Portworx has best practices for both global container level optimization, as well as volume granular optimization.
Global performance tuning
As of Portworx version 1.3, it is recommended to use a journal device to absorb Portworx metadata writes. Journal writes are small with frequent syncs and therefore only SSD/NVME should be configured as a journal device.
In 1.x, the journal device should be 2GB, and in 2.x it should be 3GB. Using a larger device will not help, since Portworx will only use these amounts of storage. The journal device can be specified via the
-j option during installation, documented here.
-j autooption. This allows Portworx to create its own journal partition on the best drive.
If you are upgrading to 1.3 and want to add a journal device to an existing node, follow these instructions.
Volume granular performance tuning
By default, Portworx will try to auto tune the IO profile setting for a given volume by learning from the access patterns. However, this algorithm can be overridden and a specific profile can be chosen.
The IO profile can be selected while creating the volume via the
io_profile flag. For example:
pxctl volume create --size=10 --repl=3 --io_profile=sequential demovolume
docker volume create -d pxd io_priority=high,size=10G,repl=3,io_profile=random,name=demovolume
It is highly recommended letting Portworx decide the correct IO profile tuning. If you do however override the setting, you should understand the operation of each profile setting.
The sequential profile
This optimizes the read ahead algorithm for sequential access. Use
The random profile
This records the IO pattern of recent access and optimizes the read ahead and data layout algorithms for short term random patterns. Use
This is useful for content management systems, like WordPress. This option applies to a Portworx shared (global namespace) volume. It implements an attribute cache and supports async writes. This increases the Portworx memory footprint by 100MB. Use
The db profile
This implements a write-back flush coalescing algorithm. This algorithm attempts to coalesce multiple
syncs that occur within a 50ms window into a single sync. Coalesced syncs are acknowledged only after copying to all replicas. In order to do this, the algorithm requires a minimum replication (HA factor) of 3. This mode assumes all replicas do not fail (kernel panic or power loss) simultaneously in a 50 ms window. Use
The sync_shared profile
Use this profile to set up the mount settings of your
sharedV4 volumes. This profile sets the following options:
This option synchronizes all the write operations with your backend storage. If you specify this option, the server and the clients will be able to access modified data immediately.
This option disables the caching for the file system and metadata (file and directory attributes) on the client-side.