In a world of large data centers called clouds, intensive software processing applications that once took hours to do can be done in minutes if not seconds. One of the problems though is transit time from the point of data collection to the data center and back again. The time to transmit data to and from a data center, the cloud, is often limited by the speed of the internet. This is not a problem for some applications, but for other applications, such as Internet of Things, applications, where a machine must respond to tens of thousands of events quickly, it can be problematic. The time to send the data can far exceed the time to process the data.
Simply put, does this mean that data centers should be smaller and placed, almost everywhere. Or should they continue as they are as a massive buildings containing thousands if not millions of microprocessors.
Recent data architecture trends suggest that the future will need both and will have both. Data scientist look at the pool of over 8 billion cells phones and see a network of over 8 billion microprocessors that remain idle most of the time. Tapping into cell phone compute power may be one way of offering a new way to process and deliver large amounts of data quickly.
But there are more unused processors in the world that can be tapped. There are a large bank of computers at companies and organizations that go unused at night. As well, during the day, they are often underused. This begs the question, should PCs be designed differently. That is should they be designed so that they can be automatically configured and connected together as data centers.
There is even more unused computer powerout there and more coming on the horizon. Automobiles. Automobiles complete with processors for GPS applications and self-driving applications. With hundreds of millions of automobile processors sitting idle isn't it about time data architects think about how to connect these processors as the ultimate moving interconnected data center - the moving cloud.
Monday, September 4, 2017
Saturday, September 2, 2017
Look Before You Leap: public clouds reported to be more costly than private clouds
Listening to all the great features and benefits of the public cloud, it's easy to get caught up in the allure of public clouds. Not only is there the attraction of powerful never-been-done before cloud apps, but also the sense that the public cloud will eliminate all your IT headaches, drama and costs.
According to recent studies (1) this is not always the case. But in all things cloud, it depends on the company and who you talk to. And it also depends on the public cloud you use. Many companies, that are IT savvy, have come to realize that they actually save quite a bit of money with a private cloud, even when considering the cost of IT upgrades and IT staff.
Apparently it's not uncommon after switching to a public cloud to wince once the monthly bills start arriving. Bills ranging in the $40,000 to $100,000 a month range cause many CIOs to think twice. And what they think about is how big of a colocation private cloud that can buy with that amount of money. And they think whether or not its worth it to have their business records in a public cloud.
Even with those considerations, CIOs must still pause and think. Will public cloud costs come down in the future? As well, just like there are hidden public costs are there also hidden application benefits of a public cloud? Finally, there is a question of whether or not a company wants to remove itself to far from the evolving world of cloud technology? CIOs might also see some value of having first class cloud professionals on-board, especially if there isn't much of a cost difference between private and public clouds
The cost debate between public and private clouds tends to end after a private cloud reaches a certain size. If a company has a huge data center, its difficult for a public cloud to beat the cost of a private cloud (2). Then again, that's not to say that even the biggest private clouds can make some use of a public cloud to save money and reduce risk.
The competing benefits of private clouds and public clouds will always tilt depending on the application and the company. Such competing benefits is what is driving the growth of hybrid cloud architectures. The hybrid model, which is part public and part private cloud. may offer the best cloud cost/performance ratios. However, again, that will depend on who you talk to, what they know and what the cloud is used for.
Cloud Learning Links
(1) Public vs. private cloud cost comparison finds enterprise winners, TechTarget, Aug 2017, Robert Gates
(2) The Complete Cloud Buy VS Build Calculator, TheCalculator,
(3) Google Cloud Pricing Calculator, Google
According to recent studies (1) this is not always the case. But in all things cloud, it depends on the company and who you talk to. And it also depends on the public cloud you use. Many companies, that are IT savvy, have come to realize that they actually save quite a bit of money with a private cloud, even when considering the cost of IT upgrades and IT staff.
Apparently it's not uncommon after switching to a public cloud to wince once the monthly bills start arriving. Bills ranging in the $40,000 to $100,000 a month range cause many CIOs to think twice. And what they think about is how big of a colocation private cloud that can buy with that amount of money. And they think whether or not its worth it to have their business records in a public cloud.
Even with those considerations, CIOs must still pause and think. Will public cloud costs come down in the future? As well, just like there are hidden public costs are there also hidden application benefits of a public cloud? Finally, there is a question of whether or not a company wants to remove itself to far from the evolving world of cloud technology? CIOs might also see some value of having first class cloud professionals on-board, especially if there isn't much of a cost difference between private and public clouds
The cost debate between public and private clouds tends to end after a private cloud reaches a certain size. If a company has a huge data center, its difficult for a public cloud to beat the cost of a private cloud (2). Then again, that's not to say that even the biggest private clouds can make some use of a public cloud to save money and reduce risk.
The competing benefits of private clouds and public clouds will always tilt depending on the application and the company. Such competing benefits is what is driving the growth of hybrid cloud architectures. The hybrid model, which is part public and part private cloud. may offer the best cloud cost/performance ratios. However, again, that will depend on who you talk to, what they know and what the cloud is used for.
Cloud Learning Links
(1) Public vs. private cloud cost comparison finds enterprise winners, TechTarget, Aug 2017, Robert Gates
(2) The Complete Cloud Buy VS Build Calculator, TheCalculator,
(3) Google Cloud Pricing Calculator, Google
GPU Cloud Computing Gains Momentum: Walmart Picks Nvidia's GPU Technology
If your designing your own private cloud or you are looking for a public cloud provider, one important consideration is cloud hardware. New advances in integrated circuit (chip technology) are quickly making traditional cloud hardware obsolete. Servers built on microprocessors, such as Intel's famous line of processors, are becoming a thing of the past. Replacing these microprocessors in cloud servers are graphic processing units (GPUs). Nvidia, which has its roots in graphic processor design, is quickly taking data center market share away from Intel. As the Market Realist points out in a telling graph, cloud builders are choosing Nvidia's data center products hands down over Intel.
As a specific telling example, Nvidia recently won over Walmart with its data center products. Walmart will use Nvidia's products in a GPU farm configuration to support its business operations. Amazon has also reported to have adopted GPU based technology for use in its data centers.
The switch in technology from microprocessor-based clouds to GPU-based clouds and even neural clouds, underscores just how tricky it can be to build and maintain a private cloud. Companies that have invested heavily in cloud microprocessor technology, now must switch over to GPU technology to remain competitive. This not only means changes in hardware, but also software. Even more importantly it means changes in the staff at the leading cloud providers. Professionals with artificial intelligence, neural network expertise and graphic processing experience are high demand items at cloud provider companies right now.
The change in cloud technology is not expected to be the last. But the switch to GPU and neural cloud technology is expected to continue. This all spells good news for companies that have a multiple cloud technology and are not locked into a specific cloud vendor. On the other hand, for companies that have locked themselves into cloud vendors that have not switched over yet, problems may be on the horizon. Technology changes at cloud companies, especially small ones, can spell trouble. If a technology change is not done right, melt downs can occur and data can be lost. Maybe even worse than that, small cloud vendors can simply go bankrupt because they lose market share too quickly.
Cloud Learning Links
Telsa Data Center GPUs for Servers, Nvidia
Data Center: NVIDIA's Key Growth Driver in 2018, Market Realist, Paige Tanner
Subscribe to:
Posts (Atom)