D3/Pick performance testing
Toby Kraft
--You know - that is a fairly meaningless performance benchmark for a database system ... It basically tests a particular type of string building and nothing else. As has been noted here in the past, different systems have different levels of optimisation for this type of string building.FWIW, on my i5 notebook (6th Gen), UV took 0.252 seconds; QM took 0.032 seconds (both local installs so no network latency).That doesn't match the commonly held view that UV is faster than QM. However, I would still want to see a more comprehensive test of "database functionality" before making any performance claims.Cheers,Brian
On Friday, 17 April 2020 05:52:27 UTC+12, Marcus Rhodes wrote:I agree. My OpenQM 2.6.6 running under Ubuntu Server 18.04 on a 7-year-old Asus G750jw did that in, on average, about 2.62 seconds. JBase 4.1 in an XP virtual machine on that same system averages around 1.3 seconds, double the speed, suggesting that, if running closer to the hardware, it might be 2.5+ times as fast as QM.
On Thursday, April 16, 2020 at 11:13:46 AM UTC-4, John Doe wrote:David Knight wrote: "Followers can always follow at a lower price."
He is right, D3 is the most expensive cart you can buy (the horse comes separately) and everybody follows in price.
Probably you don't want a cart but a car that is at least five times faster and you don't have to deal with manure (GFEs) nor with the high cost of barley (maintenance fee)
If you decide to get a practical car you should take a closer look at QM and Universe.
Note that while Universe is on average about 50 % faster than QM, apparently is about 4 times more expensive.
P.S. jBASE is not for the faint hearted.
Some may say that Revelation could be a contender. Think not.
Not long ago I've seen on You Tube a demo of Revelation running the following snippet:
BUFF = ''
FOR I = 1 TO 15000
BUFF<I> = I
NEXT
It took 5.60 seconds to run on a desktop much younger than my 8 years old laptop.
On my old laptop (Intel B960) the same code runs for:
UV 1.09 sec.
QM 2.70 sec.
… or maybe that shiny desktop was just a 386 in new clothes? Who knows?
jlab...@paradigm-systems.us
The problem with the Prime Number Sieve is that many system already have a built in function to generate primes. Granted a long hand method of generating the first 1024 prime number may have some value testing the systems math functions. In order to create a benchmark standard we would need to define each area of the environment and assign it a specific name or value and define the method for testing that method.
For example a Dynamic Database Write test may be define as creating a file of a specific modulo and separation and then create 1,000,000 records of a specific size using randomized data and logging the start and end time down to the millisecond. This could be the first test and the generated database could then be used to test selection and read performance among others.
Regards,
Jay LaBonte
--
You received this message because you are subscribed to
the "Pick and MultiValue Databases" group.
To post, email to: mvd...@googlegroups.com
To unsubscribe, email to: mvdbms+un...@googlegroups.com
For more options, visit http://groups.google.com/group/mvdbms
---
You received this message because you are subscribed to the Google Groups "Pick and MultiValue Databases" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mvdbms+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mvdbms/7662cc28-7bcc-466e-ac95-8b87afe06641%40googlegroups.com.
Toby Kraft
To unsubscribe, email to: mvd...@googlegroups.com
For more options, visit http://groups.google.com/group/mvdbms
---
You received this message because you are subscribed to the Google Groups "Pick and MultiValue Databases" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mvd...@googlegroups.com.
