Browse Prior Art Database

Prolonged life and performance of SSD based storage array

IP.com Disclosure Number: IPCOM000220445D
Publication Date: 2012-Jul-30
Document File: 2 page(s) / 47K

Publishing Venue

The IP.com Prior Art Database

Abstract

Using SSDs in high performace storage arrays is subject to wear levelling issues that influence End Of Life [EOL] determination. Premature replacement of SSDs can be costly either to the customer or to the service provider directly. Using readily available metrics it is possible to utilise the full life cycle of SSDs and prevent early replacements of these costly parts. By rotating the function of each SSD within an array, it is possible to achieve the full utilisation using existing supported function.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 2

Prolonged life and performance of SSD based storage array

With currently known technology, SSD storage devices have a well known wear levelling limitation. The device has a finite number of writes that can be supported by that device before wearing out. Furthermore, the device is likely to have a maximum age at which point the device needs replacing. Furthermore, devices are likely to provide either a warning indication for approaching EOL or alternatively some sort of endurance management system that will by capable of limiting the number of write operations in order to achieve expected EOL.

    SSD technology at this point is an expensive technology and it is very important to make the most of the technology from a performance perspective and a cost perspective.

    Use of SSDs in a RAID array is not uncommon. Furthermore, due to SSDs having a very superior IOps performance, there is very likely to be some sort of IO hot spot feature or otherwise that attempts to use the SSDs in order to maximise the benefit to the customer. A consequence of this is a natural imbalance of IO across the RAID array.

Consider:


- A RAID array with some hot spots that cause repeated writes to the same physical SSD. The consequence of this is that there will be a subset of devices in the array that will have had more writes to them, than the others.


- Therefore the SSD devices will need replacing ahead of the others which may well be in advance of the service offering and either provide increased cost to the vendor, or will cost the user due to early replacement.


- Although the other devices in the array may well have been written to less, there's an age restriction that forces device replacement to. Therefore it is not possible to get back the benefit of an early replace with a later replace on other devices in the array.

- It is also possible that there has been an array component failure too
The device suppliers have built-in algorithms that can for the life of the device

to be the normal target age. This is done by limiting the write IOps capability of the device such that the number of writes can not exceed an average rate. Drives have a 5 year life (typical), so will ensure that on average the number of writes has not exceeded its entitlement. This has two disadvantages. Firstly there may well be cases where the array IOps can be limited by one device and thus under-achieve the customer's expectations. Secondly, there is no adjustment for current trends, just a general limitation. So in the case where the customer's workloads were initially very light for a year and then rapidly increased, the array would cope perfectly well until the over-subscription catches up and then IOps of the array would then drop to the nominal workload, causing a sudden surprise for customer workloads.

    All of these devices can be queried by controller code. In order to be able to provide the customer with information that makes sure that sudden surprises are avoided, requires...