I briefed with KickFire recently and talked my way through the benchmark shown in the picture below. It’s worth commenting on. KickFire, in case you didn’t know, is data warehouse technology delivered as an appliance. It uses the MySQL database as it’s underlying database technology, but the truth is that it does a good deal of compression so it tends to consume less storage than a vanilla implementation of a traditional relational database.

If you’re familiar with fast database technology you can bracket KickFire with Vertica and GreenPlum, both of which are optimized for data warehouse speed, but KickFire is a little different to those for two reasons:

• It is an appliance.
• It isn’t aimed at the VLDB (Very Large Data Base) market. It’s aimed at the QLDB (Quite Large Data Base) market.

The sweet spot for this technology is in the up-to-6TB market, where it kicks sand in the face of the traditional database giving query responses that are at least on order of magnitude faster (i.e. 10x) but range up to and above two orders of magnitude faster (i.e. 100x).

Where does the performance come from?
There are are two distinct areas of acceleration over the normal database:

• KickFire is a column store implementation (like Sybase IQ, Vertica, Paraccel, etc.)
• KickFire has a purpose specific chip with SQL burned into the silicon.

If you ask KickFire what each of these technical features give you, then they’ll attribute 10x to the chip and a further 10x to the column store database approach.

The Benchmark
I’m no great believer in benchmarks, but query based benchmarks have greater credibility. What we see here is a comparison with data warehouse benchmarks, using Oracle on HP hardware and Microsoft SQL Server on IBM hardware. Ignoring KickFire for the moment, the Microsoft solution has the preferable TCO, but that’s down to product cost which is often not the deciding variable. What’s interesting about KickFire, apart from the fact that for 300GB databases it clearly comes in at a fraction of the price for roughly the same response times, is that it consumes so little space in the data center (less than one tenth of the space) and it consumes so little electricity (one twentieth or better).

This is the first time that I’ve noticed data center stats like that appear on a benchmark, but it makes obvious sense – especially when some data centers are reporting electricity costs as a high proportion of amortized hardware costs. I don’t happen to know whether these costs are factored in to the 3 Yr TCO figure. I doubt it because data center space and electricity costs are not standard across the  globe or across the US.

The Bottom Line
The bottom line is that KickFire offers a relatively inexpensive start point for the medium-size company that wants to get started in data warehouse of the larger company that might be able to migrate to this technology. The reduced cost of ownership is compelling.