FOCUS STRATEGY
Issue 1, December 2008
can enjoy econonmies of scale saying such large providers enjoy a wide set of buy/build options for the wide area network to support internal and external data transport to their data centers, and can create and manage dedicated networks.
Terremark describes a cloud data center as something that is agile, based on a pay for use model, that is scalable, built for rapid deployment and that can meet capacity on demand.
But proving that in the cloud it is not always clear what is being compared with what Microsoft adds the caveat: “By “embarrassinglhy distributed” applications, we mean applicatiosn whose implementations do not require intense communicatsions within large server pools.” As examples it lists Skype, P2P file sharing, eBay and Xbox Live.
RAIN DANCING
Whether at a mega or modular scale everyone agrees that cloud computing data centers will require a lot of management. At the infrastructure level, Terremark is attempting
MICROSOFT’S CLOUD DATA CENTER POSITION?
On delivering embarrassingly distributed cloud applications A paper by James Hamilton, Albert Greenberg and Ken Church of Microsoft compares the costs associated with building and operating mega scale data centers with modular or “condo” style data centers.
Abstract: Very large data centers are very expensive (servers, power/cooling, networking, physical plant.) Newer, geodiverse, distributed or containerised designs offer a more economical alternative. We argue that a significant portion of cloud services are embarrassingly distributed – meaning there are high performance realisations that do not require massive internal communication among large server pools. We argue further that these embarrassingly distributed applications are a good match for realisation in small distributed data center designs. We consider email delivery as an illustrative example. Geodiversity in the design not only improves costs, scale and reliability, but also realises advantages stemming from edge processing; in applications such as spam filtering, unwanted traffic can be blocked near the source to reduce transport costs…. Given how much the industry is currently investing in the mega model, the industry would do well to consider the micro alternative.
Large Tier II+ Data Center Servers Specs Power (Peak) Capital
Annual Ex- pense
Building Power
13.5 MW (= 250 Watts/server × 54k servers)
over $200M
$3.5M/year (= $0.03 per kw/h ×24×365 hours/year ×13.5MW)
Annual Income Rental Income None Figure 3. 40
www.datacenterdynamics.com 54k Condo Farm (1125 Condos)
54k (= 48 servers/condo × 1125 Condos)
13.5MW (= 250 Watts/server × 54k servers = 12 KW/condo × 1125 Condos)
$112.5M (= $100k/condo × 1125 Condos)
$10.6M/year (= $0.09 per kw/h×24×365 hours/year × 13.5MW)
$8.1M/year (= $1000/condo/ month × 12 months/year × 1125 Condos − $200/ condo/month condo fees. We conservatively as-sume 80% occupancy)
to carve a management niche for itself. Its position is that if you are to build a cloud data center then at very least it has to deliver such things as an ERP system, a CRM system, email and possibly a transaction web site that is available 24/7 meaning the infrastructure may need to be restructured to more readliy adapt to constantly changing service demands. “Service management must be incorporated as part of the underpinning to ensure quality service delivery. Management has evolved from systems based to business process based,” says Joost Metten, CEO of Terremark Europe.
Terremark describes a cloud data center as something that is agile, based on a pay for use model, that is scalable, built for rapid deployment and that can meet capacity on demand. Terremark is pushing “Enterprise Cloud”, based around its Infinistructure model. Accessed through a web portal the Enterprise Cloud is an application that allows you to dynamically provision servers from a pre-allocated pool of dedicated computer resources. Preconfigured server templates are available across Microsoft Windows, Linux, and Sun Solaris operating systems. Plug- and-play bandwidth is available from 160 carriers, and a full reporting interface allows instant insight into resource utilization and application performance.
The firm admits that a more mature pricing model is needed and says that it is working on one.
And don’t think the networking suppliers are leaving this market to others. Andy Scannell, virtualisation solutions group manager at Cisco, said: “We look at a future where there are whole host of pools of resources, and then ensure they are linked by fibre fabric, 10 gigabyte – 40 then 100 – if you have apps moving across differ servers, you probably don’t know the complications and flexibility required in ensuring the bandwidth is able to respond rapidly to change.”
Cisco spent $5.2 billion on research and development last year and is looking into automated data center of the future. “As you move towards this dynamic data center concept everything from configuration of resources, ensuring that the right resources are available to capacity planning becomes part of the this paradigm. If you look at aggregating silos of resources under unified fabric we expect to touch more points in the data center.”
WHY SO SERIOUS Despite all the proselytising and money being invested there remain are serious issues to be overcome. Users are nervous about sharing server time (security?) and companies who
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72