October 3, 2010
Notes and links October 3 2010
Some notes, follow-up, and links before I head out to California:
- HP hired a software guy, Leo Apotheker, as CEO, and a software guy with a liking for high-end services, Ray Lane, as chairman. Now a Leo Apotheker conference call suggests HP will increase its emphasis on software, and maybe high-end services as well. No surprise. The article suggests, however, that HP at this point has no clear strategy along these lines. That’s no surprise either.
- And then there’s Sarah Lacy’s take, of which the interesting part reads “Separately, Andreessen has said that he thinks enterprise software is ripe for disruption and his firm is going to fund a new generation of Oracle-killers.”
- I added more on Ray Lane’s tenure at Oracle over on Software Memories.
- Netezza had a falling out with its original supplier of geospatial technology, Intelligent Integration Systems (IISi), and a lawsuit ensued over alleged copying. Now ISSi has upped the stakes, essentially alleging that Netezza’s new geospatial software doesn’t work, and that hence the CIA (evidently a Netezza user) is killing the wrong people via drone strikes. Netezza has wisely selected from its short list of acceptable responses, including versions of:
- “All our classified customers are happy, and if we told you anything more than that, that would kind of defeat the purpose of being classified, wouldn’t it?”
- “Copy, schmopy. A polygon is a polygon, and has been since Euclid.”
- “We don’t have no steenking bugs.”
- OCZ, whoever they are, are trying to offer solid-state drives with PCIe-like bandwidth, which makes sense in that most observers except Teradata think the SAS interface isn’t fast enough for solid-state.
- Speaking of Teradata, I’d been wondering somewhat as to why they just shut down Kickfire’s product line after acquiring its assets. Well, somebody who tested a Kickfire box told me that — great TPC-H results notwithstanding — it turned out not to be nearly as fast as one might think, on real-life data sets that didn’t fit entirely into RAM. Hard though such a thing may be to imagine, it turns out that Kickfire’s TPC-H results were yet less significant than I thought they were.
- I haven’t been looking at High Scalability nearly as much as I should, and that’s an understatement. It’s an outstanding blog.
- A couple of Google execs offered some predictions about the future of online advertising, which might be of interest to anybody selling analytic (or text analytic) technology to the online/digital media market.
- The BBC shows us what a single 133-character tweet plus its metadata look like in JSON. (All 1582 characters.)
- Huffington Post’s CEO made some comments about influencers which are additive to what I’ve been saying about influencers over on Strategic Messaging. (If you don’t read that — well, it’s my blog about marketing.)
- Speaking of my other blogs, I’m not bothering to put up a separate post like this over on Text Technologies, where thee posts I have put up recently tend to be (at least by my standards) relatively link-heavy anyway, but I have a couple more to share even so:
- Paul Carr’s 7 rules for TechCrunch/AOL employees are really funny.
- Some major search engine marketing experts are sounding defeatist about web spam.
Categories: GIS and geospatial, Google, HP and Neoview, Humor, Kickfire, Netezza, Solid-state memory, Teradata, Web analytics
Subscribe to our complete feed!
Comments
3 Responses to “Notes and links October 3 2010”
Leave a Reply
With Ray Lane joining HP, and HP moving harder into software, how long before Vertica is the new non-Stop SQL? A good technology buried in a machine and hidden from the market.
Minor correction: OCZ are trying to offer PCIe-like bandwidth for 3.5 inch form-factor drives.
That is different from PCIe SSD, available from same OCZ and other companies for many years.
The articles argues about some faster 20Gb/sec link interface which is apparently same as external PCIe interface standard. That 20Gb/sec link is than further compared to 6Gb/sec of SAS.
However, the article (and OCZ press release) doesn’t mention that the new link runs 5Gb/sec versus 6Gb/sec per SAS. So in terms of pure wire efficiency it is in fact in disadvantage.
It is not clear also why by the same token one cannot combine multiple SAS links and achieve even greater speed per link with same number of lanes and without requiring invention of yet another storage interface?
Another interesting detail is that SAS interface haven’t gone anywhere. Flash chips still talks to multiple SAS controllers and SAS controllers still talk to RAID chip and RAID chip still talks to PCIe multiplexer. Business as usual here. So the proposed arrangement just moved RAID controller chip to the drive case and added another multiplexer chip in the critical data path.
In my opinion: it is pointless to move chips back and forth between PCIe card to the drive case in the quest of optimal arrangement (or just novel arrangement?).
What the industry really needs is a commodity FusionIO-like solution where SAS/SATA is eliminated altogether. Where native PCIe SSD controller talks directly to flash chips on one side and PCIe on the other. Such solution is leaner and doesn’t require any new standards. Neither any wires, neither 3.5 inch form-factor cases complicating airflow. A lot better power-efficiency. Lower latency. Higher throughput. More software control over flash-management issues that enables better optimization such as lighter wear-leveling for read-mostly analytic applications.
Peace,
Camuel.
[…] Notes and links October 3 2010 (dbms2.com) […]