I haven't posted in a while as I've been given the added responsibility (and privilege) of running marketing at GigaSpaces. It means a lot more work for me, but it's also fun. One fun aspect of the job is telling the GigaSpaces story to different audiences in ways that are relevant to their world. And, yeah, getting some press.
So a few months back, Derrick Harris, now the former editor of GRIDtoday, posted this column, asking himself why the "grid vendors" don't have transactional grids on their radar. I wrote Derrick, telling him that the so-called "traditional grid vendors" that he's been talking to are not positioned to solve the particular issues related to transaction processing on the grid, namely managing data, workflow and achieving low-latency (and a couple of other issues). But GigaSpaces actually is, and if he'd like, I'd tell him all about it. Which I did.
Yesterday, GRIDtoday published the resulting story -- a very well-written piece on transactional grids. Greg Nawrocki also compliments them in his own well-written and insightful blog, Grid Meter, and says: "This is truly a glimpse of what the future holds for grid." I love the support, but of course, from where I am sitting, the future is now.
The GRIDtoday piece is especially great because the writers did not just count on what the vendors say, but talked to analysts and customers as well. And Bec Wilson from Sempra Energy Trading said some wonderful things about GigaSpaces that even surprised me (100X performance improvement!). Gartner's Massimo Pezzini also does a good job of giving context to the whole phenomenon. He's one of the few analysts out there who are actually trying to size, give names and categorizes these emerging technologies.
We've always been a bit ambivalent about the term "grid" at GigaSpaces. Partly because the term (not the technology) has been widely hyped, misused and misunderstood. But to me it has quite a simple definition. I arrived at it when I read this blog from Lee Gomes at the Wall Street Journal who defined virtualization as "using software to make one computer act like many." Grid, therefore, is reverse-virtualization: using software to make many computers act like one.
That's fine when you are just dealing with compute-intensive tasks. Transactional systems are a different story. They are made of multiple tiers: presentation, messaging, business logic and data. You cannot parallelize and distribute transaction processing unless you fundamentally solve this issue and collapse the tiers into one unit -- virtualize the tiers, if you will. Otherwise, the resulting system is broken up both vertically (into tiers) as well as horizontally (across machines) -- and you end up with a spaghetti-architecture that doesn't scale and is too slow. And that is the crux of what GigaSpaces solves and how it grid-enables transactional applications.
Nati, Owen and Guy have all posted great detailed explanations on how this works.
