Comments on: Cloud computing is foggy thinking

By: AU Cloud Computing

AU Cloud Computing — Wed, 19 Aug 2009 23:30:58 +0000

Always like to see info on Cloud Computing! Looks like Australians are starting to wake up to it now with Telstra announcing a $500m spend this week on cloud computing services.

By: Randall

Randall — Mon, 05 May 2008 00:47:36 +0000

You make very good points, some I agree with in my post:
http://www.qrimp.com/blog/blog.The-Open-Cloud—-the-future-of-cloud-computing.html

The important issue right now though, is that there are myriad applications for cloud computing where the network latency will be an issue regardless of the hosting environment. What Cloud computing offers over traditional single server hosting is the scalability for multitenant apps. You don’t get that with colo, shared server, or managed hosting.

And that is why it isn’t hype — it’s awesome.

By: The Gathering Storm: Cloud Computing « Information Spillway

The Gathering Storm: Cloud Computing « Information Spillway — Tue, 05 Feb 2008 17:19:33 +0000

[...] Storage expert Robin Harris counters noting that networking is currently a bottleneck and that only Google is able to make cloud computing work. I think Mr. Harris is correct that networks are going to hinder this process. I, however, disagree concerning the Google magic that Yahoo and Microsoft are unable to reproduce. Google’s two competitors are unable to compete head to head, but there are other providers like Amazon Web Services, Sungard, EMC, and SAP already pushing into the services product space. [...]

By: Steve Jones

Steve Jones — Wed, 30 Jan 2008 09:27:14 +0000

I’m always a bit concerned when cost-curves for different product groups are combined as it depends on what factors you choose. Yes the cost per MB of disk storage has reduced several orders of magnitude faster than networks (although comparing a 10Gbe switched network vs a 10Mbps shared bus topology isn’t quite right; there’s far more that 1,000 x the data carrying capacity in a 10Gbe switch than a (what was very expensive) coax LAN.

However, if we look at a different performance metric – IOPs and MBps on the hard disk market then you get a very different picture. Today’s 15K hard drive can do maybe 180 random IOs per second and perhaps 80MBps. Those are respectively around 5 and 100 times better than what was available 25 years ago. Given that many applications are limited by storage performance this is a real issue.

So when you make you comparison, be careful over the choice of metrics. You get wholly different answers depending on what you chose. Many of the reasons are governed by physics and geometry. Disk storage capacity increases to the square of density (as, famously does semi conductor capacity). However, sequential read speed only goes up linearly with bit density when constrained by the physical limits of moving parts. What governs basic network speed for a single serial links will be constrained by all sorts of things including transistor switching speeds (which have increased by maybe 1,000 in the 25 years). That’s another of those linear relationships.

So, square relationships (storage capacity, RAM, aggregate switching capacity etc.) proceed at a different rate to linear relationships (like sequential read speeds) with mechanical ones far behind those two.

By: Gavin

Gavin — Tue, 29 Jan 2008 09:13:18 +0000

Also none of this is really new even conceptually. The grid folks have been pushing the ‘compute-anywhere’ vision for years.

The issues that limit grid aren’t being solved by calling it ‘cloud’. Data is valuable – companies need to control where it is, who gets access. Algorithms being executed can also be extremely valuable and are trade secrets in many industries. Moving data is slow and expensive and the trend isn’t for this to improve (data sets are doubling per year in many industries). However CPU cycles are cheap and getting cheaper.

Google is a very special case with many advantages when it comes to ‘cloud’ type computing – however even they are secretive about algorithms used, and I doubt they’d be interested in letting their computes be shared by just anyone.

By: Dan

Dan — Mon, 28 Jan 2008 20:42:41 +0000

Cisco is taking a step announcing a 15 Terabit backplane switch that will support over 100 Gigabit ethernet. But, it’s going to be a long time before the world goes to a 100G standard. http://www.cisco.com/en/US/products/ps9402/index.html

By: nyrinwi

nyrinwi — Mon, 28 Jan 2008 18:36:23 +0000

And what about the security implications of “cloud computing?”

Security has been an afterthought in so much of technology history, it’s embarrassing. We’ve got computer viruses on picture frames now for god’s sake! When we can make appliances that are virus-free, then I’ll trust my data the the clouds.

Cloud computing? I can’t get the visions of Skynet and the terminator out of my head.

By: scalability.org » Blog Archive » Computing in the clouds ...

scalability.org » Blog Archive » Computing in the clouds ... — Mon, 28 Jan 2008 17:35:06 +0000

[...] Robin at Storagemojo tears into the latest buzzword-enabled marketing phrase, cloud computing. Robin’s thesis is that there are impediments to moving to the cloud, those being bandwidth and the “non-magic” nature of Google’s infrastructure. [...]

By: Daniel Lemire

Daniel Lemire — Mon, 28 Jan 2008 15:05:39 +0000

For once, I disagree with you. See my rebuttal:

http://www.daniel-lemire.com/blog/archives/2008/01/28/the-network-is-the-bottleneck/

By: The network is the bottleneck?

The network is the bottleneck? — Mon, 28 Jan 2008 15:04:34 +0000

[...] There is a really nice article on StorageMojo about Cloud Computing. Cloud Computing is more or less the idea that you can offload your storage and processing tasks to a very large set of computers, typically maintained by some large company (such as Amazon). The novelty is that you abstract out where the data is held and which machine does the processing — not unlike what MapReduce does. [...]