Cortex: On-Demand Compute
Cortex is a service that provides dedicated Linux and Windows virtual servers with root access. There were errors rendering macro:
|
About Cortex
The Zuckerman Institute Research Computing Team has recently launched Cortex, an on-demand computing service that provides easy-to use private servers for labs and researchers. The servers can either be Linux- or Windows-based, and can directly connect to our Engram data storage system. These servers give you access to powerful computational capability (in a flexible and cost-effective way) without having to invest in self-managed hardware.
If your requirements change in the future, you can submit a simple request to modify the number of CPU cores, the amount of memory (RAM) committed, and/or the size of the scratch space on your server. You would have full root access and the Research Computing team can help with the installation of any software you need. Because the hardware is located on-site and managed by us, all data and code on the servers are automatically backed up and secure.
We can help you decide what type of server would be best for your needs. Email us at rc@zi.columbia.edu.
Cortex can be accessed remotely by using the CUIT Virtual Private Network (VPN) service. (More VPN setup details provided here (Limited access UNI login required))
Cortex Details
Cortex is an on-demand compute platform that provides private servers. It has the following features:
- Located in the Zuckerman Data Center (ZDC)
- Backed up and professionally managed by the Research Computing Team
- Connect your Engram Data Storage to your Linux/Windows server
- Servers are connected to Engram with dual 10 Gb fiber network
- Have your own, private server with root access
- No need to wait for your jobs to get scheduled or share your server with anyone else
- Use it for days, weeks, months or years at a time
- Use it for any purpose: compute/analytics, running Docker containers, developing new software, as a web server
- Linux supported: CentOS, Ubuntu and Debian
- Linux servers come with a graphical user interface (VNC)
- Detach and re-attach your connection/session
- Your jobs and processes keep running
- Connect to your server from Columbia campuses or anywhere in the world (with VPN)
- Easily change your server size:
- Add/reduce the number of CPU cores
- Add/reduce the amount of memory
- Add/reduce the size of disk
Types of Servers and Rates
There are three server sizes to fit your needs. In addition, we can create a custom server for the exact size you need.
Billing occurs every 3 months using a supplied chart string. Rates are provided in the table below.
The setup fee includes provisioning your server and one week of use.
Minimum server lifetime is one week. In this case only the setup fee applies (no additional charge per day).
If you do not need your server for a long period of time, you can request we put it in hibernation mode:
- Your server will continue to exist but will not be available
- When you need to use it again, we will wake it up
- There is no minimum amount of time your server can hibernate
- There is no additional charge for putting your server in hibernation mode or waking it up
Server Sizes
Size | CPU Cores | Memory (GB) | Disk (TB) | Rate per Day |
---|---|---|---|---|
Small | 1 | 2 | 0.25 | $0.20 |
Medium | 4 | 8 | 0.5 | $0.56 |
Large | 8 | 16 | 0.5 | $0.89 |
Custom | $0.04/core/day | $0.02/GB/day | $0.49/TB/day | TBD |
Example Use Case
You need a Large Server to compute on your data for three weeks.
- Order a server and attach your Engram to it
- Compute and analyze your data
- Total cost: $18.06
- Calculation: $12.46 (14 days @ $0.89/day)
- Request we put your server in hibernation mode
- Your server still exists but is not available
- You do not get charged
- Four months later, you need your server again
- Request we wake it up
- Your server comes back online in the same state as you left it
- You don't have to set up your server environment again, or install software
- No cost for hibernation and waking up the server
- When you don't need your server any longer, we can delete it
There were errors rendering macro:
- An unknown error occurred.
Accessing your VM
Your VM can be accessed from JLG or via the CUIT Virtual Private Network (VPN) service. (More VPN setup details provided here (Limited access UNI login required)).
Linux
To access a linux VM, use an SSH agent, i.e. putty, the built-in terminal or iTerm2 for MacOSX. Login with your UNI and UNI's password.
Windows
To access a Windows VM, please use Remote Desktop Connection (from Windows) or any RDP client from MAC, i.e. Microsoft Remote Desktop 10.
Login with username adcu\UNI and UNI's password.
Configuring your VM
Research computing currently offers the following guides for performing common configuration updates on your lab's VM: