This test validates the basic functionality of the alba block cache and should prove its performance benefit.
Cover area
// What area of the product is affected?
Alba
// What type of functionality was used
An alba accelerated vpool using block cache on read.
// Which type of issue is this
Enterprise functionality
performance
User story
User has a HDD Alba backend accelerated by a SSD Alba cache backend used to cache subsets(blocks of 256KB in size) of a fragment(typically 1 or 4MB) on read for faster access upon reread.
In case necessary distinction can be made between different devices by creating different cache backends for read and write.
Problem that arose
Inefficient use of the cache capacity leading to lower than expected performance.
Additional information
Setup
// Be as specific as possible
Multiple nodes(2SSD & 3HDD) part of each backend
-- 1 HDD backend: preset No Compression, No Encryption, policy 6/2/6/3
-- 1 SSD cache backend: preset No Compression, No Encryption, policy 2/0/2/1
vPool with accelerated backend
-- SCOsize = 16MB / Limited global write buffer = 2GB /
-- Use SSD cache backend for 'BlockCache on Read', 'No fragment Cache'
Volumes
-- Create volumes with size > 3 times global write buffer(eg: 10GB)
-- Write complete volume once
--- Verify the block cache namespace is still empty(show-namespace)
-- Read the complete volume and capture results(r1 = iops/bandwidth)
--- Verify the block cache namespace exists and got populated, keep track of the amount of objects.
-- Reread the complete volume and capture results(r2 = iops,bandwidth)
--- Verify the block cache namespace has the same number of objects cached.
--- Verify the results r2 of second tests are better than r1.
--- As an extra the alba proxy statistics could be tracked and compared.
Test description
This test validates the basic functionality of the alba block cache and should prove its performance benefit.
Cover area
// What area of the product is affected?
// What type of functionality was used
// Which type of issue is this
User story
User has a HDD Alba backend accelerated by a SSD Alba cache backend used to cache subsets(blocks of 256KB in size) of a fragment(typically 1 or 4MB) on read for faster access upon reread. In case necessary distinction can be made between different devices by creating different cache backends for read and write.
Problem that arose
Inefficient use of the cache capacity leading to lower than expected performance.
Additional information
Setup
// Be as specific as possible