What is LMS process in Oracle RAC database

What is LMS process in Oracle RAC databases?

LMS process called as Lock Manager Server Process , also called the GCS (Global
Cache Services) process. Main work of this process is to transport blocks across
to the nodes of Oracle RAC for ¡§cache fusion¡¨ supportability. If any consistent
read request found from another connected node then LMS process makes consistent
read image of block and transport it to across the nodes for full feeling
client¡¦s consistent read request. This images transports using High Speed
Interconnect process to and from remote nodes.

If we are using 8K block size then transporting consistent read images across to
remote node using HSI using interconnect protocall becomes more busy and high
CPU utilization. Which we can check from AWR reports, Statspack reports or Wait
event called ¡§global Cache Services¡¨ wait event. If this wait event consumes
more CPU utilization and simultaneously found performance degradation of Oracle
Database causes ¡§cache fusion¡¨ and consistent read request traportation across
to RAC nodes.

How to avoid this kind of performance degradation of Oracle RAC?


In Oracle Real Application Clusters, the cluster interconnect is a dedicated or
stand alone network This interconnect network is using to communicate the
connected nodes across the clustering of RAC.

Ethernet is using in networking of Cluster interconnect protocol. Ethernet¡¦s
default frame size of 45-1500 bytes is transfering unit between the all
participant of nodes and hosts. If request of tranferring data is larger than
1500 bytes then it will be split ed into smaller units and send multiple packets
across to nodes using interconnect because lower range is 45 to 46 and upper
limit of data is 1500 (maximum transfer unit) bytes.

Deployment of Jumbo Frames is very best option to such drastic situation of
performance bottleneck. Jumbo Frames allows Ethernet packets to be sent in a
larger size rather than default size.

Generally Ethernet frames use a 1500 bytes size. Jumbo Frames allows this to be
extended to 9000 bytes frame. This 9000 byte frame actually allow to send or
transport 8k or larger block size data into single request of packets across to
remote nodes instead of multiple frames which size default is 1500 bytes. This
scenario is improving performance of RAC nodes specially ¡§global Cache
Services¡¨ wait event because it directly connected with consistent read request
block image transporting across to remote nodes using interconnect protocol.

Multi block read request (called full table scan) is also affecting with same
scenario while we are using lower bound limit or default bound limit of MTU
(maximum transfer unit) of interconnect of RAC. It is also stripping data in
smaller units and send across to RAC connected nodes for ¡§cache fusion¡¨ or
¡§cr global cache request¡¨. It is most targeting for deployment of same Jumbo
Frames.


If small data is being transported then Jumbo Frames are not efficient and not
recommended.


Generally it is advisable to deploy for large amount of CR image transfer (cache
fusion) using LMS process (means Global Cache Services) and your data block size
is 8k or larger. But it is not advisable to implement in general IP address of
Ethernet card of server otherwise it may be degrade performance while small
packet request is processing.

Said article from following source.

http://www.dbametrix.com