2008 aio application bandwidth bl460c blade bladesystem c3000 cleaning connection Disk drive edition enclosure error home hosting hp iscsi lamp letogolais lto memory microsoft mpio multipath network oil oracle performance proliant recovery recycler recycling SB600c sb920c server solvent speed state storage storageworks system tape togo togolais ultrium virtual web windows

RAID Memory - using redundancy to prevent memory errors

Back in 2001 we had a representative of Compaq visit our workplace and give us extended education on available servers and their clustering options and high availability systems. He went on and on about redundant power supplies, fans and RAID drive configurations. Then he also mentioned hotpluggable PCI bus and ability to decommission CPU while system is running. Of course I had to ask about redundant memory. He stopped, smiled, thought for a second and then he told me he is not allowed to talk about it but I should stay assured that I'm not the only one asking the question. Then HP came and bought Compaq.

First server that was released under HP was the DL380 G2 (second generation). Still running on Pentium III cpus capable of 1.466GHz. But one amazing feature was included. The DL380 was capable of accepting either 6GB of ECC memory or 4GB of ECC memory with 2GB hot standby. With ECC, the bit error can be detected and repaired but when in 4+2 configuration on very first bit error an orange LED lights up next to memory chips and both banks of the faulty memory are copied onto backup two sticks. Data is not lost if not too many errors appear in memory.

Scroll down for additional info labeled Next generation proactive memory RAID protection

More info: Next generation proactive memory RAID protection

permanent link to article http://freecash.hogger.net/raid_memory_using_redundancy_to_prevent_memory_errors

It is clear that if a major destruction of data occurs, this system cannot prevent a loss of data. But for most practical applications, the system was already phenomenal. Then the DL580 came out and it also supported standard hot standby memory but it also had an added feature of RAID 3 configuration. In RAID 3 configuration, the data is stored in main banks and checksum is included in backup parity bank. If a bank fails, the data is then reconstructed as parity minus sum of all remaining banks. This level of protection will provide data consistency even if the memory chip were to fall out of the socket somehow. Naturally this opens a possibility to replace memory while the system is running. That was then, now all carrier class servers from Compaq/HP of the ProLiant line include the redundant memory chipset, which includes latest 1U rack mountable DL360 and some of the blade machines as well.