Once you have an Amazon AWS account, you can create one or more
"instances" in the cloud. Each instance is a virtual machine running
a standard operating system such as Ubuntu or CentOS. From your local
machine (laptop/desktop), you can connect to your instance and
interact with it using ssh, |IE| the same, usual way that one
interacts with any remote Linux machine.
AWS offers a very large number of services. For our purposes (FPGA development), please select "EC2" ("Elastic Computing"). From the EC2 web page, please select "Instances", which should land you in the EC2 Instances Dashboard, similar to that shown in :numref:`Fig_AWS_EC2_Instances_Dashboard`.
AWS EC2 Instances Dashboard showing one's instances
In the dashboard, the "Instances" panel will list all the instances that you have created. The figure shows two instances called "RSN_AFI_Build2" and "RSN_AFI_Run".
.. index:: AMI
.. index:: Instance
.. index:: Instance; type
.. index:: F1 instance
.. index:: Instance; F1
.. index:: AWS Access Key
From the EC2 Instances Dashboard, please select "Launch Instances" (button near top-right) which will start a "wizard" that takes you through the following steps for defining and creating an instance.
On the "Launch Instances" wizard page, the first section is titled "Name and tags", see :numref:`Fig_AWS_Launch_Instance_010_Name_and_tags`:
"Name and tags" section of Launch Instances screen.
Type a name for your instance into the text box.
On the "Launch Instances" wizard page, the next two sections are titled "Application and OS Images (Amazon Machine Image)", see :numref:`Fig_AWS_Launch_Instance_020_Choose_AMI`, and "Instance type", see :numref:`Fig_AWS_Launch_Instance_030_Instance_Type`:
"Application and OS Images (Amazon Machine Image)" section of Launch Instances screen.
"Instance Type" section of Launch Instances screen.
An "AMI" is a software template from which you create your instance. It is a pre-load of your virtual memory software, in particular an operating system like Ubuntu or CentOS, plus any application software pre-loaded into that operating system, such as Xilinx Vivado software.
An "Instance Type" specifies the type of computer hardware and its configuration, in the AWS cloud, on which your instance should run: x86 or ARM, number of processors, amount of memory and disk, |ETC| Instance Types named "f1.XXXX" are the only ones that have attached FPGAs.
When creating an instance, we choose both an AMI and an Instance Type on which it will run.
When we are developing FPGA hardware and running applications using FPGAs, we can regard our activities in three categories, each of which can be satisfied by a different combination of AMI and Instance Type:
Developing the hardware that will eventually run on the FPGA. This involves creation of Verilog RTL, simulating it and verifying it.
This does not need an attached FPGA. On AWS, almost any AMI and Instance Type with sufficient compute power will do. It can also be performed on a local machine (does not have to be on AWS).
We recommend selecting a standard, free, Ubuntu AMI, and the
m6i.2xlargeorz1d.2xlargeInstance Type.Synthesizing the Verilog RTL into a bitfile for the AWS FPGA.
This is most conveniently performed on an AWS instance based on the AWS "FPGA Developers' AMI", which is pre-loaded with the AWS HDK (Hardware Develpment Kit), Vivado software and Vivado licenses, and scripts needed to create the AWS FPGA bitfile. It does not need an attached FPGA.
We recommend selecting the "FPGA Developers' AMI" and the
m6i.2xlargeorz1d.2xlargeInstance Type.Loading a bitfile into the FPGA to program its functionality, and running software to interact with the FPGA hardware. This, of course, needs an "f1" instance with the attached FPGA. The AMI could be Ubuntu or CentOS; the key requirement for the OS is that should support the XDMA device driver (about which more, later) for communicating with the FPGA.
We recommend selecting a standard, free, Ubuntu AMI, or the "FPGA Developers' AMI" (both support the XDMA driver), and the
f1.2xlargeInstance Type.
To select an AMI, you can search for "Ubuntu" and "FPGA Developers' AMI" in the "Search our full catalog ..." text box. It will offer you a choice like this, for example:
Ubuntu Server 22.04 LTS (HVM), SSD Volume Type ami-052efd3df9dad4825 (64-bit (x86)) / ami-070650c005cce4203 (64-bit (Arm))
Ubuntu Server 22.04 LTS (HVM),EBS General Purpose (SSD) Volume Type. Support available from Canonical (http://www.ubuntu.com/cloud/services).
Make sure the "64-bit (x86)" button is selected (and not the "64-bit (Arm)" button), and click "SELECT". This should now show your selection, as in :numref:`Fig_AWS_Launch_Instance_020_Choose_AMI`.
To select an Instance Type, the selection box will have preselected
the t2.micro Instance Type. Change this by clicking on the caret
at the right end of the box which opens a search box. Enter
f1.2xlarge or m6i.2xlarge and select it from the list.
Why not make just one instance for all our development activity?
