I need to deploy a very large (100TB+) file server to support 15-20 video editing workstations, on a very limited budget. Uptime requirements are not high - it'll serve nine-to-five users, so if I need to reboot or shut down for maintenance after 5
PM, nobody will care. Even if it does go down in the middle of the day, it's still not a big deal as long as it is recovered within a reasonable timeframe. Therefore, I'm looking at using a single file server node rather than a cluster, and either Intel JBOD2312S2SP
or Supermicro 847E26-RJBOD1 DAS shelves. Using SATA drives, such as Seagate NAS series, or Western Digital Red, as opposed to Seagate Constellation ES or Western Digital RE NL-SAS drives will provide significant cost savings, and if I add an SSD tier, the
differences are huge - Intel DC S3500 series are very affordable, while even the cheapest SAS SSDs cost thousands per unit, but the question is, what are the trade-offs? I know I will lose clustering support, but what about MPIO? For example, if I use the
Intel JBOD boxes, get eight of them, populate each with 8x4TB SATA HDDs and 4x80GB SATA SSDs, then daisy chain them into two groups of four, and use two dual-port SAS HBAs to connect a cable (from different HBAs) to both ends of each chain, will the resulting
configuration use all four SAS ports to access the drives, thus providing shelf and HBA redundancy? An HBA loss will cut off one of the loop ends, while a shelf loss (for instance, dead expander) will break it in the middle - will it gracefully recover with
SATA drives, or do I need dual-port SAS drives for that? Are there any other caveats with SATA HDDs and SSDs that I need to be aware of?
↧