Frequently Asked Questions

To find the most relevant information, select one of the following:

• I am a user that wants to run an application on PlanetLab.

• I am the designated Principal Investigator (PI) for my site.

• I am a system administrator or the Technical Contact for a PlanetLab site.

• I am interested in getting involved in the day-to-day support of PlanetLab.

• I am interested in running a "private" PlanetLab on my own facility.

PlanetLab User

Individuals are not allowed to directly join PlanetLab. Your home institution must be a member of the PlanetLab Consortium. To see if your institution is already on PlanetLab, click here. This page also identifies your site's Principal Investigator (PI), who is the person that approves PlanetLab accounts at your institution.

My institution is already on PlanetLab. How do I get an account?

The first step is to click on the "Create Account" menu item on the right side of this page. After reading the Acceptable Use Policy (AUP) and filling out the registration form, you need to ask your site PI to approve your request. Once that has happened, you can log onto your PlanetLab account.

Note that as part of the registration process you are asked to provide a public key. (Follow the "What is this?" link to learn more about how to do this.) Your public/private key pair will be used when you log into your PlanetLab slice using SSH, but that slice is created through another process. (See the answer to the next question.)

I have a PlanetLab account. How do I create a slice for my application to run in?

Creating a slice and assigning users to it requires action on the part of your site PI. (Information on how to do this is given in the next section of this FAQ.) Once this has been done, you can manipulate the parameters of your slice by first logging into your account on this web server and selecting one of the operations under the "Slices" menu item.

I have slice. How do I access it?

First, realize that your slice is not actually instantiated on a set of PlanetLab nodes until you invoke the "Manage Nodes" operation available on the slice management page. Once this has been done, and the operation has had a chance to complete (this should happen within 15-30 minutes) you can SSH into any of the nodes assigned to the slice using the keys you supplied when you created your account.

I have a working slice, but I cannot SSH into some of the nodes I assigned to the slice. Why?

There are several possible explanations. One is that it is a rare event that 100% of a set of hundreds of nodes distributed all over the planet are accessible at any one time. The applications you build will need to robustly deal with this situation. Another possibility is that you can actually SSH into the node, but the node is not reachable using ping. This is because some sites filter pings. Don't assume that not being able to ping a node means that you cannot SSH into it.

If a problem with a particular node persists, or if you see an unusually large number of unreachable nodes, please send email to PlanetLab Support. Be sure to identify the nodes in question by name (rather than by IP address).

I logged onto the nodes in my slice, but several packages that I depend on (e.g., gcc, python, java) are not available. Why?

This is by design. Every user wants a different set of packages, or different versions of the same package. Your slice is created with a near-empty file system. You need to load whatever packages (RPMs) you depend on, using for example, the YUM package manager or the Stork environment service.

Note that several other users have already worked through the same situation you have just encountered, and have built tools to help them manage their slices. See the Services page for a list of that the user community has make available for others to use.

In general, you should post any questions you have about using your slice to the PlanetLab User's Mailing List. Likewise, if you know the answer to another user's question, please post a response.

Regarding compilers, in particular, you should not use PlanetLab nodes to compile your software since cycles on those machines are a scarce resource. Also, you should not leave the only copy of your source code on PlanetLab since nodes can have their disks wiped at any time.

I have developed a tool or service that others might find useful. How do I make it available?

Wonderful! The first thing you should do is send an announcement to the PlanetLab User's Mailing List telling other users how great your tool or service is. Then contact the PlanetLab staff and ask to include a link to a web page describing your service to this web site.

I'm trying to implement a service, but need a feature from the underlying OS that is not available. How can I get the missing capability added?

PlanetLab's architecture is constantly evolving. Everyone is welcome to contribute to this process. The first step is to start a discussion on the PlanetLab Architecture Mailing List. Describe the capability you need and discuss the various options for providing it. It's possible that this topic has come up before, and you'll find a discusson in the mail archive. Another possibility is that the issue has already been decided, in which case you should be able to find a PlanetLab Design Note (PDN) that explains how the question was resolved.

Principal Investigator (PI)

You need to do two things to join PlanetLab: (1) sign the consortium membership agreement, and (2) connect two or more nodes to PlanetLab.

What role do I play as PI?

