My “two cents” view of a Data Centre
I have often wondered what a real data center looks like; now I have some firsthand experience to share, thanks to Mr Ross Summerfield who kindly provided me an opportunity to visit the Hume Data Centre of DHS (Department of Human Service).
To me, it’s far enough from the city to be called in the middle of nowhere, view in Google Map.
Just a bunch of storage warehouses and the rest are farms, mountains and high way.
I was all prepared to take some photos, but when I arrive at the place, I was told mobile phones are not allowed to be brought into the computer/equipment room. Unsatisfied that I have no picture to show, after I came back I searched online and found this article.
First Data Center I have visited, and the place with the most strict security policy too. (If I’m over exited, you can understand why.)
The facility is ASIO T4 certified as a “secure area” under the Government’s Protective Security Manual, with 24×7 staff and time-stamped IP cameras. (from the article)
What you probably couldn’t tell from the picture is that it is really loud and windy inside.
There are multiple separated rooms in the building, I was told that not all equipment belongs to DHS, they occupies 2 buildings though, it’s a collocated site. After I came back I dug up a bit, found this company: Canberra Data Centres. They have a nice video:
It’s interesting to see how the robot arm fetching tape from the tape rack. A bit of the background, tape holds more data than disks, so data not accessed so often are stored in taps. I found a picture from the internet; the actual tape cartridge library looks very similar to this:
Then there is also the “switchboard”, sorry if it’s not the right word for it, basically it allows you build a direct connection between any rack anywhere (on the same floor probably).
I heard that they have equipment upgrades all year round, “you start from one end, by the time you finish at the other end, and it is probably time to start upgrading the oldest one all over again.” A few of big hardware vendor name came up so often that I actually remember are IBM and Hitachi.
One of the new terms I picked up along with the fact that “you buy a number of MIPS from IBM rather than a number of mainframe computers”, here is a rather interesting explanation of the term:
MIPS: Million Instructions Per Second, (or Meaningless Indicator of Processor Speed, Many Important Selling Points, etc). A crude and not very meaningful way of expressing raw computer power. Not often used by IBM, which prefers to use ITR to express the power of a machine by comparing it with another machine in the same range.
Unfortunately I wasn’t able to get the details about what kind of software structure the data center operates on, namely whether they use OpenStack (which is the awesome project I’m doing my internship with), CloudStack or similar sorts, the answer is roughly something like “We use open source technologies and build in house solutions on top of them.”
I did get a high level overview of what kind of tasks are running inside those monstrous machines. First there are a lot of data stored here about Australian citizens, if I remembered correctly “it will take about one month to just read the entire index catalog”, I couldn’t remember the exact figure, DHS manages the record for Medicare (medical record) and Centrelink (everything related to get welfare support from government). “High end mainframes are good for batch processes, which handles regular payment…middle ranged servers hosts virtual machines handles websites, remote desktop service…”. There is a number of other tasks which assumed run on the mainframe: “we also run regular cross checks with the tax office’s records (in an anonymous, privacy preserved way) to catch identity frauds”, “we make sure the medical practitioners are not forge claims to resell government subsidized medical supply to other countries where they are not subsidized”.
== End of Main Article ==
Something I don’t pay attention to until I need to write something “formal”:
I always feel it’s such an over complicated grammar rule to have plural vs singular, now I finally got some proof !
(It’s probably worth noting, there is no such worry in Chinese.)
For those non-native English speakers who may find interesting: