Exadata Storage cell servers are basically performing the work of storing database’s data and in addition to that due to the self-intelligence some other work also perform at the storage level like smart scan, HCC decompression etc. which basically boost database performance. Exadata RACK comes in various mode like quarter rack, half rack, full rack and now from X8M onward it also comes with elastic configuration in which we can increase storage and database server both, the no. of cell servers exists in the Exadata Rack vary from configuration to configuration. Below is the cell server configuration as per Exadata RACK.
|
EXADATA RACK |
No. Of Cell Server |
|
Eight Rack |
3 Cell Server |
|
Quarter Rack |
3 Cell Server |
|
Half Rack |
7 Cell Server |
|
Full Rack |
14 Cell Server |
Irrespective of the Exadata Rack, in every configuration
of Exadata rack cell servers comes with 12 SAS disk drives of uniform size in
each and every cell storage server. For E.g. In a quarter rack Exadata machine there
are 12 physical disk of uniform size resides under each and every cell storage
server which provide the feature of high availability with the help of using
the mirroring by ASM technology of RAC.
Prior to understand how the Exadata cell server work and how it provides high availability in addition to some other feature to boost performance as well, first of all we’ve to under the storage hierarchy of Exadata storage server.
Exadata RACK broadly comes with any of the storage server’s configuration in addition to Extended Storage.
|
High-Capacity Storage Server |
|
Extreme Flash Storage Server |
|
Extended Storage |
Difference Between Eight Rack & Quarter Rack
The main difference between eight rack and quarter
rack is the number of cores available at compute node and storage server both
and also the storage capacity however the no. of database server & storage
server will be as same as quarter rack in order to support maximum availability architecture (MAA).
|
Rack Size |
Database Server And
Cores |
Storage Server And
Cores |
HC Storage Server Capacity |
Extreme Flash Storage Server Capacity |
|
Eight Rack |
2xserver,48 cores |
3xserver, 48cores |
252TB Disk,38.4 TB Flash &4.5 TB PMEM |
76.8TBFlash And 4.5 TB PMEM |
|
Quarter Rack |
2xserver, 96 cores |
3xserver,96 cores |
504 TB Disk,76.8 TB Flash,4.5 TB
PMEM |
153.6 TB Flash And 4.5 TB PMEM |
High Capacity Exadata Cell Storage Server
1
Disk’s Size= 14 TB
12
disk * 14 TB each = 168 TB (Total Physical Disk Size in a single cell storage
server)
LUN: LUNs are the logical abstraction unit which are
mapped to Physical disk drives.
CELL DISK: Cell disks are mapped to LUNs which basically segregate
the storage of LUNs in the form the cell disk.
GRID DISK: GRID Disks are mapped to cell disks and on the top of
grid disk we consume storage in the form the asm disk group which is visible to
ASM.
Flash Storage
Device In HIGH-CAPACITY Storage Server
FLASH LUN: Flash LUNs are the logical abstraction unit which are
mapped to Cell disks in the flash card and are used to allocate or segregate the
storage area of flash card like the same we do with physical disks.
Cell disk: In Flash card we again map Flash LUNs to cell disks.
Flash Cache: In PCI NVMe flash card we mapped flash cache to cell
disks
Flash Log: In PCI NVMe flash card we mapped flash log to cell
disks.
Extreme Flash Storage Server
Extreme
flash storage Exadata RACK configuration comes with 8 flash drives of 6.4 TB
each on the contrary of using physical SAS disk drives in high-capacity storage
server in extreme flash storage server we use flash drive.
Flash LUN: In Extreme flash storage server flash drives are
mapped to flash LUNs instead of physical disks.
Cell Disk: Flash LUNs are mapped to cell disks to segregate
flash storage’s space.
Grid Disk: Like High-capacity storage server, Cell Disk are
again mapped to grid disk to consume storage area and to make it available for
the databases running in compute node.
Flash Cache: On the contrary to high-capacity storage server, in
extreme flash PCI NVMe flash cards are not available and smart flash cache and
smart flash log also mapped to flash drives.
NOTE: In Extreme flash storage, 95% of the PCI NVMe flash
drive’s space is use for data storage and only 5% is allocated for smart flash
cache and smart flash log.
Persistent Memory in High Capacity & Extreme Flash
From
X8M onward persistent memory also comes into the role which is also a major
contributor to boost database’s performance and it comes with Exadata RACK
irrespective of the storage server we choose and it’s a sharable cache between servers.
DB host directly access the data
from persistent memory using RDMA n/w which eliminate the CPU & OS cycle of
compute node.
Persistent memory is essentially a
memory card and it comes with a size of 1.5TB out of which 128GB is allocated
to each and every cell disk in a single cell storage server and it place in
between the flash cache and storage server.
128 GB Of Persistent memory allocated to one cell disk
128
GB X 12 Cell Disk = 1536GB (1.5 TB Approx)
PMEM LUN: PMEM device mapped to PMEM LUNs to allocate storage
area
Cell Disk: PMEM LUNs are mapped to cell disk to segregate
storage area of PMEM LUN
PMEM CACHE: Cell disks are mapped to pmem cache to consume
storage area
PMEM LOG: Cell disks are mapped to pmem log to consume storage
area
NOTE A: Almost 95% of the storage area of PMEM LUNs are
allocated to PMEM cache
NOTE B: In Exadata RACK storage server cell disk are one of
the mandatory part which we use to segregate allocated space whether it’s cell
disk, flash card, flash drive, PMEM Memory Card etc.
Extended storage server
It’s
basically use t achieve a lower cost as flash is not included into it and also
storage server software is optional in this storage server this type of storage
server is primarily used to store rarely accessed data.
