01-accessing.Rmd

---
title: "Accessing the server clusters"
output: html_document
---

# Accessing the server clusters

You will naturally be able to use the server clusters as a University member. More tutorial could be found at <https://www.wiki.ed.ac.uk/display/ResearchServices/GPUs>. The job type selection should probably be Serial or Parallel, but select the option that best describes your needs. In the job type description you should describe any relevant information related to user groups, access to specific folders, or special software usage. If you need Stata, then you should request to be added to the `econ` group, which holds the license to use Stata. The system to be used should be Pod and Knot, which are the main Linux-based server clusters. Pod is the newest cluster and offers greater capabilities, but Knot is still adequate for most analysis needs and typically has less active users (i.e., a shorter job queue).

## On campus

The clusters can be accessed on campus via a terminal emulator using the `ssh` command or via a GUI SSH app, such as [X2Go](https://wiki.x2go.org/doku.php). On macOS or Linux the default terminal apps can be used. On Windows the [PuTTY](https://www.putty.org/) or X2Go apps can be used. Once a terminal is started, connect to Knot by entering the command

```
$ ssh user@eddie.ecdf.ed.ac.uk
```

where `user` is your server username. You will then be prompted for your password. SSH keys can also be set up to skip password entry.

To use GUI applications on the server, a slightly different process is needed in order to enter an XWindows environment. On macOS, [XQuartz](https://www.xquartz.org/) must be installed. On Windows, use the X2Go app to connect to the server. More info on X2Go can be found at <http://csc.cnsi.ucsb.edu/docs/using-x2go-gui-login-knot-or-pod>. To use GUIs, add the `-X` option to the `ssh` command, such as

```
$ ssh -X user@eddie.ecdf.ed.ac.uk
```

Once connected, you will interact with the Linux server using the command line interface. The default working directory is your home folder `/home/user`. Note that this is the login node, but all analyses should be run on compute nodes. These analyses should be submitted as jobs to the queue (more about this [later](#submitting-jobs)), which allocates compute nodes among users.

To disconnect from the cluster, use the command `exit`.

## Off campus

The clusters can be accessed off campus using the VPN service for secure remote access. Instructions for installing the VPN app can be found at <https://www.ed.ac.uk/information-services/computing/desktop-personal/vpn>. Once the forticlient app is installed, open it and select connect for Remote Access. You will then be prompted to enter your userid and the associated password. Once connected, indicated by a green checkmark, you can then access a cluster as you normally would on campus by using a terminal or GUI SSH app.

### More about VPN

Note that forticlient VPN could be installed on Linux machine. More details could be found here: