Posted February 14, 2018 12:29:22 The next wave of cloud-powered data storage will be the first to be distributed across multiple physical data centers, and in a way that is less costly, more reliable, and less vulnerable to the whims of an aging infrastructure.
These cloud-hosted data centers will be called “dynamics,” and they will be designed to be more reliable and resilient than physical data centres, according to a recent presentation at the International Data Centers Conference in Berlin.
In addition to the hardware, the cloud-enabled data will be managed by software, meaning that there will be no physical storage for those servers that are not being used for analytics.
“We are looking for a cloud that can handle the volume of data that is being generated today, but also that can be distributed with a level of reliability and resilience that you don’t get with physical data,” said Andrew Prentice, a research scientist at MMB who presented the research at the conference.
Prentice and his team used the technology of the company, Dynamics, to simulate a large data centre in the cloud.
“In the simulation we have a large number of servers, each with a different workload, each of which will be hosting different kinds of data,” Prentice said.
“When you look at each server in terms of the physical storage that it has, you’ll see that they’re all physically isolated from each other.
And as the workload increases, it becomes easier to see what kind of workload each server will be able to handle.” “
So when we have the same workload, it’s very easy to see which server will do which job.
And as the workload increases, it becomes easier to see what kind of workload each server will be able to handle.”
This allows the Dynamics servers to be used to provide redundancy and to support data storage operations that are more reliable than physical servers, Prentice explained.
“You can do something like this, for instance, and we have data that’s going to be stored on our cloud server and we’re going to want to put it on our server where the other server can be used.
So that server can keep up with that workload,” he said.
This type of redundancy can be important for large data sets that need to be shared across multiple data centers.
“This means that it becomes much easier for you to have redundancy between your physical data center and your cloud data center, because if the physical server fails, you can replace the physical data server and move the workload to the cloud server, and that’s very efficient,” Prentices said.
The researchers used a simulation of a cloud-scale data center to demonstrate how the Dynamices could be used for data storage.
The Dynamics simulations have two cores that can both host multiple virtual machines running the same software application.
“These cores can be connected to each other and the data can be moved between them in real time,” Ponsent said.
These virtual machines, which are called “fusion” machines, can be deployed in different virtual environments and are then able to be controlled by other virtual machines that are also running the software application on their cores.
The first virtual machine was deployed on a Dynamics core.
In the second simulation, a second core was deployed, but the first core was only used to host a single Fusion machine.
The Fusion machines were used to move the data to the second Dynamics virtual machine.
“The second Fusion machine was configured to host the data that we wanted to move to,” Ponson said.
For instance, if the first Fusion machine hosted a dataset of historical data and the second Fusion server was used to create an infographic, the Fusion server would then be able move that data to that Fusion server.
“That data would be transferred from the Fusion machine to the Dynamical server, which is a physical server, but we can also move the same physical data from the Dynamic server to the Fusion servers,” Ponisent said, adding that this process can be accelerated using a “fuse hub” that is connected to a cloud computing service provider.
“Now, in this case, the Dynamically server would have to be in a different data center,” Poon told The Hill.
“And because we have two Dynamics cores, we can use that to replicate a single data center in the future.”
The researchers were able to simulate using multiple Dynamics instances, but they were only able to move data between them at a limited level of redundancy.
“At the very, very high level, you’d only be able really move data to two Dynamic cores,” Pounse said.
At the very low level, it would be possible to move multiple Dynamic instances, with each Dynamic instance being able to transfer data between two Dynamical cores.
“It’s really about being able move data in a consistent way across multiple cores, which makes the process easier,” Pone said.
Dynamics is a spin-off of IBM’s