Jerry Held on cloud data warehousing and how business intelligence will be transformed by it
Vertica Chairman Jerry Held has a pair of blog posts on analytics and data warehousing in the cloud. The first lays out a number of potential benefits and consequences of cloud data warehousing, under the heading of “Transforming BI”:
- New BI technology adoption will accelerate. (Because evaluation will be easier)
- Organizations will conduct more short-term ad-hoc analysis. (Because of quick set-up and lack of capital costs.)
- Lines of business will have the flexibility to fund more data mart projects. (Because of lack of capital costs.)
- Data warehousing will increase within medium-size businesses. (I guess because of lack of capital costs.)
- The analytic SaaS market will develop faster. (Because a suitable platform now exists, and because of lack of capital costs.)
On the whole Jerry makes a compelling argument, and I recommend the post highly.
The second post in the series lays out what kind of a DBMS architecture you need to get these benefits. Basically, the criteria boil down to “Runs efficiently and is reliable,” and — surprise! — the specific features needed are exactly those found in Vertica.
Edit: But I do take issue with Jerry’s bright-line distinction between cloud computing and SaaS.
Comments
5 Responses to “Jerry Held on cloud data warehousing and how business intelligence will be transformed by it”
Leave a Reply
Pure balderdash: High speed analytical databases require tight integration to the hardware platforms that they run on. All of the leading hardware vendors offer a myriad of configurations to their platforms, Sun is offering 8GB DIMMS recently for instance. The software must be optimized to execute on the platform provided. How will that ever happen in the cloud? It is a fantasy. The whole concept is that you don’t know what platform the application is running on. Non-starter
Bill,
That depends a lot on the DBMS design. To a first approximation, what you said is true for row-based systems but false for column-based ones. (I’m not sure about array-based ones.)
CAM
So you are saying that users of the column based approach don’t need to care about the disk to memory ratio on the machine they are running their query on?
They don’t need to care as much as users of other kinds of DBMS do. E.g., reference architectures aren’t as important as they are for most row-based systems.
CAM
[...] Vertica Chairman Jerry Held’s views on cloud data warehousing, and my reaction to same Share: These icons link to social bookmarking sites where readers can share and discover new web pages. [...]