Genesis II V2.0

From VCGR Wiki

This is a list of ideas - not committments to implement.

• New Certificate life times. The current deployed system (XCG) has root certificates that are valid for only one year. They expire in June. All certificates that use that certificate as root (pretty much all of the system components) will expire at that time. We will be extending the root certificate lifetime, and changing the way certificates are used for authentication.

• X.509 key size discussion - are 2K keys really the way to go The key sizes are killing us on signing at the message level.

• X.509 proxy transport AuthN consistent with the OGF Production Grid Interoperation (PGI) profile (if it is out in time.)

• Resource Fork implementation of export_dir, and possibly replicated export dir

• Suspend/resume jobs so that we get better usage of ITC machines.

• BES support of streams to the grid. In other words, named-pipe-like streamable byteio resources as inputs and outputs.

• Info in SOAP header to say the client is either aware of rebinding, or is simply a Genesis II client.

• FUSE support back into BES, and in particular for PBS sequential job backends.

• Throttle staging

• Installation for lay users - may require purchasing decent installer software.

• Additional metadata in BES-factory attributes

• Test plan and test harness

• New version of patch system This will allow us to send patches to specific machines reather than the whole net.

• Change the way that clients look for patches so that it looks less often unless explicitly asked for.

• RNS 1.1 Support. We currently support a modified version of RNS 1.0. We will upgrade to RNS 1.1. This should in principle allow us to interoperte with other RNS 1.1 systems.

• Resource Usage Record (RUS) support in the BES containers. RUS is an XML standard for resource usage.

• Accounting record resource fork in BES containers that will allow clients (subject to access control) to peruse the accounting records.

• X.509 AuthN modification to not compare the who certificate, but rather definable sub-fields. This will get us around the problem that ITC (and perhaps others) change a users DN slightly everytime they issue a new certificate.

• MPI issues at UVA will be resolved.

• Application deployment supported again, this time for queue submitted jobs as well.

• Activity endpoints as resource forks.

• Resource routing in queues. Currently queues do not check if the BES container they send a JSDL document to can actually execute the job - i.e., whether the resource matches the job.

• Replicated RNS directories for critical directories.