I’m an intermediate PC specialist in higher ed and do everything from help desk, printers, audio/video, imaging, light server management, automation, etc. for a small department of 100 users. For those users it is just me and my IT director. He handles the macro stuff and I do the micro day-to-day.
The elephant in the room for us is our aging infrastructure so we’re tackling this project together because of my likely ‘inheritance’. Currently, we have an old Xserve as our primary server which we mainly use for LDAP. Primary storage for users is split between multiple local RAID arrays attached to the Xserve over FireWire 800/eSATA which users access over mapped network drives. Primary and off-site backups are also a hodgepodge of RAID arrays attached to Mac minis. Each tier is about 50TB, combined 150TB. The arrays have four bays and are in RAID 5. Yea, it is messy. The main reason we are doing this is because the Xserve is showing its age and is becoming unreliable. The second reason is because of just how much of a headache our storage solution has become to manage. It is “scalable” in a rough sense because we can just add more arrays, but keeping track of what data is on which array and juggling storage is just unmanageable long-term.
My boss is interested in moving over to cloud block storage. Although we have a 10 Gb/s connection from my limited research using cloud block storage doesn’t seem like a good choice for us. Pricing calculators show AWS EBS/EFS and Google Cloud volumes will cost between $1,200 and $2,000 monthly for 50TB of storage. I think the people recommending cloud storage are only familiar with large businesses/corporations. We’re higher ed so we don’t pay for bandwidth, electricity, rack space, etc. only hardware maintenance. Comparing $1,200/mo vs. a one-time purchase of a $10,000 array that lasts 5-8 years just doesn’t compare. The only thing it costs us to maintain them are time (which we have plenty of) and dead HDDs and PSUs. In terms of availability/up-time local vs. cloud is acceptable. I would guess 97% vs. 99.9%.
If I were to pull the trigger on a purchase right now I’d go with a beefy 2x Xeon SuperMicro server for LDAP and any other service we might need to run in the future. It would be attached to another SuperMicro chassis with a medium tier motherboard/CPU, 100+ GB of RAM, and 30+ hot swap bays we could half fill up now, and expand later on. I’d run Ubuntu because I’m familiar with it and put the drives into ZFS RAID. Another 2 similar servers would act as primary backup and off-site backup (we have free off-site rack storage as well).
There’s a few things I’m unfamiliar with in this scenario: How would I attach the primary storage to the main LDAP server? USB 3.0/eSATA? We could just get a 10 Gb/s switch and NICs and do it that way, but at that point FreeNAS might be a better option for the primary storage vs. my first choice of Ubuntu + ZFS.
What would you guys recommend in our situation? Should we be looking into pricey SANs? Money is always hard fought in higher ed. One certainty is that the Xserve will be replaced with a Linux server. We prefer *nix + LDAP. Probably either Ubuntu/Debian (my preference) or openSUSE (my boss’ preference).