When compared with software, we are not that forward in terms of hardware. Amazon has recently introduced EC2 F1, the computing instance with Field Programmable Gate Arrays (FPGAs).
An FPGA is just like a black box - it consists of "inlets" and "outlets". The FPGA integrated circuit can be programmed, strictly speaking "configured", by the customer or a designer, based upon the specifications. The Amazon EC2 F1 is the architecture based on FPGAs that replicates a given hardware for performance acceleration of an application. The F1 instance are handy to program and is available with all the tools you need to develop, simulate, debug, and compile a given hardware acceleration code. It also comes with FPGA developer AMI and Hardware Developer Kit (HDK), that helps the developer make better and efficient hardware.
Using the F1 instance is easy - actually, it holds only three steps.
The advantage of EC2 F1 lies in reusability of the AFI. It can be reused as many times, across many F1 instances as you like. It is capable of holding the same performance.
Amazon has launched the F1 instances in two instance sizes - F1.2xlarge and F1.16xlarge, that include up to eight FPGAs per instance. The F1 instances are fabricated with 16nm Xilinx UltraScale Plus FPGA. In terms of logic elements, each of these FPGAs are designed
to hold 2.5 million logic elements. The hardware capabilities of the F1 instance include 64GB DDR4 ECC protected memory, with a dedicated PCIe x16 connection and 6,800 DSP engines. Acquiring on-demand instances for F1 is same as that of the other EC2 instances. However on-demand, you need not change the AMI; it still can be reused.
Benefits of Amazon EC2 F1 implementation:
Quick deployment of custom hardware accelerations. As mentioned earlier, it is easier to deploy the images onto the F1 instances. Also, the F1 instances can have one or more AFIs associated with them. This helps you to run multiple hardware accelerations on the same instance without degrading the performance of the hardware acceleration code.
Predictable performance. All the FPGAs available on the F1 instance can communicate at a data rate of 12GBps, all of them connected through a dedicated PCI express fabric. This helps the instance to deliver consistent performance, throughout.
Changing the economics of FPGAs. The Amazon EC2 F1 eliminates the expenses for specialised hardware, as F1 can run several instances at a time. Also, this reduces the cost for deployment.
Using the existing FPGA algorithms. Though the approach for customised hardware has changed, there will not be much change in the existing FPGA algorithms. They can be used as is, with minor modifications.
Features of Amazon EC2 F1:
Applications of Amazon EC2 F1:
1. Genomics Research
2. Financial Analytics
3. Real-time video processing
4. Big Data Search and Analytics
5. Security
Using the F1 instance is easy - actually, it holds only three steps.
- Design and Development of FPGA, based upon the specifications
- Registering it as an Amazon FPGA Image (AFI)
- Deployment of the image onto the F1 instance
The advantage of EC2 F1 lies in reusability of the AFI. It can be reused as many times, across many F1 instances as you like. It is capable of holding the same performance.
Amazon has launched the F1 instances in two instance sizes - F1.2xlarge and F1.16xlarge, that include up to eight FPGAs per instance. The F1 instances are fabricated with 16nm Xilinx UltraScale Plus FPGA. In terms of logic elements, each of these FPGAs are designed
to hold 2.5 million logic elements. The hardware capabilities of the F1 instance include 64GB DDR4 ECC protected memory, with a dedicated PCIe x16 connection and 6,800 DSP engines. Acquiring on-demand instances for F1 is same as that of the other EC2 instances. However on-demand, you need not change the AMI; it still can be reused.
 |
| Amazon EC2 F1 instance specifications |
Quick deployment of custom hardware accelerations. As mentioned earlier, it is easier to deploy the images onto the F1 instances. Also, the F1 instances can have one or more AFIs associated with them. This helps you to run multiple hardware accelerations on the same instance without degrading the performance of the hardware acceleration code.
Predictable performance. All the FPGAs available on the F1 instance can communicate at a data rate of 12GBps, all of them connected through a dedicated PCI express fabric. This helps the instance to deliver consistent performance, throughout.
Changing the economics of FPGAs. The Amazon EC2 F1 eliminates the expenses for specialised hardware, as F1 can run several instances at a time. Also, this reduces the cost for deployment.
Using the existing FPGA algorithms. Though the approach for customised hardware has changed, there will not be much change in the existing FPGA algorithms. They can be used as is, with minor modifications.
Features of Amazon EC2 F1:
 |
| Programmable Hardware Accelerations |
 |
| Dedicated FPGAs |
 |
| Reusable Amazon FPGA Image |
 |
| Hardware Development Kit |
 |
| FPGA Developer AMI |
1. Genomics Research
2. Financial Analytics
3. Real-time video processing
4. Big Data Search and Analytics
5. Security
0 comments:
Post a Comment