storage school 1
TRANSCRIPT
Storage School I
An introduction to the world of storage
Storage School I: An introduction to the world of storage
Presented by Stephen FoskettDirector of the Data Practice
www.contoural.com
The world of storage can be daunting the uninitiated. This session provides all of the
background that you will need to get started in the world of storage. We will start with the basic
concepts: SAN vs. NAS, block vs. file, RAID levels, and other basic topics. These are woven together in a lively lesson, explaining how we got here and why it all matters. We will finish up with a brief
discussion of how storage fits within the big picture of enterprise IT.
WHAT I ASSUME YOU KNOW
• Storage School I assumes no prior knowledge of storage topics, but a basic comprehension of computing and networks would be helpful.
BY THE END OF THE SESSION, YOU’LL KNOW THE FOLLOWING:
• A bit of history and context• Five Important Concepts:
• Storage outside the computer• Blocks and files• The importance of SCSI• What RAID is and why it’s important• The three kinds of storage arrays
• How it all fits together in a storage architecture
How involved with storage are you?1. I’m soaking in it! (it’s my job)2. Touch and go (I’m involved but not all the time)3. I just stepped in quicksand! (I’m new to all this)4. Cardboard boxes and tape! (what’s this all about?)
NIL NOVI SUB SOLE(There Is Nothing New Under The Sun)
• The basic concepts of storage are not new and most are easy to grasp once the reasoning and history behind them is understood
• Simply put, the storage world of today is the result of consolidation, networking, and sharing of resources
• We mostly talk about open systems now, but much of the work was pioneered in the world of mainframes and minicomputers
Important Concept #1:Storage is Outside the Computer
• Mainframe storage has always been located in a separate cabinet• IBM introduced the first disk drive system in 1956,
the 350 disk storage unit• The storage industry was born with “plug-
compatible” storage for the System/360’s 2311 and 2314 DAS in the 1960s
• Bus-and-tag became ESCON in 1990• Open systems storage moved outside later
• Seagate’s 1980 introduction of the ST506 brought hard disk storage to the personal computer
• Introduced in 1986, SCSI allowed personal computers and servers to access external storage
IBM 350
IBM 2314
ST506
ESCON
SCSI
We’re Used to External Storage Today
• External disks are common from PCs to servers• FireWire and USB storage is used on PCs• External Serial ATA (eSATA) is becoming more
common• Servers still use SCSI, but also commonly use Fibre
Channel• Networked storage is also gaining attention
• NAS and iSCSI use common Ethernet and IP protocols
• Enterprise storage generally consists of SCSI, Fibre Channel, and Ethernet
FireWire and USB
eSATA
Ethernet
Fibre Channel
Important Concept #2:Blocks and Files
• Disk drives (and things like disk drives) organize data in blocks• Equal-sized units have unique addresses on the disk
• People (and most applications) organize data as files in folder hierarchies
• Filesystem drivers in the operating system translate file requests to block addresses
Most Enterprise Storage Systems and Protocols are Block or File Based
• Block protocols require a filesystem driver in the computer to locate files• SCSI, Fibre Channel, and iSCSI• Also USB, FireWire, thumb drives – anything that acts like a
disk drive• File-based devices handle the file translation and
organization themselves• File servers and NAS arrays return data based on directory
location and filename• Content-addressable storage (CAS) is something
else altogether• CAS uses a hash of the content itself (block or file) to
create a unique address for data
SAN and NAS
• A storage area network (SAN) is a block storage network• SCSI “initiators” (servers) talk to “targets” (disks or arrays)
and request access to logical sets of block storage (LUNs)• SAN implies FC or iSCSI storage
• Network-attached storage (NAS) is a file storage network• Clients request files from file servers or NAS arrays
(“filers”)• Common NAS protocols include CIFS/SMB for Windows and
NFS for UNIX• A network of NAS devices has been dubbed a file area
network (FAN)
Important Concept #3:Most Enterprise Block Storage Uses SCSI
• SCSI is the foundation of all current enterprise block storage protocols• “SCSI” is both a command set and physical
specification• Thick parallel SCSI cables of old have been
replaced by new connections• “Fibre Channel” = SCSI commands over Fibre
Channel Protocol on optical fiber or copper cables• iSCSI = SCSI commands over TCP/IP, commonly
over Ethernet• SAS = SCSI commands over some FCP services and
a serial transport based on Serial ATA (SATA)• Mainframes now use FICON which is like
ESCON over FCP (not SCSI)
SCSI commands
FCP
Optical/Copper
“Fibre Channel”
SCSI commands
Partial FCP
SATA Copper
Serial Attached SCSI (SAS)
SCSI commands
TCP/IP
Optical/Copper
iSCSI
Ethernet/other
…But other protocols are used by disk drives
• Serial ATA (SATA) is used in lower-end drives • Replaced parallel ATA, also called IDE• SATA is quick, common, and cheap
• Serial-Attached SCSI (SAS) is the next tier• Replaced parallel SCSI as higher-end drive• Shares common components with SATA but adds SCSI
command set (and “command queueing”)• Native Fibre Channel drives are still tops
• Have non-optical FC interconnect• Enterprise drives versus desktop drives
• Enterprise are more sturdy and pass more rigorous tests• Spinning speed (RPM) has huge impact on performance
Important Concept #4:RAID Combines Disks
• A Redundant Array of Independent Disks (RAID) is a combination of disk drives acting as one• RAID can improve performance and reliability
• RAID is as old as storage• IBM patented the general concept in 1978 • David Patterson, Garth Gibson, and Randy Katz defined
five idealized RAID “levels” in 1988• The “I” originally stood for “inexpensive”, but this proved
to be inaccurate once arrays were produced for sale!• Today there are literally dozens of different
implementations of the RAID concept
Common RAID Levels
RAID 0 “Stripe”• Poor reliability – no data
protection and double the chance of failure!
