Dilawaer Duolikun

Power Consumption and Computation Models of a Server with a Multi-core CPU and Experiments

The power consumption of servers has to be reduced in a cluster to realize eco society. In this paper, we discuss power consumption models of servers. We take a macro level approach to reducing the total power consumption of servers to perform application processes. A server is equipped with a multi-core CPU. Through measuring electric power consumed by types of servers to perform application processes, we newly propose a multi-level power consumption (MLPC) model of a server with a multi-core CPU. Here, the power consumption of a server depends on the number of active cores and active threads where at least one application process is performed. We also discuss a computation model which gives the expected execution time of a process on a server. Based on the MLPC model and the computation model, we discuss an energy-aware (EA) selection algorithm to select a server for each process requested by a client in a cluster so that the total electric energy consumption can be reduced. We evaluate the EA algorithm and show the total energy consumption is reduced in the EA algorithm compared with round-robin and random algorithms.

Multi-level Computation and Power Consumption Models

It is critical to reduce the electric power consumed by servers in a cluster in order to realize eco society. In the multi-level power consumption (MLPC) model of a server with a multi-core CPU, the power consumption of the server depends on the number of active cores and active threads where at least one application process is performed. In our previous studies, we discuss the energy-aware (EA) selection algorithm to select a server for each request process. Here, a server which is expected to consume the minimum electric energy is selected in a cluster. A server consumes the basic electric power even if no process is performed. The ratio of the basic energy consumption to the total electric energy consumption is large, e.g. 40 to 50 %. In this paper, we newly propose a globally energy-aware (GEA) algorithm to select a server for each process in a cluster. Here, not only the total electric energy consumption of the servers but also the ratio of basic electric energy consumed by servers to the total energy consumption can be reduced. We evaluate the GEA algorithm and show not only the total energy consumption of the servers but also the average execution time of processes are reduced in the GEA algorithm compared with the EA, round-robin (RR), and random (RD) algorithms.

Evaluation of Energy-Aware Server Selection Algorithms

The electric power consumed by servers has to be reduced in a cluster in order to realize eco society. We take a macro level approach to reducing the total electric energy consumption of servers to perform application processes in a server cluster. Servers are now equipped with multi-core CPUs. In this paper, we discuss a multi-level power consumption (MLPC) model of a server with a multi-core CPU. Here, the power consumption of a server depends on the number of active cores and active threads where at least one application process is performed. We also discuss a multi-level computation (MLC) model which gives the expected execution time of a process which is concurrently performed with other processes on a server with a multi-core CPU. Based on the MLPC model and the MLC model, we discuss an energy-aware (EA) algorithm to select a server for each process requested by a client in a cluster so as to reduce the total electric energy consumption while satisfying deadline requirements of the processes. We evaluate the EA algorithm and show not only the total energy consumption but also the average execution time of each process is reduced in the EA algorithm compared with the round-robin (RR) and random (RD) algorithms.

Power Consumption Models for Migrating Processes in a Server Cluster

Application processes have to be efficiently performed on servers in a cluster with respect to not only performance but also energy consumption. In this paper, we consider a process migration (MG) approach to energy-efficiently performing an application process on servers in a cluster. In this paper, we propose a model to estimate the energy consumption of a server to perform processes. First, a process is initiated on a server named home server in a cluster. A process performed on a current server is migrated to another server if the server is expected to consume a smaller amount of electric energy to perform the process than the current server in the estimation model. A process takes checkpoints and sends the checkpoints to the home server. If a process is faulty, the home server recreates the process on an operational server and the process is restarted on a state saved at a checkpoint most recently taken on the home server. In the evaluation, the total energy consumption of servers is shown to can be smaller in the MG algorithm than the other algorithms.

Energy-Aware Server Selection Algorithms in a Scalable Cluster

It is critical to reduce the electric energy consumed in information systems, especially server clusters. In this paper, we extend the multi-level power consumption (MLPC) model and the multi-level computation (MLC) model to a server with multiple CPUs. In this paper, we newly propose a totally energy-aware (TEA) algorithm to select a server for a process in a cluster. Here, servers in a cluster are first classified into subclusters. Each subcluster is characterized in terms of the electric power and computation rate. One server is randomly selected in each subcluster. Then, one server is selected so that the expected electric energy is minimum in the selected servers. We evaluate the TEA algorithm and show not only the total electric energy consumption of the servers but also the average execution time of processes are reduced in the TEA algorithm compared with other algorithms.

