On 09/14/2013 09:59 AM, Grant wrote: >> It's time to switch hosts. I'm looking at the following: >> >> Dual Xeon E5-2690 >> 32GB RAM >> 4x SSD RAID10 > If I make this 6x SSD RAID10 with redundant power supplies, what is my > weakest link as far as hardware? If a CPU craps out, will the system > keep running? > > - Grant > consider making the main memory ECC too and flick the correct switches in kernel to ensure ECC is monitored. no point in ensuring the data is resilient if the content is garbled.
and also consider what happens if the raid controller fails due to a popped capacitor five years from now will you still be able to get a like for like replacement ? bear in mind that you may have to keep the raid card firmware up to date in order to be compatible with newer cards of course, this is all relative to how long you stay with your host but you have to decide how much resilience you want to build in. it's the mechanical parts of spinning rust or pseudo mechanical nand gate switching for SSD that will tend to fail, secondary to that in most places the PSU acts as a static cling with a dust blower attached, and any slight knock knocks the dust off causing a short circuit especially if any humidity is caught in the air also consider the fans blowing around the air inside the machine you can start thinking what about earthquakes or flooding in the area - surely you want to ensure two geographically diverse locations cpu / motherboard failures on server spec tend to not be very likely, especially if the environment is controlled (air filters/temp/power) a great many things happen that are beyond anyone's sphere of control - just look at the new york datacentres during hurricane Sandy; would it be better to have had more diesel on site or just everything replicated at another site ? the real question is what is your expectation of uptime and how can your budget match that. uptime is affected by software as well as hardware don't forget.