On 2/19/2013 10:39 AM, chamara wrote:
Hi Thanks Shawn for the Input, Yes i am using SolrCloud to replicate the index to another server running with the same spec with 32cores and 72GB RAM on each machine. I have to test the performance of RAID 10? Have you ever done a deployment with RAID 10? The indexing will be NRT as far as i can see , so it is going to be lots of writes to the DISK.
I use a six-drive RAID10 on each of my four Solr servers - with 1TB drives it gives me a usable capacity of 3TB. RAID10 has very good performance on both reads and writes, and neither creates any significant penalty for the other. The drawback is that it comes at a significant penalty in capacity - if you have 8TB of raw disk space, you get only 4TB usable.
RAID5/6 has excellent read performance in mostly-read environments, but *all* performance (including reads) suffers when doing a lot of writes, due to the need to calculate parity and write it to all disks. A RAID controller with battery-backed (or NVRAM) cache memory is very effective at reducing the cost of this penalty when the amount of data written at once is smaller than the cache, but if you write a LOT of data, you can't buy a cache big enough.
With very careful design, you might be able to make a RAID6 solution work for you, but RAID10 works well for any situation.
Thanks, Shawn
