Cloud computing technical notes


This purpose of this page is to present you with important technical details of cloud computing.


  • The material on this page can be incredibly useful but out of context may appear a bit ‘so what?’; we suggest skimming through here to help build your cloud context web if these topics are new to you.
  • Our glossary may help with unfamiliar terms.
  • Be warned! We have placed – where appropriate – smaller “Warnings” sections below with each sub-topic.

Virtual Machines

How fast is a given instance? You can read about instance types on the vendor’s cloud console and/or information pages. For example AWS EC2 instances are written up here. Here are a couple ways to look at CPU details from the command line of a Linux machine:

% lscpu
% cat /proc/cpuinfo

It is possible that (from evidence circa 2012) a certain degree of CPU type heterogeneity can introduce on the order of 10% performance variation.

Data transfer to cloud storage


Run the ingress (local to S3) transfer in parallel on a wired connection (not WiFi). Being peered with a high speed netwokr like Internet2 helps; or being in a Science DMZ environment is better still. If the data is going to the Oregon region of AWS: It might not save any time to go from Portland to Seattle and back to Portland again. Finally: You can use the console (browser) but the risk is that this will be slow. On the other hand you can use the AWS command line; e.g. the cp command. There are GUI-based interfaces like Cyberduck as well.

  • Note the AWS CLI installation page; this includes making sure the cli is in your path.

Ports, sockets and tunneling

Following a URL such as you may see a colon followed by a number, as in: This following number is a port, per wikipedia ‘a networking endpoint in an operating system’. A port in this context means a dedicated signal wire. By directing applications to a port you build a dedicated send/receive connection. This is particularly useful in getting two computers to talk to one another.

For completeness there is a related concept called a socket which we do not need in this discussion.

We proceed with a particularly important cloud-oriented use of ports. Suppose you have a cloud VM instance You wish to run some software that uses a Graphical User Interface (GUI). In this case let’s consider the common example of a Jupyter notebook. Here a remote Jupyter notebook server is accustomed to talking to your machine via your local browser. The commands below follow this reference.

Let’s assume you log in to the cloud VM at ip address using ssh.
There you issue

% jupyter notebook --no-browser --port=8889

This is directing the jupyter application to pay attention to port 8889. Next, on your local machine, say a laptop for example, you would issue:

% ssh -N -f -i ~/.ssh/credential_filename.pem -L localhost:7005:localhost:8889 username@

This uses the ssh command. If you are using a Windows PC it is common to connect using PuTTY (see below). The point of the command is to create a tunnel from port 7005 on your local machine to port 8889 on the cloud VM.

The -i switch specifies an identity file (key pair access to the cloud VM).

The -L switch in the command stands for Local.

Double hop into a private subnet

(kilroy this migrates to HIPAA)

Suppose we want to work on an EC2 instance on a private subnet in a VPC. On that machine we have three tasks:

  • Make sure any updates are installed
  • Pull in data from an S3 bucket
  • Start a jupyter notebook (with browser disabled).

We assume the files are already in place on the S3 bucket.

From a starting point ‘laptop’ run PuTTY or ssh to reach the bastion server. From the bastion server:

bastion% sudo yum update
bastion% ssh -i ec2_private_keypair.pem ec2-user@

This gets us to the command line on the private subnet EC2 instance. We proceed with the three steps given above.

ec2_private% sudo yum update

<diagnostic print outs>

ec2_private% aws configure

(enter both public and private IAM User access keys; nothing else required)

ec2_private% aws s3 cp s3://s3-bucket-name/filename1 .
ec2_private% aws s3 cp s3://s3-bucket-name/filename2 .
ec2_private% aws s3 cp s3://s3-bucket-name/filename3 .

ec2_private% jupyter notebook --no-browser --port=7777

On our local machine:

laptop% host1=ec2-user@
laptop% host2=ec2-user@
laptop% ssh -i hipaa_bastion_keypair.pem -L 7777:localhost:7777 $host1 ssh -i hipaa_ec2_private_keypair.pem -L 7777:localhost:7777 -N $host2

In our local browser address bar:


This brings up the Jupyter notebook view from the machine on the private subnet.

