Evolution of the Storage Brain 筆記 (二)

Chapter 4. A Journey to the Center of the Storage Brain

The Past: Refrigerators and Cards

• Skip

The Middle Ages: The Age of RAID Controllers

RAID controllers combined the basic functionality of disk controllers with the ability to group drives together for added performance and reliability

• Communication between the underlying disks and the attached host computer
• Protection against disk failure through the creation of one or more RAID groups
• Dual-controller systems, which could be used for added performance, availability and automated failover

The Modern Age: Storage Array Controllers

• The storage industry became very volatile during the 90s
• Modern-day networked storage was born. 
• EMC and NetApp capitalized on increased storage intelligence and grew at record paces

Enterprise Array Controllers: A More Intelligent Brain

As the storage industry moved from the late 90s into the mid 00s, the market stabilized into a smaller albeit more mature set of suppliers

Table: Data Management Tasks Addressed by Modern Arrays

Function	Description
Performance	Storage arrays are required to quickly process a multitude of data I/O requests arriving simultaneously from hundreds (or thousands) of desktops and servers
Resiliency	Ø  Automated RAID Rebuilds Ø  Data Integrity Checks Ø  “Phone Home” Alerts Ø  Environmental Monitoring Ø  Non-disruptive Software Updates
Virtualization	Ø  Transparent Volume/LUN resizing Ø  Thin Provisioning Ø  Data Cloning Ø  Data Compression Ø  Data Deduplication

Chapter 5.  Without Memory, You Don’t Have a Brain

Human	PC
Sensory memory	buffers and registers
Short-term memory	cache, Random Access Memory (RAM), flash and solid-state disks (SSDs)
Long-term memory	hard disk drives

Era	Memory Type
40s-50s	Cathode Ray Tubes
50s-60s	Magnetic Core Memory
70s-present	Random Access Memory (RAM) integrated circuits

NetApp add some interesting algorithm-level intelligence to its PAM cache

• Priority-based caching
• Non-redundant data caching
• Predictive caching
• Immediate caching
• Metadata caching

Here are the trends we’ll see in the industry’s efforts to reach this goal

• Application servers and workstations will cache more and more of their own data
• Storage networking switches will cache more and more data as it travels through their path
• Storage systems will cache more and more front-end data
• SSDs will become the first line of defense for large amounts of data that can’t be stored in cache
• Hybrid disk drives will be the final step

Chapter 6. The Storage Nervous System

Different between NAS, SAN, and iSCSI, 看了這麼多解釋, 我還是覺得鳥哥的圖最棒

NAS Communications Protocols
NFS	NFS stands for the Network File System protocol used by UNIX- based clients or servers
CIFS	CIFS stands for Common Internet File System. CIFS is another network-based protocol used for file access communications between Microsoft Windows clients and servers.
SAN Communications Protocols
Fibre Channel Protocol (FCP)	Typically occurs via specialized Fibre Channel cabling, Fibre Channel host bus adapters (HBAs) and Fibre Channel switches operating between the SAN and its hosts
iSCSI	iSCSI stands for Internet Small Computer Systems Interface. iSCSI is “a transport protocol that provides for the SCSI protocol to be carried over a TCP-based IP network.”
FcoE	FcoE stands for Fibre Channel over Ethernet. FcoE allows Fibre Channel storage traffic to be sent over Ethernet networks

What is data package collision in network? 

http://en.wikipedia.org/wiki/Collision_domain

• A network collision occurs when more than one device attempts to send a packet on a network segment at the same time
• Collisions are resolved using carrier sense multiple access with collision detection in which the competing packets are discarded and re-sent one at a time
• This becomes a source of inefficiency in the network
• Collision domains are found in a hub environment
• Collision domains are also found in wireless networks such asWi-Fi.
• Modern wired networks use a network switch to eliminate collisions.

Early Heritage Evolved into Separate Paths & Growing Confusion

10 Megabit per second (Mbps) 

Ethernet was the norm with 100 Mbps Ethernet just 

emerging. 

SANs came out of the gate with fiber optic cables and a protocol that could move data at 1 Gigabit per second (Gbps), a ten-fold improvement over the fastest Ethernet-based NAS transport

• anyone with a “need for speed” simply had to use SAN storage
• Databases, transaction processing, analytics, and similar applications fell into this category
• NAS, although less costly and easier to implement than SAN, was usually relegated to “slower” applications
• User files, images and Web content

New Realities Point to SAN/NAS Convergence and Unification 

10-Gigabit Ethernet (10-GbE or 10 Gbps) is common. 100-Gigabit Ethernet (100-GbE or 100 Gbps) devices have also been demonstrated.  Fibre Channel networks operating at 4-Gbps are common today, with 8-Gbps having also been newly delivered.

• Unifying block and file data transport onto the same network fabric
• Unifying SAN and NAS data storage functionality onto the same storage system.

Advancing Intelligence for Internal Communications

SAN Virtual Storage

• add a second logical abstraction layer to the physical disk drives
• did not map LUNs to physical drives, it mapped LUNs to the logical blocks stored on these drives

NAS Virtual Storage

• NetApp had created an intelligent internal communications network specifically designed for storage systems. The WAFL file system provided a communications network based on logical file-based objects

Next breakthrough in storage protocols and communication becomes object-based storage, SONET, RDMA