Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The SuperMicro 3U blade server looks very promising - MicroCloud!

If i configure RAID 1 on the 2 SATA drives in each of the blades, how would that work out in terms of reliability?

Compared to 8 1U servers with hardware RAID, the price difference seems huge - $10K vs $32K - so i would love to try this approach.

Does anyone have first hand knowledge of the quality of this SuperMicro product?

share|improve this question
Question is unclear; are you asking about mdadm raid, or this particular hardware? If the latter, then this is probably too narrow to be on topic, and in either case, is more of an opinion poll than an answerable question, and so is off topic. – psusi Jan 24 '12 at 23:50
You're right. I ended up asking two questions in one – andy318 Jan 25 '12 at 0:02
up vote 2 down vote accepted

Mdraid has proven quite stable for a couple of years now and shows quite trouble-free handling.

The main reliability problem you might run into is rebooting with a failed primary boot drive.

  • you need to have your boot loader redundantly written to the second disk as well - a fact many admins forget
  • the BIOS of your server machine needs to be able to do a failover to the second hard disk for booting, should the first fail
  • you might see a scenario where a drive is not quite dead but taking ages for reading even a single sector - thus giving you a reboot taking a whole weekend.

All of these problems are alleviated by using a hardware RAID controller as it includes the necessary abstractions and fault detection logic.

Apart from that, you should be fine with a software RAID, it works quite well performance-wise too. If you need more detailed performance considerations, read the numerous performance reviews you can find on Google or in this article by Jeff Garzik - a Kernel developer at RedHat.

As for your second question, I have not much experience with Supermicro gear, but what I have seen so far was mainly commodity desktop-grade stuff built into cheap 19" enclosures. It might work out well as long as there is redundancy, I also know that large hosting providers like Strato (Germany) were using Supermicro stuff in their datacenters, but personally I would rather look for "proper" blades.

share|improve this answer

Linux software RAID is completely reliable. A few things to note, however.

Most hardware RAID arrays have a battery-backed write cache. For software RAID you don't have this. A UPS would not come near the length of time a BBWC can hold data, which is months. Thus you need to enable write barriers which would cause a performance hit. See Your kernel has to be fairly recent (2.6.33?) to actually have this support.

Most hardware RAID arrays support hot swapping. You need to make sure the SATA controller, Linux drivers, and drives support hot plugging and you will need to manually issue the commands to achieve something like hot plugging.

Most hardware RAID arrays have a skulker that will continuously (or periodically) scan all the disks in entirety to detect errors before both disks go bad. Linux md has this as well by writing values to "/sys/block/md0/md/sync_action" but again you have to manually configure this.

share|improve this answer
All SATA drives and the linux drivers support hot swapping. It is just a question of whether or not you have the physical chassis to swap the drives easily. – psusi Jan 25 '12 at 15:06
BBWCs typically are designed to hold the data for 24 hours only. There are flash-backed variants (FBWC) which will hold the data indefinitely. – the-wabbit Jan 28 '12 at 7:46

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.