Browse Prior Art Database

System and method of array scrub adaptive adjustment

IP.com Disclosure Number: IPCOM000244567D
Publication Date: 2015-Dec-22
Document File: 8 page(s) / 190K

Publishing Venue

The IP.com Prior Art Database

Abstract

This disclosure shows an array scrub adaptive adjustment system. This system has two part: “Predictive adjustment” and “Real-time adjustment “. 1) Predictive adjustment is based on historical IO statistic regularity and according to some regularity pattern, scrub manager can adjust the scrub rate ahead of heavy IO period. 2) Real-time adjustment is based on current IO strength, if heavy IO comes unexpected, scrub manager will change the array scrub from full-scan mode into priority-scan mode.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 40% of the total text.

Page 01 of 8

System and method of array scrub adaptive adjustment

RAID technology has array scrub, which is a background process which verifies that each data and parity block can be read at least once periodically. This function can reduce the possibility of IO timeout and data loss due to disk damage, and therefore maximize the protection of data.

But array data scrubbing could consume too much IOand XOR bandwidth in large configurations along with the disk capacity growth. If there are some heavy host I/O, array scrubbing impacts host I/O. In another hand, if some area of hard disk just have hot IO on it, then we don't need to scrub it in a certain period of time again because host IO has already verified the block can be read or not.

In current solution, array scrubbing didn't do any adaptive adjustment. Scrubbing is performed in short bursts once every defined period performed in accordance with the address sequence without disk priority or block priority, no matter host I/O is heavy or not.

So the drawbacks are: 1) In current solution, operator will set a threshold of used bandwidth and a defined period for array scrubbing. This is a very simple way to constraint array scrub. If the host I/O is heavy, array scrub is still on-going and work as normal. This will impact host I/O. When host I/O is light, array scrub won't use additional resource, despite there are enough resource can be used by array scrub.

2) In current solution, array scrub performed just in accordance with the address sequence without any priority selected. So in some situation, array scrub may performed in the area of disks that just has very hot IO on it and this kind of scrub is totallyunnecessary and delayed the scrub of the other more potential broken area of some disks which may have not any IO on it for some time.

3) Along with hard disk capacity grow bigger and bigger, it is hard to scan all the blocks in a defined period without impact host IO. For example, the scrub rate for an array built with 8*6TB disks will be increased several times than the array built with 8*300G disks which will have more chance to impact host IO.

This disclosure shows an array scrub adaptive adjustment system. This system has two part: "Predictive adjustment" and "Real-time adjustment ".


1) Predictive adjustment is based on historical IO statistic regularity and according to some regularity pattern, scrub manager can adjust the scrub rate ahead of heavy IO period.

1


Page 02 of 8

Applications always have regularity. In the storage, arrays I/O statistics data can be gathered and saved in repository. With the historical data, arrays workload regularities can be concluded by intelligent analysis. According to the workload regularity, array scrubbing can be performed more during light workload and perform less during heavy workload. Frequency of array scrub and bandwidth of array scrub can be adjusted according to workload regularity.

The advantage of predictive adjustment are:


With the met...