Silo data storage facility

The Maths department has a large capacity local disk storage facility for research users called the silo. This system can accommodate up to 6 TB (6 terabytes) of disk storage although about 4.5 TB is currently fitted and an external disk enclosure allows the installation of an extra 4.2 TB. Hosted on a Sun Enterprise 450 server, this service has been operating continuously since May 2006.

Why use the silo?

In these days of cheap USB external drives with large capacities, it is tempting to ask why use the silo when you can buy a 1.5 TB external drive down the high street for under £100? The reason is reliability and data integrity; USB external drives use SATA disks which offer a very high capacity through high bit densities but that is about all that can be said for them - they are intended for low duty cycle consumer applications, cannot be used continuously for long periods, don't check and rewrite failing sectors and have relatively high hard and soft error rates and gradual data loss over a period of time is common. Where SATA disks are used in servers and storage arrays, they are always used in multiples, usually in some kind of RAID (Redundant Array of Inexpensive Disks) configuration where errors on one disk are corrected by reference to the remaining disks in the array. The manufacturer's warranty for SATA disks is 1 year.

The enterprise SCSI disks fitted in the silo have a maximum capacity of 300 GB but are much better engineered and designed to operate continuously for at least 5 years, which is the standard warranty period. At 10000 or even 15000 rpm spindle speeds, they are also much faster in operation; when the disks are not reading or writing data, they constantly run internal self-tests, seeking out defective sectors and rewriting them or mapping them out and transferring the data to fresh sectors and there is also a regular magnetic refresh of all the sectors in use. The disks are hot-swappable, meaning they can be removed and refitted or replaced without shutting the server down. All this accounts for the relatively high cost of these disks (typically £350 for a 300 GB 15000 rpm disk).

Following system upgrades in May and June 2011, the disks are now organised as three ZFS RAIDz1 pools - there are two pools called 'maths' and 'maths2' for use by all Maths users and these allow up to two disks to fail without corruption or loss of data. There is an additional pool known as 'nucleus' which is dedicated to the nucleus group in Stats and allows one disk to fail without data loss or corruption.

How do I use the silo?

To use the silo you simply need an account on it and to arrange this, send an email to Andy Thomas requesting an account. This will then be set up and the account details will be mailed to you along with some information on its usage. It's as simple as that.

Accounts on the silo are local machine accounts and not directly connected to the College's central computer accounts system but usernames and the groups to which users belong will be identical to college accounts making it easy to transfer data between Maths & College computer systems and the silo, and to make possible the mounting of NFS exports from silo onto Maths systems.

You can copy data to and from your account on silo using a variety of familiar methods - scp, sftp, rsync and by mounting the silo disk(s) via NFS on a Linux or UNIX workstation or PC are all supported options. You can also log into silo using ssh and carry out the usual file and directory tasks, such as moving, renaming, copying and deleting, etc. However, the range of applications and packages installed on silo has been deliberately kept to the bare minimum to discourage the use of this system for computational work, etc as it is strictly intended as a bulk data storage facility.

Are disk quotas imposed on silo?

No! Per-user disk quotas are not imposed on silo as this would defeat the object of this facility. However, there is a finite limit set by the available space in each disk pool - the maximum contiguous file space that can be provided in the largest maths2 pool to any user is 1.7 Tb and this is the maximum single file size possible.

Is data on the silo backed up?

Yes, it is mirrored every 4 hours to a mirror server fittingly called 'silo-shadow', which contains an exact copy of silo's 3 ZFS pools on a single large ZFS pool. silo-shadow runs OpenIndiana, an open source fork of the Solaris operating system. In addition an occasional tape backup is made using a set of dedicated tapes on the LTO 3 tape library system.

Mounting silo disks on your Linux/UNIX computer

Both disk pools on silo are exported to Maths systems via NFS and can be easily mounted on any Linux or UNIX system so that it appears to be a part of your computer's filesystem. To do this you need to decide where you would like to mount the silo disk pools within your systems' filesystem; /home/silo1 and /home/silo2 might be good places so do the following (as root) to mount the pools on your own system:

1. first, create the mount point /home/silo1 on your computer:

   mkdir -p /home/silo1

2. using your favourite text editor, add the new mount by adding the following lines to the bottom of your /etc/fstab file:

   silo.ma.ic.ac.uk:/maths/home    /home/silo1      nfs rw,bg,nosuid,soft,intr,rsize=32768,wsize=32768 0 0

   silo.ma.ic.ac.uk:/maths2/home    /home/silo2      nfs rw,bg,nosuid,soft,intr,rsize=32768,wsize=32768 0 0

3. save your updated /etc/fstab and then mount the new mount by typing:

   mount -a

Your silo data should now be accessible in /home/silo1; if you are a member of the Stats nucleus group and want to acess the nucleus data share, you can repeat the above sequence of commands to mount the nucleus NFS share on a different mount point, eg:

silo.ma.ic.ac.uk:/nucleus/home    /home/nucleus      nfs rw,bg,nosuid,soft,intr,rsize=32768,wsize=32768 0 0

About the silo server

silo is a heavyweight enterprise server engineered for longevity, maximum reliability and very long uptimes. Containing 4 SPARC processors, triple redundant power supplies and a total of 9 disk controllers, it runs the OpenSolaris operating system making full use of the ZFS/RAIDz facilities offered by Solaris since version 10. So if you are comfortable with using standard Linux commands and utilities, you will have no problem using these on silo. Also, if you mount silo's disk pools on your Linux or UNIX system, you will be able to work with your data on the silo just as you do now with the files and folders on your local system's hard disk or your College home directory and you can use your favourite graphical filemanagers such as nautilus, konqueror, etc.