Google's dominance over how we search and consume information may be meeting its first real challenge. This article at GigaOM (http://bit.ly/GeZ7C) explains why. And no it's not Wolfram Alpha or something..
One of the other things that come out of this is; Twitter does not have one coffee shop. It has hundreds and thousands of them, with new ones opening every minute. So how do I know which coffee shop to go? And more importantly, How do I know what conversation is taking place at which coffee shop?
It would be interesting to see, what kind of features Twitter's rumored search engine comes up with. If this is just a replica of Google Search, in the sense it just shoves keywords into a BigTable data structure, then it would be no use at all. Go to a real coffee shop; not the sluty Cafe Coffee Day ones but a real one like India Coffee House, take a pen and paper and then try to create an index out of all the keywords you overhear or recognize. I bet you will give up. But you can of course gauge trends, and that's what Twitter Trends is already doing with Twitter.
I am not sure why this has to be a X VS Y, but an interesting presentation nevertheless.
I believe this choice should be made on the basis of sound Data Modeling rather than anything else. The Relational Model is well known, but very few people know that the Big Table and HBase storage strategy originates from the Entity Value Attribute model.
I find the EAV model ideal for storing all kinds of metadata. Consider metadata for a photograph. While there exists known formats like EXIF and IPTC which embed the metadata inside the file itself, businesses having vast amounts of Digital Assets often add extra metadata, and store it separately. The same thing applies to other domains as well. There could be tons of metadata associated with products like a books, tables, chairs, lamp shades, you name it. Merchants usually markup individual products using keywords, or using some form of key, value pairs.
This metadata could then be indexed or in HBase's case are sorted and stored near to each other, so that extraction is fast.
About Me I enjoy solving user interface and usability problems. I also have proven expertise in areas like middleware, encryption, web services and systems engineering.
About Me