Loopback transport support has been
added to transport-independent RPC.
Automatic user-space thread generation has been
enabled in the RPC library.
NFS server-side performance enhancements have been
added.
Loopback
Transport Support
Loopback transport provider devices (tlclts, tlcots, and tlcotsod) have been added to the TI-RPC definition file, /etc/netconfig. Also, the system now has the following new loopback
transport-specific directories and files:
/etc/net/loopback_transport_name/hosts
/etc/net/loopback_transport_name/services
The following ONC/NFS daemons support loopback transport requests:
/usr/sbin/rpcbind
/usr/sbin/keyserv
/usr/sbin/rpc.nisd
/usr/sbin/nis_cachemgr
The netid and address fields in the rpcinfo call (which queries /usr/sbin/rpcbind to determine what services have been registered) now
give the loopback device name plus an address name, rather than
just the netbuf address provided by udp and tcp transport.
Additionally, the ticlts loopback transport device has a randomly generated string address.
User-Space
Thread Generation
To process incoming RPC requests, the svc_run() function call in the RPC library automatically generates
a thread on behalf of the application. The threads are managed by
the RPC library software. RPC threads may be created when calling
the RPC library.
NFS Server-Side
Performance Enhancements
The NFS server daemon, /usr/sbin/nfsd, has been modified to enhance performance. As a result,
the user may see more nfsd daemon processes running than requested, depending
on the number requested and the number of processors configured.
This change is documented in the nfsd(1M) manpage.
The NFS client-side buffer cache management has been modified
to improve server performance from a VxFS file system mounted on
the client.
The performance enhancements included have given HP industry
leading NFS SPECsfs benchmark values on our V-Class platforms.
Printable version
