AMD EPYC vs. Intel Xeon Cascadelake With Facebook's RocksDB Database

[phoronix]

Phoronix: AMD EPYC vs. Intel Xeon Cascadelake With Facebook's RocksDB Database

Following the benchmarks earlier this month looking at PostgreSQL 12.0 on AMD EPYC Rome versus Intel Xeon Cascade Lake there was interest from Phoronix readers in wondering how well Rome is doing for other modern enterprise database workloads. One of those workloads that was recently added to the Phoronix Test Suite / OpenBenchmarking.org is Facebook's RocksDB, the company's embedded database that is forked from Google LevelDB. With RocksDB being designed to exploit many CPU cores and modern SSD storage, here are some benchmarks looking at how the Xeon Platinum 8280 stacks up against various new AMD EPYC 7002 series processors.

http://www.phoronix.com/vr.php?view=28378

[chithanh]

we have yet to test any PCIe 4.0 enterprise SSDs to measure that uplift.

Probably you will also have to test with a dataset that doesn't fit into RAM then, in order to see any difference...

[edwaleni]

Just looking at the cost for performance metrics, you can see why Rome is going to be a huge mover in the datacenters for the next 18 months.

[Drago]

It is all good and all testing these server processors, but I wonder is there any professional user reading Phoronix and chooses platform for his server farm based on these articles?

[andrei_me]

It is all good and all testing these server processors, but I wonder is there any professional user reading Phoronix and chooses platform for his server farm based on these articles?

Not related to hardware as it is too expensive where I live, but in my company I made a report using phoronix articles as evidence to which distro we should choose to use on everyone's laptop to achieve good perf and low maintenance burden (Clear linux), and for those hardware without UEFI or couldn't have Clear installed, we had evidence that there isn't a significant difference between Debian/Fedora/CentOS, so we choose Debian as it is the distro we are most used to

[Drago]

Not related to hardware as it is too expensive where I live, but in my company I made a report using phoronix articles as evidence to which distro we should choose to use on everyone's laptop to achieve good perf and low maintenance burden (Clear linux), and for those hardware without UEFI or couldn't have Clear installed, we had evidence that there isn't a significant difference between Debian/Fedora/CentOS, so we choose Debian as it is the distro we are most used to

Sure, choosing OS for personal laptops is a thing, choosing server platform is entirely different thing.

[edwaleni]

It is all good and all testing these server processors, but I wonder is there any professional user reading Phoronix and chooses platform for his server farm based on these articles?

I can assure you that several enterprise server architecture teams read Phoronix. There are also many enterprise server evaluation teams that run PTS in their labs to validate performance profiles. I have seen Phoronix articles being passed around corporate email systems and sometimes followed by fierce discussions just like you see in the forums.

Just because you don't see posts from these companies employees or contractors is more indicative of the internal data policies of their employers to protect their competitive positions in the marketplace.

Most server farm decisions are made based on several data points, not just PTS results or an article in Phoronix. But it is a data point that is used frequently.

[Drago]

I can assure you that several enterprise server architecture teams read Phoronix. There are also many enterprise server evaluation teams that run PTS in their labs to validate performance profiles. I have seen Phoronix articles being passed around corporate email systems and sometimes followed by fierce discussions just like you see in the forums.

Just because you don't see posts from these companies employees or contractors is more indicative of the internal data policies of their employers to protect their competitive positions in the marketplace.

Most server farm decisions are made based on several data points, not just PTS results or an article in Phoronix. But it is a data point that is used frequently.

If so, I am glad to hear that

[Spacefish]

Well, there are companies which still prefer intel cpus, because they are "used to them" or whatever the reason..
If you look at the raw data, AMD is just blowing intel out of the water.

[edwaleni]

Well, there are companies which still prefer intel cpus, because they are "used to them" or whatever the reason..
If you look at the raw data, AMD is just blowing intel out of the water.

I think AMD Rome is going to put a big dent in "Intel Only" in the datacenter. There is pent up demand to reduce the number of sockets in the 2U form factor and Rome really nails it. A lot of companies are putting Rome based designs on their RFP's.

The only weakness that I can see is that Rome will be less impressive in white box.