Energy-efficient dynamic clusters of servers

Electric power consumed by servers has to be reduced in order to realize green societies. We consider computation (CP) and storage (ST) types of application processes performed on servers in this paper, where CPU and storage drives are mainly used, respectively. In the storage- and computation-based power consumption model proposed by the authors, the power consumption rate of a server depends on what types of processes are performed but is independent of how may processes are performed on the server. In the storage- and computation-based processing model, the execution time of an ST process depends on the number of concurrent CP and ST processes but the execution time of a CP process depends on only CP processes and is independent of ST processes. In our previous studies, the energy-aware algorithm is discussed to select a server in a cluster of servers for each request so that the total power consumption of the servers can be reduced. However, a server consumes electric power even if the server is idle, i.e. no process is performed. In this paper, we discuss a dynamic energy-aware (DEA) cluster which includes only active servers where at least one process is performed. A server for each request is selected in a dynamic cluster so that the total power consumption of servers in the cluster can be reduced. We evaluate the DEA algorithm in terms of the total power consumption and average execution time and show the total power consumption can be reduced.

Energy-Aware Clusters of Servers for Storage and Computation Applications

It is now critical to reduce electric energy consumed in a cluster of servers, especially scalable systems like cloud computing systems. In clusters, most application processes like web applications use not only CPU resources but also files and databases. In this paper, we consider storage processes which read and write data in files in addition to computation processes. We propose a PCS model (power consumption model for a storage server) which shows how much electric power a server consumes to perform storage and computation processes. We also propose a CS model (a computation model for storage server) which shows how long it is expected to take to perform storage processes and computation processes. By using the PCS and CS models, we propose a local energy-aware (LEA) algorithm to select a server for a request process in a cluster so that the total electric energy consumption of the servers can be reduced. We evaluate the LEA algorithm in terms of total electric energy consumption of the servers. We show the electric energy consumed by servers to perform computation and storage processes can be reduced in the LEA algorithm.

Asynchronous Migration of Process Replicas in a Cluster

Application processes have to be efficiently and reliably performed on servers in a cluster. A process is replicated to increase the reliability and availability. However, the more number of replicas of a process are performed, the more reliable and available the system is but the more amount of electric energy is consumed. In this paper, we take a process migration (MG) approach to energy-efficiently and reliably performing an application process on servers in a cluster. A process performed on a current server is migrated to another server if the server is expected to consume a smaller amount of electric energy to perform the process than the current server while the deadline constraint of the process is satisfied. In order to reliably perform a process, multiple replicas of the process are performed on different servers. We consider synchronous and asynchronous ways to migrate process replicas to servers. In the synchronous migration, every replica moves to another server on a same computation state. In the asynchronous migration, each replica by itself makes a decision on when the replica is migrated to another server.

Energy-Efficient Replication and Migration of Processes in a Cluster

Application processes have to be efficiently performed in presence of server faults in a cluster. Multiple replicas of a process are performed on multiple servers. However, the more number of replicas of a process are performed, the more reliable and available the process can be performed but the more amount of electric energy is consumed. In this paper, we take a process migration (MG) approach to energy efficiently and reliably performing multiple replicas of an application process on servers. A replica of a process performed on a current server st migrates to another server su if the server su is expected to consume a smaller amount of electric energy to perform the replica than the current server st while the deadline constraint of the process is satisfied. We consider synchronous and asynchronous ways for process replicas to migrate to servers. In the synchronous migration, every replica migrates to another server on a same computation state. In the asynchronous migration, each replica makes a decision on when the replica migrates to which server.

Reduction of Unnecessarily Ordered Event Messages in Peer-to-Peer Model of Topic-Based Publish/Subscribe Systems

A distributed system is considered in an event-driven model where a state of a process transits on occurrence of an event. In this paper, we discuss a peer-to-peer topic-based publish/subscribe (P2PPS) model where each peer process (peer) can both subscribe and publish event messages. The subscriptions and publications are specified in terms of topics. Each event message e carries a vector e:TV = (tv1, , tvh) of topics t1, , th in a system. An event message e1 causally precedes an event message e2 with respect to a topic subset T iff not only e1 causally precedes e2 but also e1.tvj ≤ e2.tvj for every topic tj in an intersection T of the publications of e1 and e2 and the subscription Si of pi. A pair of event messages e1 and e2 are unnecessarily ordered if e1.TV ≤ e2.TV but e1 does not causally precede e2. In this paper, we newly propose a topic-based-causally delivering (TBC) protocol where the linear clock and physical clock are used with topic vectors to reduce the number of pairs of messages unnecessarily ordered. We evaluate the TBC protocol and show the number of pairs of unnecessarily ordered messages is reduced.