June 29, 2009
Xtreme Data readies a different kind of FPGA-based data warehouse appliance
Xtreme Data called me to talk about its plans in the data warehouse appliance business, almost all details of which are currently embargoed. Still, a few points may be worth noting ahead of more precise information, namely:
- Xtreme Data’s basic idea is to take a custom board and build a data warehouse appliance around it.
- An Xtreme Data board looks a lot like a conventional two-socket board, but has only one four-core CPU. In addition, it sports some FPGAs (Field-Programmable Gate Arrays).
- In the Xtreme Data appliance, the FPGAs will be used for core SQL processing, after the data is ingested via conventional I/O. This is different from Netezza’s approach to FPGA-based data warehouse appliances, in which the FPGA sits in the place of a disk controller and touches the data first, before passing it off to a more or less conventional CPU.
- While preparing entry into the data warehouse appliance business, Xtreme Data has sold its board to 150 other outfits, many quite impressive. Buyers seem to be FPGA users who previously had to craft their own custom boards. According to Xtreme Data, major uses by these customers include:
- Military/intelligence/digital signal processing.
- Military/intelligence/cybersecurity (a newish area for Xtreme Data)
- Bioinformatics/high-throughput gene sequencing (a “handful” of customers)
- Medical imaging
- More or less pure university research of various sorts (around 50 customers)
- … but not database management.
- Xtreme Data’s website has a non-obvious URL. 🙂
So far as I can tell, Xtreme Data’s 1.0 product will — like most other 1.0 analytic database management products — be focused on price/performance, without little or no positive differentiation in the way of features.
Categories: Data warehouse appliances, Data warehousing, Netezza, Theory and architecture, XtremeData
Subscribe to our complete feed!
Comments
6 Responses to “Xtreme Data readies a different kind of FPGA-based data warehouse appliance”
Leave a Reply
[…] Patented SQL on a chip! Xtreme Data readies a different kind of FPGA-based data warehouse appliance […]
Can you provide any info on how this differs from Kickfire’s SQL Chip? (NDA permitting) Thanks!
Zman,
I’m sorry, but I think any more detail would be covered by the embargo.
Zman,
It is a little unclear as to what the capabilities the “Xtremedata” box will actually have, given the partial stealth nature of Xtremedata.
Kickfire is selling a product today and is performing extremely well for our customers. We feature a small form factor and we don’t cosume a lot of power. We can run queries in seconds that MySQL takes DAYS to process.
For example. We have one customer who bought our appliance. On their old MySQL servers, one of their most important queries ran for more than 4 DAYS. Before engaging in a POC with Kickfire, they tried out Infobright. They got pretty good results. Their 4 day query ran in four hours.
In the Kickfire POC, this query ran in four minutes. This very important query can now be run any time they want, it can be changed to look at patterns and it doesn’t lock their database for hours or days at a time.
Kickfire has show proven real performance numbers on benchmarks like TPC-H(tm) and in customer POCs. We’ve proven our solution works. We’ll just have to wait and see if Xtremedata can match the very high bar which we have set.
Xtremedata claims to run the TPC-H(tm) queries, but they have not submitted audited results, and they aren’t allowed to use the TPC-H(tm) name as far as I understand.
The architectural differences are
moribundmanifold. The Kickfire SQL chip is a true dataflow engine with patented in-memory compression and SQL execution technology. Just having an FPGA doesn’t mean you can run “SQL on a chip”.These are my own opinions and I am not speaking directly for Kickfire.
Hi Justin,
Please help me understand the benefit of a SQL on Chip engine as comared to ParAccel and Vertica, who can provide stellar performance on any data volumes with SW only? They can also package it as an appliance with very small number of nodes to very large number of nodes.
Thanks
The big difference between the MPP column stores and Kickfire is the price/performance ratio. If you look at the benchmarks you will see that Kickfire provides great performance in a small form factor for a lot less money. If benchmarks are good at only one thing, gauging price/performance is probably it. The SQL chip allows us to literally do more with less.
But beyond that, Kickfire decided to go in a different direction from the other guys. In order to get performance at very large data volumes, there are tradeoffs. MPP databases often don’t feature UNIQUE keys, FOREIGN KEY constraints and other indexes which can be very useful.
There are technical challenges to implementing such keys in a MPP system, since the constraints must be maintained across the entire cluster. When you have volumes of data in the tens of terabytes or higher, it makes sense to go with ETL instead.
Big companies with vast amounts of time and money will be happy to invest in an ETL process to accompany their new multi-TB warehouse. Smaller shops with less data are not so willing to spend effort and hard earned capital on such processes.
Small companies want fast results without spending a lot of money. They might only have a hundred gigabytes of data, but they want answers on it now, and they don’t want to spend a million dollars to get them.
There are a lot more small companies with tens or hundreds of gigabytes than there are large companies with tens or hundreds of terabytes.