How many Google machines
An interesting tidbit coming out of the Google S-1 filing is that they have spent about $250 million on hardware equipment. From there, we can get a few guesses at the magnitude of the Google system. Based on quick back of the envelope calculations, it looks like Google is managing between 45,000 and 80,000 servers. Here’s how I arrived at this conclusion:
According to calculations by the IEEE, in a paper about the Google cluster, a rack with 88 dual-CPU machines used to cost about $278,000. If you divide the $250 million figure from the S-1 filing by $278,000, you end up with a bit over 899 racks. Assuming that each rack holds 88 machines, you end up with 79,000 machines.
However, one must recognize that equipment is not all CPUs. As a result, you must discount the figure of $250 million to account for routers, firewalls, machines for employees, etc… So let’s assume for a minute that only about $200 million is going to the CPUs. That still leaves us with 719 racks or a bit over 63,000 machines.
Even if we discount other equipment to be costing $100 million, we end up with a bit over 31,654 machines on 359 racks.
So how much processing power is that? Well, once again, the Google cluster document provides some interesting tidbits. Per the document, the racks that were used were
88 dual-CPU 2 Ghz Intel Xeon servers with 2 Gbytes of RAM and an 80-Gbytes hard disk.
That means that, on the low end, the Google cluster has the following stats:
- 359 racks
- 31,654 machines
- 63,184 CPUs
- 126,368 Ghz of processing power
- 63,184 Gb of RAM
- 2,527 Tb of Hard Drive space
In the middle range of my estimates, the cluster would have:
- 719 racks
- 63,272 machines
- 126,544 CPUs
- 253,088 Ghz of processing power
- 126,544 Gb of RAM
- 5,062 Tb of Hard Drive space
And on the high end of my estimates:
- 899 racks
- 79,112 machines
- 158,224 CPUs
- 316,448 Ghz of processing power
- 158,224 Gb of RAM
- 6,180 Tb of Hard Drive space
Assuming that the 1Ghz chip is going at about a third the gigaflops of a 2Ghz processor (3.3Gflops), we can then guess at the size of the Google supercomputer. Just for the sake of argument, let’s go with 1 Gigaflop per processor. This means that the Google supercomputer has about 126 teraflops of power on the low end of my estimates, 253 teraflops on the middle end, and 316 teraflops on the high end. This would easily put it on top of the list of fastest computers in the world.
Any way you slice it, that’s a lot of power.
Related Terms
Google12 Comments
1Random Thoughts — September 19, 2006 at 6:07 am
The hippocampus keeps strengthening the associations among these new elements until, after a while, it no longer needs to do so. The cortex will have learned to associate these various properties itself to reconstruct what we call a memory. Soooo??? What is the power of google supercomputer? Do you see any connection? Consider the current working model of google … google introduced a revolutionary system into search engines … that propelled google to the number one position in search business …
Google — by celardore (Score: 5, Interesting) Thread There doesn’t seem to be any mention of the GoogleNet. While it may not be used for figuring out sums and what-not, it does have an estimated 126 terraflops of computing power. I’d say that’s notable. I bet at least half those terraflops are devoted to advertising aswell. how many aren’t listed? — by rritterson (Score: 5, Interesting) Thread How well does this represent the real top 500?
[…] with del.icio.us | Email this entry | TrackBack URI | Digg it | Track with co.mments | Click here for copyright permissions! Copyright 2006 Mathew Ingram […]
how big is the Googleplex? The Times says the number of servers the company is currently operating at its 25 locations around the world is in the 450,000 range. That figure has more than quadrupled since 2004, when Google’s server operation was alreadyestimated to be one of the world’s most powerful distributed supercomputers — rivalling anything that NASA or the NSA have. Based on estimates of the power that half a million servers would consume, that means Google’s electricity bill is likely somewhere between
for my professional site and my portfolio, but I thought Blogger would work better for my blog. Part of my descision stems from my current love affair with Google, but I also want to get as far away from the server admin business as possible. I figure Google’s servers should prvide me with plenty of data security. Plus, I’ve been annoyed with iPowerWeb for sometime now. I plan to keep earthshine.org, but Autumn and I might be ditching autumnjerry.net in favor of another Blogger site
. The series got attention from some media outfits and my readership grew. The bet on longer, more analytical, types of blog entries seemed to have worked. I followed suite with a quick overview of Google’s S-1 filing, which itself was followed by some projections as to how many computers Google had at the time (that entry, by the way, is still one of the more popular ones on the site as few other people have attempted to do the math again). Media attention grew further. In November 2004, I said that Apple should consider Intel and Windows
(and most generalizations of FLT have been), then a computer search may have a chance of coming up with the counterexample. Peter Norvig did an initial hunt but came up empty. I extended his results, and also came up empty. Now that I have access to lots of machines, I’d like to extend the search a bit further. The old article I wrote is still valid, though it contains a misstatement that’s made all the more embarrassing by being in ALL CAPS. I’ll be lazy and leave it as one of those pesky
logistics involved in selling consumer electronics. As mentioned before, that’s a completely different business from selling adwords. Nokia is pumping out 11 phones per second worldwide. You can have all the Ph.D:s in the world in your staff and asupercomputeron top, that’s still a huge challenge. Nevertheless it’s a telling sign that the two most talked about phones at the moment comes from the internet and computer industry. Updated: New York Times
9portaleco — November 21, 2007 at 7:19 am
Quite interesting. However, Google will certainly not buy all machines. Some or even a large part of the equipment may be leased. A modern INTEL x86 has in the range of 10 MFlops. My estimate would yield about 1300 Teraflops being installed in the new Google data center in The Dalles, Oregon. Thus consumers are driving the IT world and not national agencies with their super computers any more. There is an analogous story for data. A couple of years ago one of the largest insurance companies in the world was struggling with 3 Terabytes of data in their data center. Today you can find that capacity installed in many living rooms.
[…] Google, but I also want to get as far away from the server admin business as possible. I figure Google’s servers should provide me with plenty of data security. Plus, I’ve been annoyed with iPowerWeb for […]
[…] the other hand, they could look at increasing the number of crawlers they are using. We know that Google has a lot of machines but trying to scale to the point where they can monitor a trillion pages via crawl would require a […]
[…] of building very large machines. Similarly, one of the largest web-based application, Google, uses tens of thousands of machines which interact with each other as a single […]
