michaelslab

My Server cluster called "Alicante" is intended as and run time environment and test environment for my computations like the many particle problem.

The cluster comprises of four machines:

  • thunder:  HP DL 140 8 Cores / 16 GB Memory + 2 x 1 TB disks
  • cyclope: 2 x Dual core / 4 GB Memor + 1 TB Raid 5
  • monster: HP NAS 2000 with a storage array for 6x300GB disks
  • askar: 2x Atom + 50GB SD RAM

The architecture of the cluster is shown in the picture below. Cyclope and thunder are operating in a cluster using pacemaker. Both machines are sharing a 500 GB gluster file system (/global). The head node of the cluster is askar.

All machines of the cluster are connected to a private GB LAN and the public LAN of my home network. The private LAN is used by pacemaker and the gluster fs.

The host monster is providing storage capacity to both computing nodes via 2GB fiber channel connections.

 

 

Logging into the cluster

Even thought each machiene has its local file system users of the cluster are created on the /global file system which is shared among all computing nodes in the cluster.