-
|
Hello, hammerdb> I couldnt find anything in the logfile, mysql processlist shows nothing useful, |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
|
I decided to let it run for as long as possible, |
Beta Was this translation helpful? Give feedback.
-
|
So the key bottleneck here is the network round trip between the client and the server whilst doing the load, so yes if this network round trip is not particularly fast then the load can and will take a long time. By default, the HammerDB schema creation and load has been designed to be as easy as possible, so you start the schema build running, and it will create and load the data dynamically without intermediate files. It does this by calling standard 'insert' calls. For the larger tables mySQL uses multi-row inserts, for the item table because it is smaller and usually finishes much faster we insert a row at a time - so as you see the item load taking longer, then we know the issue is the network round trip as it is not using multi-row. Additionally, MySQL networking is not as performant as other databases, so the impact is amplified with MySQL. Note that even loading locally, there is always the overhead of the database maintaining consistency by logging all of your transactions in case of recovery, so most of the load overhead is on the database side (for all databases). For these reasons, HammerDB supports a datagen option - this allows you to create the data locally as flat files, this will run much faster than even a local database load as it is only dependent on the local CPU and disk performance. You can then ship these files to the remote server and do a bulk load that is not logged. The link above explains how to do this and complete the schema build for each database. This is more manual overhead initially, but much quicker to do the actual bulk load itself. So if you are doing repeated builds and/or working remotely this is the approach to take and the reason this datagen feature was added. Also be aware that if testing remotely, then the network round trip will impact the test performance as well. With MySQL for reasons noted above you will need a larger number of Virtual Users than testing locally to reach equivalent performance each on a LAN. The highest performance for MySQL will come through connecting through a local socket connection. |
Beta Was this translation helpful? Give feedback.
-
|
thank you for the prompt reply |
Beta Was this translation helpful? Give feedback.
So the key bottleneck here is the network round trip between the client and the server whilst doing the load, so yes if this network round trip is not particularly fast then the load can and will take a long time.
By default, the HammerDB schema creation and load has been designed to be as easy as possible, so you start the schema build running, and it will create and load the data dynamically without intermediate files. It does this by calling standard 'insert' calls. For the larger tables mySQL uses multi-row inserts, for the item table because it is smaller and usually finishes much faster we insert a row at a time - so as you see the item load taking longer, then we know the issue is …