Storage facilities

You may have noticed that hard disk storage continues to get cheaper, and it’s been doing so for a very long time now. I see that back in 1982, I paid fellow RWCer Mark Newton £500 for an 82MB SCSI hard drive from Control Data, but today that money would buy me around 7.5TB of storage. We’ve covered the long, long road from megabytes via gigabytes up to terabytes over the past 27 years. Clearly, you need a box in which to put such a pile of storage, and there’s little point wiring it up to just one server. Far better to mount it directly into your network as a NAS device. In this operation mode the storage box appears as another server on the network, so instead of talking to \\Myserver\SharedData\ you now talk to \\Nasserver\SharedData\ instead, and all the Active Directory accounts stuff will just work. If you want to get really flash you can set up the box as an iSCSI device, and talk to it as if it were just a SCSI drive fitted in your local machine, but in reality it’s over there in a rack (where “there” could be in the next building).

To celebrate the current falling prices, I’ve just ordered a Thecus N7700 Ultimate Tower Of Storage. This is a solidly built box that sits on my Gigabit Ethernet LAN and works as a NAS device, or as iSCSI too, apparently. Seagate is making 1.5TB hard drives nowadays, and the Thecus will accept seven of these, so allowing for one disk’s-worth lost to RAID5 I’m looking at a volume size of 9TB. Of course, I could slice and dice this 9TB into more usefully sized lumps, or I could just use it as one huge bit bucket (I suspect it will be the latter, when the device arrives later this week and is pressed into service). And the cost? Less than £2,000 for the box, fully populated with all those disks.

If you want to save on the pennies then shove in three 1.5TB drives for a total of 3TB of storage: you can add extra disks later without disturbing the existing RAID, simply expanding it automatically to take advantage of the extra space. Of course, this box represents a single point-of-failure, and 9TB is a lot of stuff to lose in one go, which is why it will be stress-tested for a few months before deployment, and a second one then purchased to go into another building across the yard.

Microsoft DHCP blog

The number of blogs currently spewing out of Redmond is quite amazing: just about every group has a blog in which they talk about the product, the stuff they’re still working on, and apologise for limitations in what just shipped (or escaped). I was particularly intrigued by the DHCP blog.

Now DHCP is pretty vital stuff, which most of us rely on all the time: think how boring it would be to manually type in a “free” IP address every time you connect to the LAN. It turns out that the DHCP server in Server 2007 R2 is clever: not only can it cope with multiple DHCP servers on a LAN, but you can cluster them for failover support. This is called a split-scope or 80/20 configuration, and there’s an excellent walkthrough on the DHCP team blog at http://blogs.technet.com/teamdhcp/default.aspx. Clustering your DHCP service ensures that if the main server fails the service is available from the other partner in the failover pair, and since you can normally have only one DHCP server on a network, clustering it is a real Jon Honeyball makes the most of tumbling prices by ordering a tower of storage, and David Moss continues setting up a course management system.belt-and-braces measure. (Sure, you can subdivide your network and ensure that DHCP pings don’t travel between subdivisions, but I’m keeping this relatively simple.)