Module 1: Data Center Infrastructure Flashcards
What is a data Center
A location holding Compute, network and storage.
Contains secure power supply, security and environment controls.
Usually for Mission-Critical and Business-Critical applications.
What is compute
The server systems or storage on which applications are hosted.
What is network
The network component such as switches, routers and cables.
What is storage
The persistent systems on which all data is stored
What is the differenence between converged and Hyperconverged?
Converged is prepackaged system of server network storage, components are discrete, preconfigured, and separate. Deployment options with reference architecture and pre-racked. Used for Private Cloud and Machine learning.
Hyperconverged provides components that are tightly coupled and abstracted through virtualisation, with greater gains through automation. Often deployed on commodity components. Ideal for remote offices or VDI.
What are the different types of storage media
Magnetic disk - large capacity, cheap
SSD - fast, more expensive
Optical Disk drive - polycarbonate disk, uses laser beams, CD, DVD, etc. WORM
Magnetic tape drives - large capacity, cheap used for backups, sequential reads and write
What are the Storage Class Memory Technologyes
Read cache
Write cache
SCM sits between the disks and a cache to give a bigger cache system
Higher performance and lower cost
used for random read-intensive workloads
Reduces latency and improves throughput
Data can be address at the byte or block level
Can SCM bet address at byte level or block level?
Block level
What is SSD
Data storage device that uses SSD memory also known as flash
Key components are flash controller and NAND flash memory chips
Commonly used SSD interfaces SAS, SATA, FC, PCIe and USB drives
Benefits of SSD compared to HDD
Higher reliability
Higher performance
Faster access to data
Weighs less
More power-efficient
Less cooling required
How do SSDs store data?
In flash cells combined into pages and then into blocks
There are 3 primary types of SDD are;
Single-Level Cell (SLC) - 1 bit/flash cell - most expensive and lowest capacity.
Multi-Level Cell (MLC) - 2 bits/flash cell - Cheaper than SLC but higher error rates than SLC.
Triple-Level Cell(TLC) - 3 bits/flash cell - High capacity but low performance.
Read/write operations of SSD
There are 3 types
Page-Read
Page-Write
Block-Erase
Page status is either free, written or Invalid.
What is Write Amplifcation?
When the number of writes to the media is higher than the writes issued by the host.
You must copy the written data to a new location. The controller reads the content into cache and erases the block.
What are the tasks that can maximise SSD endurance?
Garbage Collection - SSD controller cleans up blocks for deletion and defrags the blocks
Wear Levelling - distribute the workload across the Cells (there are two types inflight and background).
Overprovisioning - the flash drive will have hidden blocks available that can be used if required when space need.
Caching - write coalescing (hold in write cache until a whole block is written), write combining (when a block is written to repeatedly in a short amount of time. It is held in cache until the final resulting IO is then written to the disk.
What type of NAD flash has strong error correction for lower error rates?
eMLC
What is a reason to use overprovisioning in an SSD?
It extends the life of an SSD
What is DAS?
Direct Attached storage.
This is where the data storage resides in the same frame as the compute.
E.g. an internal hard disk on a PC or server
What is a network storage environment
Either
SAN ( storage device connected via FC switches to hosts to provide LUNs )
NAS ( storage device connected via IP switches to host providing network shares in CIFS or NFS)
Common storage interfaces
SATA - common interface on consumer desktops and laptops. Cheap, low performance, high capacity. Direct attached
SCSI - enterprise storage interface providing parallel transmission and used for high performance and mission critical workloads
SAS - Serial SCSI is serial point to point variant of SCSI. SAS is used for high-end computing.
NLSAS - same as SAS but in the back end uses SATA drives which are cheaper and slower but offer large capacites.
FC - Fibre Channel the protocol for sending data on a SAN network with very high throughput support up to 65 Gbits/sec
NVMe is a flash storage based interface that reduces I/O overhead and includes performance improvements.
What is RAID
RAID means “Redundant array of independent disks”
RAID technology combines multiple physical drives to create one logical drive.
Function is to protect data and improve performance (through more available spindles).
How is RAID implemented?
Software RAID -= implemented at the OS level, performance dependant on the host
Hardware RAID - implemented at the host or storage level, dedicated hardware so faster.
What are the 5 RAID types
Striping - RAID 0 - all disks combined to make one logical drive, no protection
Mirroring - RAID 1 - two or more disks hold duplicate copies of the data best for protection
Parity - RAID 3 - one disk holds parity, rest hold data. Good for large sequential writes and reads. RAID 5 splits parity across all disks for better performance.
RAID DP - RAID6 - this uses parity across the disks in horizontal and vertical slices two disks lost to parity but better protection
RAID-TEC (tripe erasure coding). Add a triple-parity disk to a RAID DP group. Protects against 3 disk failures (where there are long rebuild times). Default for NL-SAS and SATA.
What is the minimum number of disks that are required to create a RAID DP data aggregate (excluding hotspares)
Five
Erasure Coding
Forward error correction. Parity based technology uses Reed-Solomon encoding algorithm.
This is used to re-create any corruption
Benefits are reliability and availability with more storage efficiency.
Best used for sequential rights.
Best used for archival storage (ideally in cloud).
Data is split into data and parity fragments. D+P=N (Storage node).
True or false. In erasure coding, if a data fragment or a parity fragment become corrupt or lost, you require the original data for data recovery?
FALSE
