Product |
Version/HW |
Cores per Node |
UoP per Node |
Notes |
Teradata | EDW 6700H |
16 |
32 |
Uses hyper-threads. |
Greenplum | DCA UAP Edition |
16 |
8 |
Recommends 1 Segment instance for each 2 cores. |
Exadata | X3 |
12 |
12-24 |
Maybe only 12… cannot find if they use hyper-threads. |
Netezza | Striper |
16 |
16 |
May use hyper-threads but limited by 16 FPGAs. |
HANA | Any Xeon E7-4800 |
40 |
80 |
Uses hyper-threads. |
A unit of parallelism (UoP) is defined as the maximum number of instructions that can execute in parallel on a single node for a single query. Since any program can start threads that wait I do not count these as UoP. On some CPUs vendors such as Intel allow two threads to execute instructions in-parallel in a core. This is called hyper-threading and, if implemented, it allows for two UoP on a single core.
Please feel free to comment if there is an update you would like to see included.
Update Log
April 24, 2013: Original Post… I’ll try to keep this updated over time.
Hi Rob
Sorry…I did try, but I can’t find a definitive reference on the Hyperthreading for Exadata either, but I can say that it uses Xeon E5-2690 (X3-2) or E7-8870 (X3-8) CPUs, which do offer HT and that there are several references (google search) from respected people/organisations indicating they use it on their X3 systems. For what it’s worth, one of my clients uses a HP DL980 G7 with E7-2830 Xeons in it and has HT enabled on that so I can’t see why they wouldn’t enable it, if it was appropriate for the type of workload.
Surprised DB2 LUW DPF (BLU or not) isn’t included here. Is there a criteria for the list it doesn’t meet?
It was not excluded for any good reason, Nathan… I just ran out of time. Using the criteria I suggest how would you evaluate it?
Great info! I’m wondering if I can help put Actian ParAccel and AWS Redshift on this list. I’ve used these in the past, and have access to the technical docs, as well as implemented clusters in physical and virtual worlds. Since they use commodity hardware, we can rely on their reference architectures. How can we have an apples-to-apples comparison? Use TPC-H or TPC-DS benchmark platforms?
Would love your thoughts on blockchain!