OU Research Disk (OURdisk)

OU Research Disk (OURdisk) is a large scale, high performance disk resource that is available both on OU IT's supercomputer (already) and on servers across OU (soon).

You can buy portions of OURdisk, where each portion is ~9.3 TB of usable capacity, corresponding to the usable capacity of a single 16 TB disk drive. OURdisk purchases are of one or more ~9.3 TB portions, at $860.03 per such portion (equivalent to $92.39 per usable TB), good for at least 5 years. These purchases are treated as permanent equipment, not as a service. So, a purchase is a one-time up-front payment, not a recurring charge. And, an individual purchase from external funds (such as a grant) of more than $5000 (at least 6 portions) isn't subject to Indirect Costs.

OURdisk's initial capacity is ~3800 TB at OU Norman and ~3800 TB at OUHSC. The Norman part is already in "friendly user" mode, available on OU IT's supercomputer, and OU IT will be rolling out other capabilities gradually.

Questions about OURdisk? Contact support@oscer.ou.edu.

Features

Speed

OURdisk's speed, on a single copy command from the supercomputer to OURdisk, is ~0.7 GB/sec, and the aggregate of all simultaneous copies is 14+ GB/sec. (This is expected to grow as more capacity is added to OURdisk.)

Size

OURdisk is starting at ~3800 TB at OU Norman and ~3800 TB at OUHSC. More capacity will be added at each campus as demand grows at that campus.

Reliability

Based on OU IT's calculations, OURdisk has a tiny chance of losing any data over 5 years.

In particular, OURdisk uses "erasure coding" in an 8+3 server-level configuration: 8 data chunks and 3 redundancy chunks per file, where each file is spread out over 11 servers, on one hard drive per server. This means that OURdisk would have to lose at least 4 servers full of hard drives, or 4 hard drives in 4 different servers, in order to lose any data (or even to lose access to the data).

To ensure high reliability, OU IT developed a Monte Carlo simulator of such failures and found that, over a 5 year period on 1000+ hard drives, the probability of enough simultaneous hard drive failures to cause data loss was immeasurably small.

Technology

OURdisk's hardware is a collection of servers full of hard disk drives, specifically Dell PowerEdge R740xd2 servers, each containing 24 hard disk drives and 2 SSDs for metadata and small I/O transactions.

OURdisk's software is Ceph: https://ceph.com/en/

This is a free, open source Software Defined Storage package that provides high speed, large size and high reliability.