Check out what’s new in this version of OurGrid.
The table below summarizes all changes made in the OurGrid 4.3.1 release.
Status |
Issue |
Description |
---|---|---|
CHANGED |
Support for VirtualBox 4.1.18 |
|
CHANGED |
Created new Worker's properties based on the VM image: OS and arch |
|
CHANGED |
Added support for numpy and scipy in the standard Virtual Box image |
|
CHANGED |
JSAGA + Rest API: File upload |
|
CHANGED |
JSAGA + Rest API: Authorization based on roles |
|
FIXED |
Bug fix: Error while taking a snapshot after a task execution in Virtual Box |
|
FIXED |
Bug fix: Worker+Virtual Box debian installation returning an error code (1) |
|
FIXED |
Bug fix: Idleness detector was not working in 64bits machines |
OurGrid 4.3.0 has changed! This version of OurGrid provides faster and simpler ways to deploy and manage the OurGrid sites!. Check out what’s new and what’s been fixed in this version below.
The table below summarizes all changes made in the OurGrid 4.3.0 release.
Status |
Category |
Description |
---|---|---|
NEW |
Issue #1305 |
OurGrid components can now be installed via Debian and RPM packages. This feature will dramatically decrease configuration overhead, mainly regarding virtualization configuration and dependency management. |
NEW |
Issue #1081 |
OurGrid is coming to Windows! Same benefits of the Deb installers plus sandboxing with Virtual Box, which wasn't supported in previous version. |
NEW |
Issue #1465 |
OurVirt is a Java API for Virtual Machine managers, developed in the LSD, UFCG. Emerged from the needs of the GenericExecutor (See below), which LibVirt didn't supply, OurVirt delivers a common interface for creating, starting, stopping, destroying, creating shared folders, mounting shared folders, getting status, executing commands, taking and restoring snapshots, cloning disks/images of virtual machines. Now it implements VServer and VirtualBox strategies. |
NEW |
Issue #1079 |
Isolates hypervisor-specific tasks from the Worker execution process. Now we use a single executor implementation for every sandbox strategy. |
NEW |
Issue #1466 |
The Worker Admin goes to the web. Now we deploy a single web application for the administrator that wants to manage its OurGrid community remotely. Management means deploying, starting, stopping, killing process, issuing certificates, updating workers remotely via SSH. |
NEW |
Issue #1433 |
Now the Worker components need an issued certificate provided by the Peer. The tools allow for easy certification generation/issuing on the Peer side. The certificate requester page has also a voluntary version, which create ready-to-donate Worker packages. |
NEW |
Issue #1769 |
The REST Broker API is a service that allows to: Submit a job; Get a status list of all submitted jobs to the broker; Get the job status specified by the id; Cancel a job specified by the id; Clean a finished job filtered by id |
CHANGED |
Issue #1794 |
The Peer configuration on Broker was changed, and the support for the GDF file was removed, by now we can set peers in the configuration file, using the broker.peer.address property. |
CHANGED |
Issue #1076 |
The Dynamic Worker feature inverts the interest between the Peer and the Worker components. Now the Worker tries to login in its Peer during its starting process. If that Worker has a certificate issued by the Peer, the login process ends sucessfully and the Worker will be managed by that Peer. This feature allows for easy addition of new Workers and for voluntary computing. |
CHANGED |
Issue #1469 |
A voluntary Peer does not check for issued Workers' certificates. In other words, it accepts any Worker who means to donate. |
CHANGED |
Issue #1485 |
A Peer that enables on-demand registration will accept any Brokers that wants to consume its resources, without any previous registration. |
CHANGED |
Issue #1305 |
Worker Idleness detector now works for USB devices. Using an adapted code of the libxprintidle, the Worker is now able to handle that kind of even. |
CHANGED |
Issue #1468 |
In the previous versions of OurGrid, when a Worker was preempted, the Broker waited for its failure. A single message called WorkerPreempted could avoid the dependency of the Failure Detector in this scenario, which will reduce makespan in certain scenarios. |
CHANGED |
Issue #1074 |
Discovery Service Availability feature introduces the concept of Discovery Service Network. Several Discovery Service may connect among themselves, creating a network and signaling Peers of the existance of others DS. The Peers now persists a list of Discovery Services and may connect to others DS if its Primary DS fails. To achieve scalability, we limit the number of Peers a DS can hold. If the DS is overloaded when a Peer tries to login on it, the Peer will be notified and will try to connect to other DS that it knows. Also, a Peer can determine the maximum number of Peers it will forward requests. |
CHANGED |
Issue #1436 |
Implemented due to the Dynamic Worker feature. Now the Worker has a WorkerSpec file where tags or properties can be added, propertis that are not dynamic or can't be inferred by the Worker. This file has a similar function to the old SDF (Site Description File). |
FIXED |
Issue #1413 |
Broker relies on a never-arriving failure message of a preempted Worker. |
FIXED |
Issue #1422 |
Worker status change notifications should be labelled with its consumer. |
FIXED |
Issue #1464 |
OurGrid charts: Out of memory on Peer Availability queries. |