• No “wasted” space• Fast reads and writes – 2x!
RAID 1 “Mirror”• Good reliability• 50% overhead for data
protection• 50% “wasted” space• 2x fast reads but slower
writes
Common RAID Levels
RAID 4 “Dedicated parity”• Good data protection• Less “wasted” space (N-1)• Nx faster reads but slower
writes• Parity across blocks means
lots of recalculation if they’re not written at the same time
RAID 5 “Striped parity”• Good data protection• Less “wasted” space (N-1)• Nx faster reads but slower
writes• Parity for each block means
they can easily be written individually
PPPP
PP
PP
RAID Mashups
• It is common to “stack” RAID levels as “RAID X+Y” where “X” is laid over top of “Y”• RAID 0+1 (or RAID 01) is mirrored stripes• RAID 1+0 (or RAID 10) is striped mirrors• RAID 5+0 (or RAID 50) is striped RAID 5
• RAID 6 or RAID DP has two parity slots – either a duplicate or an alternate calculation
• RAID E mixes a hot spare disk into the striping• Some vendors use RAID on a region of a disk
instead of a whole disk drive
Grid array
Important Concept #5:There are Three Kinds of Storage Array
• Monolithic arrays are large cabinets with many disk slots, controllers, and I/O paths• IBM, EMC, and HDS started with the mainframe
• Modular arrays use a 1- or 2-controller “head” and generic disk shelves that can be added as needed• 3Com was followed by NetApp, CLARiiON, etc…• Clustered heads and SAN storage can be used
• Grid arrays have small nodes with a few drives that team up in flexible clusters for performance and reliability• Upstart CAS and iSCSI arrays were first to use this
concept
Monolithic array
Modular array
Choices Abound for Networked Enterprise
Storage• No matter if you’re looking for SAN, NAS, iSCSI, or
CAS there are lots of options• There are monolithic, modular, and grid devices that
support most protocols• Every type of equipment and protocol could
support every type of application• Databases can run great on NAS or RAID 5• You can build a cheap SAN with Fibre Channel or iSCSI• NAS filers can make great archiving targets• Workstations can share SAN storage• CAS can be accessed with NFS or CIFS• You can put tier-3 bulk storage on an enterprise array• Modular arrays can outrun their big brothers
Mixing Up the Right SAN
• The best choice is the one that makes the most sense in your environment
• Select the right tool for the job instead of using a wrench as a hammer• Just because something can work doesn’t make it a good
idea• Always pick the simplest and most straightforward solution
• Look for the best fit for your budget and scale• If you only have a few terabytes, buy just one networked
array that will work for most of your applications• Match your chosen technologies with the platforms
and applications you have
Architecture Example:Small Web Company
• A small but growing business focused on a web-based product• Wants stability, flexibility, scalability, low cost, and DR• All Windows, mostly file but with some block storage
• Selected a modular NAS/iSCSI array• NAS replaced all current Windows file servers• iSCSI replaced internal storage for email and database• Picked a midrange device with lots of growth potential• Used SATA drives with RAID 6 for reliability and “good
enough” performance• Built-in snapshots and replication of both file and block
data from a single interface
Architecture Example:Large Financial Company
• A household name in the world of finance• Wanted to implement tiered storage to save money• Hundreds of TB, mixed Windows, UNIX, and mainframe
• Selected a modular FC SAN device• Sufficient staff and money to bring in a new storage
platform• Spent time and money on data classification to move less
critical apps off Tier 1• Decided to consolidate Windows systems with
virtualization and blades rather than use iSCSI• Deferred all enterprise storage purchases for two years• Kept all mainframe data on Tier 1 enterprise storage for
now
Closing Thoughts
• Bring in the storage that is right for you• Don’t let “rules of thumb” and bogus “best practices”
prejudice your choice• All storage devices work pretty well these days -
but none are perfect• Don’t try to do anything exotic with basic devices• Use the right tool for the job
• Make the vendors prove it works• Talk to references who are doing what you want to do• Create a proof of concept before buying
• Remember that it’s not all about the technology – even the best storage can’t fill an uncertain need!
Questions?
• Audience Q&A: 10-15 minutes
• Contact me at [email protected]
• Come talk to me after the session or at lunch
• I'll be available at the Ask-the-Expert booth today and tomorrow from 5 PM to 6 PM
For More Information
• Contact me:• Stephen Foskett – [email protected]
• Visit SearchStorage.com and read Storage magazine
• Get SNIA’s "Network Storage Terms and Acronyms" book
• Ask others here at the show or at user groups• Storage Networking User Group (SNUG)
• http://storagenetworking.org• Association of Storage Networking Professionals (ASNP)
• http://asnp.org• Ask the vendors (really!)