Jump to content

GPU-Based Acceleration For PostgreSQL

Source
 

A developer has modified the PostgreSQL database software so that it can exploit GPGPU computing for faster performance. 

Kobei KaiGai, a software architect at NEC, has been working on GPGPU acceleration for PostgreSQL going back to 2011. Originally he was using a CUDA-based approach but he's shifted now to using the vendor-neutral OpenCL. His implementation supports many PostgreSQL functions and data types. The implementation can also perform full-table scans, hash joins, and pre-process of aggregation all with GPUs. 

This project to yield GPU acceleration with PostgreSQL is called PG-Storm. Going in the future there is a desire to expose more logics and functions on the GPU implementation. 

PG-Strom is open-source and is hosted on GitHub. Those wishing to learn more about this GPGPU support for PostgreSQL can also see the PG-Strom slides embedded below.

 

gpgpu-accelerates-postgresql-english-17-gpgpu-accelerates-postgresql-english-19-

 

Thought this might be news worthy as lots of us deal with database servers on the daily. I personally like PostgreSQL for its strong MySQL similarities.

Link to comment
https://linustechtips.com/topic/276145-gpu-based-acceleration-for-postgresql/
Share on other sites

Link to post
Share on other sites

Damn it! If we had this 3 years ago I could have talked my previous boss into upgrading my PC to a workstation with gpus for performance purposes. I need to find another job in BI and get me a system with a couple of 980s for business purposes of course

-------

Current Rig

-------

Link to post
Share on other sites

this screams for hsa! gpu accelerate what can be and fall back to the cpu, or there might be workloads where can work together sans pcie latency.

Link to post
Share on other sites

This is awesome. A server with a four processors, SSD storage, and 10Gb network connections is already an amazing database server. Adding a GPU or two would be awesome.

 

HSA can definitely help here, but this seems like an application where a dedicated GPU would be much better -- huge databases will require huge amounts of GPU power to accelerate effectively. Adding a discrete GPU is a much more effective solution than changing CPU setups. That being said, adding HSA CPUs will make smaller databases perform better as well.

 

Off topic, I'm a MySQL guy, no particular reason :)

I do not feel obliged to believe that the same God who has endowed us with sense, reason and intellect has intended us to forgo their use, and by some other means to give us knowledge which we can attain by them. - Galileo Galilei
Build Logs: Tophat (in progress), DNAF | Useful Links: How To: Choosing Your Storage Devices and Configuration, Case Study: RAID Tolerance to Failure, Reducing Single Points of Failure in Redundant Storage , Why Choose an SSD?, ZFS From A to Z (Eric1024), Advanced RAID: Survival Rates, Flashing LSI RAID Cards (alpenwasser), SAN and Storage Networking

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×