MySQL connection

As another example concerning MySQL database connections: Here is a paraphrased quote from an online ssh tutorial

SSH can forward communication across an SSH session that you establish. For example, you can set up a port forward for your connection from your home machine to such that connections to localhost port 3306 forward to the remote machine’s port 3306. Port 3306 is the port that the MySQL server listens on, so this would allow you to bypass the normal host checks that the MySQL server would make and allow you to run GUI MySQL programs on your local computer while using the database on your remote machine. Here is the command to accomplish this:

ssh -L

The -L (which means Local port) takes one colon-delimeted argument.


You specify what host and port the connection will go to on the other side of the SSH connection. When you make a connection to the local-host:local-port port it sends the data through the SSH connection to the connect-to-host:connect-to-port port. From the point of view of connect-to-host it is as if the connection came from that machine.

In our example the ssh command for the Jupyter notebook connection stipulated port 7005 on localhost. Hence on your local machine you can now open a browser and enter in the address bar:


If all is well you will see the Jupyter notebook (from the cloud VM) on your browser.

Jupyter notebooks have associated passwords; and in the case above our colleague had to set this password to an empty string in the Jupyter config file due to Jupyter permission default settings.

The main point here is that the ssh protocol can be used to create tunnels into cloud VMs and that this particularly applies when you want to see GUI content on your local machine.


The Windows PC environment does not yet support the ssh command and consequently cloud VM connections typically use the PuTTY application. (There are also alternative options ). PuTTY is a slightly circuitous but perfectly reliable means of connecting with a cloud VM, supporting as it does several communication protocols including ssh.

The main point here is that a VM credential file with a .pem extension does not work with PuTTY. You can use the companion application PuTTYgen to convert this file to an equivalent file with a .ppk file extension. The PuTTY program can then use this credential file to connect to a cloud VM.

kilroy need screencap of PuTTYgen in operation and the menu location for specifying a .ppk file.

kilroy need text and screencap showing use of PuTTY to accomplish the Jupyter tunnel given above; and copy to the all Jupyter pages as well.

Python connections

Python has a Library to help simplify connecting to a URL. The following code is a simple program that demonstrates an API call to our AralDIF project:

import urllib2
import time

def dif(calltype, date1, date2, station):
    baseAPI = ''
    apiExtension = calltype + '?' + 'start=' + date1 + '&' + 'end=' + date2 + '&' + 'station=' + station + '&'
    api_call = baseAPI + apiExtension
    print api_call
    t1 = time.time()
    u = urllib2.urlopen(api_call)
    t2 = time.time()
    data =
    t3 = time.time()
    print 'URL open required', t2 - t1, 'seconds; read() required', t3 - t2, 'seconds'
    return data

data = dif('gethydrograph', '1995-01-01', '2000-12-31', 'UCHKU')
print data[0:198]

The output of this program (the first 198 characters returned) is as follows:
URL open required 9.09429216385 seconds; read() required 0.0734670162201 seconds

Note that establishing the connection required about 9 seconds and the data transfer less than one tenth of a second.

The importance of being able to write this code is high and can be understood as follows:

If your research requires you to regularly go to an external data resource to make new data queries (updates, new ideas for looking at existing data and so on) then you traditionally carry this out by manual methods: Lots of clicking and dragging and waiting. The above Web Client demonstrates that this process can be automated; provided that the person maintaining the data service (in our example ‘’) has built and maintained their side of the automation contract.


This section describes using exclam in front of a Linux command from a Jupyter frame.

Linux connections


  • ** Using curl requires a proper URL and – when hitting GitHub – this can be tricky to figure out. Please see our dedicated GitHub page and the Azure Jupyter page for more deatails. **

This section describes ssh / ftp / wget / curl stuff





Per Wikipedia, curl is a command line tool for getting or sending files using URL syntax. See this link for Curl usage.

curl can be used in the context of an Azure-hosted Jupyter notebook to load a particular data resource (e.g. from GitHub) as an initialization step every time the notebook is restarted.