Browse Prior Art Database

METHOD AND MEANS OF ENSURING DATA INTEGRITY UPON FAILURE OF NVRAM CACHE

IP.com Disclosure Number: IPCOM000013544D
Original Publication Date: 2000-Mar-01
Included in the Prior Art Database: 2003-Jun-18
Document File: 1 page(s) / 36K

Publishing Venue

IBM

Abstract

RAID adapters can be configured to run in write-back mode to increase performance. This means that although the adapter tells the host that it is done writing the data to the hard files, it isn't really. The data still remains in the adapter's cache and will be written to the hard files later. If the cache, however, is volatile and the system loses power before the data is flushed to the drives, data loss occurs. To address this problem, the ServRAID II adapter and later uses an NVRAM cache daughter card, that plugs into the host adapter. Data in cache, but not yet flushed to the physical drives, is mirrored in the NVRAM cache. If a system powers down with dirty data in cache, the data will be reconstructed from the NVRAM cache upon power-up. Additionally, the card is made removable such that if the host adapter dies, the NVRAM card can be removed and placed on the new host adapter, and the data reconstructed as before without any data loss. The problem which presents itself, is that a user could take an NVRAM cache daughter card, that he has previously used on another system that was powered down abnormally, and place it in a new system. In this case, the data that's in the daughter card has nothing to do with the new system. If we reconstruct it, we will corrupt the hard file data on the new system. So...sometimes we want to reconstruct data...sometimes we don't. This disclosure discusses the method we use to decide. The premise for the method is that the NVRAM cache daughter card should be logically tied to the drives of the adapter it is connected to and not to the adapter itself. We thus use a piece of the NVRAM cache to store the configuration signature that is written on the hard drives. When we come up, we check the configuration signature of the NVRAM against that on the drives. If they match, we flush any NVRAM data. If they don't match, we clear out the NVRAM cache, write the drive configuration signature to it, and start using it like normal. At the next boot, the configuration signature will match and data will be reconstructed. 1

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 70% of the total text.

Page 1 of 1

  METHOD AND MEANS OF ENSURING DATA INTEGRITY UPON FAILURE OF NVRAM CACHE

RAID adapters can be configured to run in write-back mode to increase performance. This means that although the adapter tells the host that it is done writing the data to the hard files, it isn't really. The data still remains in the adapter's cache and will be written to the hard files later. If the cache, however, is volatile and the system loses power before the data is flushed to the drives, data loss occurs. To address this problem, the ServRAID II adapter and later uses an NVRAM cache daughter card, that plugs into the host adapter. Data in cache, but not yet flushed to the physical drives, is mirrored in the NVRAM cache. If a system powers down with dirty data in cache, the data will be reconstructed from the NVRAM cache upon power-up. Additionally, the card is made removable such that if the host adapter dies, the NVRAM card can be removed and placed on the new host adapter, and the data reconstructed as before without any data loss.

The problem which presents itself, is that a user could take an NVRAM cache daughter card, that he has previously used on another system that was powered down abnormally, and place it in a new system. In this case, the data that's in the daughter card has nothing to do with the new system. If we reconstruct it, we will corrupt the hard file data on the new system.

So...sometimes we want to reconstruct data...sometimes we don't. This disclosure discusses t...