Intel Computer Systems Cluster (the "fish machines")

You will be doing your 15-213 lab assignments on a cluster of Nocona Xeon servers, donated by Intel, called the "fish machines". Intel engineers traditionally use the names of North American rivers as internal names for their processor projects. So it seems fitting that we, as denizens of the Intel cluster, name our machines after fish of North America. The fish machines run Linux, are rack-mounted in the Wean Hall 3rd floor machine room, and are administered by the CS facilities group. They reboot every day at 7am.

Here is the current fish machine status.

Fish machines available to students:

flounder.ics.cs.cmu.edu	grouper.ics.cs.cmu.edu	kingfish.ics.cs.cmu.edu	mackerel.ics.cs.cmu.edu
tuna.ics.cs.cmu.edu	pompano.ics.cs.cmu.edu	sailfish.ics.cs.cmu.edu	seabass.ics.cs.cmu.edu
seatrout.ics.cs.cmu.edu	swordfish.ics.cs.cmu.edu	tarpon.ics.cs.cmu.edu

Fish machines available to teaching staff only:

bluefish.ics.cs

bonito.ics.cs (Autolab server)

cobia.ics.cs

Quick Start Login

Suppose your user name is bovik. First, login to an Andrew Linux or Unix machine:

ssh bovik@linux.andrew.cmu.edu

Then, from your Andrew home directory, type:

unix> /afs/cs/academic/class/15213-s10/bin/checkin (very first time only)
unix> ssh -x -l bovik@ANDREW.CMU.EDU fish.ics.cs.cmu.edu

Note: If prompted for a password, enter your Andrew password.
Note :The uppercase 'ANDREW.CMU.EDU' is significant.
Note: The string "-l" is dash el, not dash one.
Note: Your top level Andrew home directory needs to be at least world listable: "system:anyuser l".
Note: By using host name "fish", you will be directed at random to one of the course machings.

Getting Help

Information about the CS computing environment is at www.cs.cmu.edu/~help.
If you are having trouble logging in, please send mail to your instructor (gkesden@cs.cmu.edu).

Frequently Asked Questions

Q: How do I get an account?
A: Accounts will be created for you automatically.

Q: How do I know that I'm really logging into the fish machines?
A: Your connection to our servers is authenticated with RSA Public Key Encrpytion. All the cluster machines share a RSA fingerprint of 60:2e:0d:8b:fd:ee:a7:7e:42:b5:9a:90:76:0f:f2:90 . If any other fingerprint is provided to you, you should stop immediately and contact the course staff.

Q: What do I need to do before logging in for the very first time?
A: From your Andrew home directory on an Andrew Unix machine (e.g., linux.andrew.cmu.edu or unix.andrew.cmu.edu), run the following one-time checkin script:

    unix> /afs/cs/academic/class/15213-s10/bin/checkin

Your top level Andrew home directory needs to be at least world listable: "system:anyuser l".

Q: What does the checkin script do?
A: The checkin script activates your account so that you can login to the fish machines using your Andrew password. It creates a hidden directory called ~/.15213 with login credentials for the fish machines. If you don't have one already, it also creates a protected directory ~/213hw where you can safely do your assignments without other students being able to see them. You only need to run the checkin script once, before your very first login to the fish machines. However, you can safely run the checkin script as often as you like.

Q: How do I login to one of these machines once I have run the checkin script?
A: If your Andrew login is bovik and you want to login to a machine, then type the following while logged in to an Andrew Unix machine:

    unix> ssh -x -l bovik@ANDREW.CMU.EDU  fish.ics.cs.cmu.edu
    [type your Andrew password to the prompt, if prompted]

Note: The uppercase 'ANDREW.CMU.EDU' is significant.
Note: "-l" is "dash el".
Note: The "-x" option disables X11 forwarding, which sometimes interacts poorly with the new 64-bit FC3 boxes. It's not always necessary, but we suggest using it just to be sure.

Q: I did everything you said but I still can't login. Now what?
A: Here are the most common reasons students can't login:

You're not logging in from an Andrew Linux or Unix machine.
FIX: ssh to linux.andrew and login from there.
You're logging in as user or user@andrew.cmu.edu instead of user@ANDREW.CMU.EDU
FIX: Run ssh using "-l user@ANDREW.CMU.EDU"
You forgot to type -l ("dash el") before your user name on the ssh command line.
FIX: Run ssh using "-l user@ANDREW.CMU.EDU"

Your Andrew ~ (home) or ~/.15213 directories are not accessible.
FIX: Make them accessible:

    unix> cd 
    unix> fs sa -dir ~ -acl system:anyuser l
    unix> fs sa -dir ~/.15213 -acl system:anyuser rl

You've somehow deleted the .login and .klogin files in your ~/.15213 directory.
FIX: Rerun the checkin script.
You've changed your Andrew password after you ran the checkin script.
FIX: Logout from the Andrew machine, log back in, and try again.

If you still can't login, please send mail to your instructor.

Technical specs

Each node on the cluster runs a 64-bit version of Fedora Core 3 (Linux kernel 2.6.11) and consists of the following hardware:

3.2 GHz Dell PowerEdge 1850
Dual 3.2 Ghz 64-bit (EM64T) Nocona Xeon processors (the processors don't share caches)
Cache block size: 64 B
L1 cache: 16 KB, 8-way associative, write-through
L2 cache: 1 MB, 8-way associative, write-back
800 MHz frontside bus
Intel E7520 (Lindenhurst) chipset
2 GB, 400MHz DDR2 SDRAM memory
146 GB Ultra320 SCSI hard disk
Embedded ATI Radeon 7000-M graphics card, w/16 MB SDRAM
1U rack height

48 bits virtual address
36 bits physical address
4 KB page size
4-level page tables with 512 entries per table
Data TLB with 64 entries, fully associative
Instruction TLB with 128 entries, fully associative
L1 instruction cache: 16KB, 8-way set associative, 64 byte blocks
1 cycle access time for cache hit
L1 data cache: 16KB, 8-way set associative, 64 byte blocks
1 cycle access time for cache hit
4 cycles latency for L1 cache hit memory access
L2 unified cache: 1MB, 8-way set associative, 64 byte blocks
~10 cycles access time for L2 cache hit
Main memory: 2GB
~50-100 cycles access time
146 GB hard disk