**5.7 AliEn Shell - aliensh**

24 Will-be-set-by-IN-TECH

processing chain is extensively monitored so a user can any time get the information on the

As mentioned before, there are several AliEn services running at each ALICE site: CE, ClusterMonitor, PackMan and MonALISA. These services are running on a dedicated

The AliEn site CE is usually associated with the local batch system. It is periodically submitting testing pilot jobs (Job Agents) to the local WLCG CE or an appropriate external Resource Broker or WMS. The role of the Job Agents is to verify the local hardware and software capacities at the site. After the usual matchmaking procedure, the JA is sent, through the site CE, into the local batch queue and then to a local Worker Node (WN). After its startup, the JA performs its task and in the case of a positive checkup, the JA requests a "real" job from the central Task Queue via the AliEn Job Broker, or dies otherwise. The PackMan automates the process of installation, upgrades, configuration and removal of the ALICE software packages from the shared software area on the site. It also advertises known/installed packages. The packages are installed on demand, when requested by a Job Agent running on a Worker Node or during the central software deployment over the Grid sites. If a package is not already installed the PackMan would install it along with its dependencies and return a string with commands that client has to execute to configure the package and all its dependencies. The PackMan manages the local disk cache and cleans it, when it needs more space to install newer packages. The Cluster Monitor handles communication with the AliEn Job Broker and gives configuration to JAs. It gets "heartbeats" from the JAs. If it gets no heartbeats from a JA, the existing job will get into the ZOMBIE status (after 1.5 hours) and then it will expire (after

Since the AliEn Workload Management does not depend directly on sophisticated monitoring, no special monitoring tools were developed in AliEn. As the monitoring solution, ALICE has adopted and further developed the Java-based MonALISA framework [44] mentioned already in the previous section. The MonALISA system is designed as an ensemble of autonomous multithreaded, self-describing agent-based subsystems which are registered as

The collected monitoring information is published via Web Service for use by AliEn Optimizers or for visualization purposes. An extension of the network simulation code which is a part of MonALISA can provide a tool for optimization and understanding of the

Experience with the performance of different types of storage managers shows that the most advanced storage solution is the native XRootD manager [37] described in section 3. It has been demonstrated that with all other parameters being equal (protocol access speed and security) the native XRootD storage clusters exhibit substantially higher stability and availability. The ALICE distributed system of native XRootD clusters is orchestrated by the

dynamic services, and together can collect and process large amounts of information.

status of his/her jobs.

machine, so-called VOBOX described in section 3.

**5.4 Site services**

3 hours).

**5.5 Monitoring**

**5.6 Storage**

performance of the AliEn Grid system.

To complete the brief description of AliEn, we mention the client called AliEn shell. It provides a UNIX-shell-like environment with an extensive set of commands which can be used to access AliEn Grid computing resources and the AliEn virtual file system. There are three categories of commands: informative and convenience commands, File Catalogue and Data Management commands and TaskQueue/Job Management commands. The AliEn shell has been created about 4 years ago and become a popular tool among the users for job handling and monitoring.