Your main responsibility is to vouch for the users at your site that use PlanetLab, and locally enforce the Acceptable Use Policy (AUP). Specifically, you approve their PlanetLab accounts and assign them to your site's slices. You do this through a set of account management pages available once you log in.

That's a lot of work. How to I appoint a "deputy PI"?

Send email to PlanetLab Support asking that another user at your site be given PI privileges.

I'm at an academic institution. Do I need to sign the Consortium membership agreement?

Yes. Academic institutions do not pay a membership fee, but someone with signature authority at your University must sign the membership agreement. This is needed to bind you to the AUP and Intellectual Property (IP) terms of the Consortium.

I'm at a for-profit company. Why do we need to pay membership dues?

Your dues help pay salaries of the PlanetLab staff, who maintain and enhance the PlanetLab infrastructure. We also pay bandwidth charges and have capital expenses. Academic institutions do not pay dues, but in effect their "share" of the costs are being picked up by the National Science Foundation, which contributes over a $1M a year to PlanetLab.

Getting someone at my institution with signature authority to sign a 4-page membership agreement is a non-trivial matter. Why is it so long?

That's what you get when you have lawyers representing two major corporations and three Universities (two public and one private) try to agree to something. If you boil it all down, the membership agreement does two things. First, it says that each site is responsible for its own users, and that everyone agrees to follow the AUP. Second, it sets out the IP terms of the Consortium, which can be summarized as follows: (1) you are free to run proprietary software in your slice without obligation to show it to anyone else; (2) software written by the PlanetLab support team will be made available under a BSD-like license; and (3) anyone is free to contribute software to PlanetLab, but the Consortium isn't likely to accept the contribution unless we can make it available to everyone else under BSD-like terms.

[Note: the above description of the membership agreement is a layman's interpretation. It does not serve as a substitute for the official document.]

Technical Contact

The best way to understand your duties is to read the Technical Contact's Guide, which is available by following the "Guides" link in the menu bar.

Access the node using your web browser. From this page, you will see what slices are running on this node and how much data they have sent/received recently. If you are worried about a particular computer being attacked or probed from this node, you can query this page by specifying the IP address and port number of the computer in question, and learning what slice sent the offending packet(s). You should feel free send email to the research group identified as being responsible for this slice asking if the behavior you saw is to be expected.

A local PlanetLab node is well beyond suspicious behavior; it's bringing our campus network to it's knees. What do we do?

Of course you can always turn the machine(s) off, but if you do, then the PlanetLab staff will not be able to figure out what went wrong. Instead, we request that you bring each suspicious machine into a Safeboot (or Diagnose) state. This will terminate all activity on the node (killing the source of the problem), but leave the machine in a safe state that will allow the PlanetLab staff to log onto the machine and diagnose the problem. You should also send email to PlanetLab Support with a report of the problem. The Technical Contact's Guide provides a specific description of how to do this.

The PlanetLab nodes we host are using a significant fraction of our site's external bandwidth. Can we do anything to limit bandwidth usage?

Yes. Log into the node(s) using the site_admin account and set the upper bound on bandwidth consumption using the pl-limitbw command, as described in the Technical Contact's Guide.

My site is very sensitive to security issues. What do you do to ensure that the PlanetLab nodes we host aren't compromised.

Several things, many of which are enumerated in the Hosting a Node document. However, it is not possible for us to run conventional intrusion detection software because our users' programs often exhibit unusual (but perfectly legitimate) behavior that trigger IDS alarms. The main thing you should know is that bad things do occasionally happen, but when they do, we are able to bring all the nodes world-wide into a safe state in a matter of minutes.

PlanetLab Support

Subscribe to the Support mailing list. To do this, and learn the ground rules for contributing to PlanetLab support, follow the "Mailing Lists" link in the menu bar.

Private PlanetLab

We distribute a stand-alone version of PlanetLab, called MyPLC. You can download and install it on your own hardware infrastructure. The best way to learn about -- and follow the ongoing development of -- MyPLC is to join the "developers" mailing list (follow the "Mailing Lists" link on the NAV bar).

Where can I find the MyPLC source code?

All PlanetLab source code is available via SVN.

Can my private PlanetLab federate with the public PlanetLab?

Yes. Send email to PlanetLab Info if you are interested in peering with PlanetLab.