processing for each new request onto it. If you are paid by customers for each request in a way that is proportional to the processing cost, then financing becomes pretty simple. This relies on being able to estimate the compute cost reliably enough, but in many cases, I don’t see this being too much of a problem. So, imagine you’re charging on average about $3 a compute-hour when it costs you $1 an hour. That’s a pretty nice margin if you’re doing thousands of hours a day. And you’d have minimal capital costs and managing your infrastructure becomes simple. Bye bye VCs as well. Sounds almost too good to be true. (Maybe it is – I’ve only been knocking around such concepts in my head for a while).

    PS Do you process the logs for your website at all? If you do, then that’s something you could possibly farm out to a utility grid. I don’t think too many web log processing programs are grid enabled though – probably only the really expensive ones…

    PPS I don’t think something like speech to text would necessarily work that well. The better programs have a learning process (ie they’re interactive). And it’s not that compute intensive, unless you have a LOT of MP3s. Even then though, if your networking isn’t too fast, it might take a long time to upload all that data. Actually, log processing might have a similar issue of data size / processing time, that doesn’t make it suitable. (But last I heard, networking speeds are still increasing at a faster rate than data processing speeds, so such issues might fade with time).

  • http://www.activegrid.com Peter Yared

    We’re looking to use a grid to test horizontal scalability of ActiveGrid LAMP Server. It’s a lot cheaper to rent a bunch of servers by the hour and hammer them than to buy a bunch of servers and maintain them. The problem of course is getting your software up there, load balancing it, etc. The Sun Grid stuff is not very appropriate for our kind of workload, but we are starting to work with another grid provider that has not launched yet. Peter