One reason is cost. AWS' usage charges vary by instance type, and unfortunately "f1" instances are generally more expensive than other Instance Types. Further, charges are mostly accrued while actually running an instance; "stopped" (suspended) instances have comparatively negligible charges. Thus, for development work (creating RTL, RTL simulation, verification with RTL simulation, host-side software development, |ETC|) we might choose an Instance Type that is not an "f1" instance. We may only start running and use the "f1" instance while we are actually using the FPGA, keeping it stopped at other times.
Another reason is preference---for example, one may prefer the Ubuntu environment over CentOS, or vice versa.
.. index:: Key Pair
On the "Launch Instances" wizard page, the next section is titled "Key pair (login)", see :numref:`Fig_AWS_Launch_Instance_040_Key_Pair`:
"Key Pair" section of Launch Instances screen.
What this is: later, when you use ssh to connect to your
instance, or use scp to copy files in and out of the instance, you
will use this Key Pair to authenticate yourself. AWS uses
cryptographic access keys for authentication rather than the older,
less secure password mechanism.
The screen shows a slot where you can select an existing key pair if you already have one from previous uses of AWS (click on down-caret on the right to select an existing key pair), or you can click on "Create new key pair" to create one.
Follow the prompts to create an access key pair, which is a pair of
rather long ASCII key strings, randomly generated to be
cryptographically secure. One of these is called your "public key"
and the other is your "private key". The prompts will ask you to cut
and paste the private key into a file on your local computer
(laptop/desktop). Typically this file is placed in your ${HOME}/.ssh/
directory with filename extention .pem, like this:
~/.ssh/RSN_AWS.pem
You will refer to this file later when using ssh to connect to
your instance (see :numref:`Sec_connecting_to_an_instance`).
Tip
You can create a new access key at any time; it does not have to be during instance creation. In your EC2 dashboard, clicking on your login id on the top-right will drop down a menu with a "Security Credentials" option. This will take you to a screen where "Access Keys" is one of the topics, with a "Create access key" button to create a new one. It is a good practice to change your access key every so often.
On the "Launch Instances" wizard page, the next section is titled "Network settings", see :numref:`Fig_AWS_Launch_Instance_050_Network_Settings`:
"Network Settings" section of Launch Instances screen.
We recommend leaving everything here unchanged. In particular:
- "Auto-assign public IP" = "Enable"
- "Allow SSH traffic from Anywhere"
Note
You can update the network settings later if you wish/need to.
On the "Launch Instances" wizard page, the next section is titled "Configure Storage", see :numref:`Fig_AWS_Launch_Instance_060_Configure_Storage`:
"Configure Storage" section of Launch Instances screen.
We recommend leaving everything here unchanged.
Note
You can update the storage settings later if you wish/need to.
On the "Launch Instances" wizard page, the next section is titled "Advanced details". We recommend leaving everything here unchanged.
"Summary" is a floating panel on the right-hand side of "Launch Instances" wizard page, |IE| it stays visible as you scroll the page, see :numref:`Fig_AWS_Launch_Instance_070_Summary`:
"Summary" section of Launch Instances screen.
Click the "Launch instance" button on the "Summary" panel. It will flash up a "Launching Instance" panel and shortly transition to "Success", see :numref:`Fig_AWS_Launch_Instance_080_Success`:
"Success" on the "Launch an Instance" screen.
Important
After clicking "Launch instance", AWS may present a failure message like this:
You have requested more vCPU capacity than your current vCPUlimit of 0 allows for the instance bucket that the specifiedinstance type belongs to. Please visithttp://aws.amazon.com/contact-us/ec2-request to request anadjustment to this limit.
If you get this message, please visit the contact-us link
shown, click on the "Support" button and lodge a request to
increase your "vCPU quota" to 8, which is required for the
f1.2xlarge instance type. Try to include a few sentences
explaining your need for an f1.2xlarge instance type.
You will get an almost instance response saying:
This specific limit increase request requires further internal review before approval ...
These requests are processed by humans, so it may take a few hours before you receive an email indicating that your request has been approved, after which you can proceed.
Click the "View all instances" button on the "Launch an instance: Success" screen. This will return you to the EC2 Instances dashboard, see :numref:`Fig_AWS_Launch_Instance_090_Instances_Initializing`, for exaample:
"EC2 Instances" screen, showing an instance initializing.
Your new instance should be listed there, with the name you supplied. The "Instance state" column will initially say "Stopped". When you start the instance (see next section) it will say "Pending", then "Running". The "Status check" column will say "initializing", then "2/2 checks passed". See :numref:`Fig_AWS_Launch_Instance_100_Instances_Running`:
"EC2 Instances" screen, showing an instance running.
All this may take about a minute; you may wish to press the refresh icon |ICON_REFRESH| to update the dashboard display.
Congratulations! You have just created your AMI for FPGA development!