elements of san capacity planning mark friedman vp, storage technology [email protected] (239)...
TRANSCRIPT
Elements of SAN capacity planning
Mark FriedmanVP, Storage Technology
[email protected](239) 261-8945
04/19/23
Overview
How do we take what we know about storage processor performance and apply it to emerging SAN technology?
What is a SAN?
Planning for SANs:SAN performance characteristics
Original test results, May 2001: 4 x 550 MHz, 1 Gb FC More recent test results, May 2002: 2 x 2.266 GHz, 2 Gb
FC SPC-1 testing using 3 nodes with 2 x 3 GHz, 1 & 2 Gb FC
Backup and replication performance
Evolution Of Disk Storage Subsystems
See: Dr. Alexandre Brandwajn, “A study of cached RAID 5 I/O” CMG Proceedings, 1994.
Cached Disk
Strings & FarmsSpindles
Storage Processors
Write-thruCached
subsystems
04/19/23
Disk Subsystem Modeling framework:
ComponentsFront end interface bandwidthInternal bus bandwidthNumber of processors and their speedCache memoryDisk interfacesSCSI Disks
Internal segmented buffering Seek, Rotational Delay, Data transfer rate
Logical:physical disk mappingLinear mappingDisk stripingRAID mapping
04/19/23
What Is A SAN?
Storage Area Networks are designed to exploit Fibre Channel plumbing
Approaches to simplified networked storage:SAN appliances SAN Metadata Controllers (“out of band”)SAN storage managers (“in band”)
04/19/23
The Difference Between NAS and SANStorage Area Network (SAN) designed to exploit Fibre Channel plumbing require a new infrastructure.
Network Attached Storage (NAS) devices plug into the existing networking infrastructure.Networked file access protocols (NFS, SMB, CIFS)TCP/IP stack
Media Access: Ethernet, FDDI
Application: HTTP, RPC
Host-to-Host: TCP, UDP
Internet Protocol: IP
Packet Packet Packet Packet
04/19/23
The Difference Between NAS and SAN
NAS devices plug into existing TCP/IP networking support.
Performance considerations:1500 byte Ethernet MTU
TCP requires acknowledgement of each packet, limiting performance.
Application InterfacesRPC DCOM Winsock NetBIOS
Named Pipes
NetBT
Redirector Server
User ModeKernel
TCP UDP
IP ARPICMP IGMP IP Filtering IP Forwarding
Packet Scheduler
NDIS WrapperNDIS Miniport NIC Device Driver
TDI
NDIS
04/19/23
Source: Alteon Computers, 1999.
The Difference Between NAS and SAN
Performance considerations: e.g.,1.5 KB Ethernet MTU
Requires processing 80,000 Host interrupts/sec @ 1 Gb/sec
or Jumbo frames, which also requires installing a new infrastructure
Which is why Fibre Channel was designed the way it is!
04/19/23
The Holy Grail!
Storage Area NetworksUses low latency, high performance Fibre Channel switching technology (plumbing)
100 MB/sec Full duplex serial protocol over copper or fiber
Extended distance using fiberThree topologies:
Point-to-PointArbitrated Loop: 127 addresses, but can be bridged
Fabric: 16 MB addresses
04/19/23
The Holy Grail!
Storage Area NetworksFC delivers SCSI commands, but Fibre Channel exploitation requires new infrastructure and driver support
Objectives:Extended addressing of shared storage pools
Dynamic, hot-plugable interfacesRedundancy, replication & failover Security administrationStorage resource virtualization
04/19/23
Distributed Storage & Centralized Administration
Traditional tethered vs untethered SAN storage
Untethered storage can (hopefully) be pooled for centralized administration
Disk space pooling (virtualization)Currently, using LUN virtualizationIn the future, implementing dynamic virtual:real address mapping (e.g., the IBM Storage Tank)
Centralized back-upSAN LAN-free backup
04/19/23
FC is packet-oriented (designed for routing).FC pushes many networking functions into the hardware layer.e.g., Packet fragmentationRouting
Storage Area Networks
Upper Level Protocol
SCSI IPI-3 HIPPI IP Fc4
Framing Protocol/Flow ControlFc2
8B/10B Encode/Decode Fc1
100MB/s Physical Layer Fc0
Common Services Fc3
04/19/23
FC is designed to work with optical fiber and lasers consistent with Gigabit Ethernet hardware100 MB/sec interfaces200 MB/sec interfaces
This creates a new class of hardware that you must budget for: FC hubs and switches.
Storage Area Networks
04/19/23
Performance characteristics of FC switches:Extremely low latency ( 1sec), except when cascaded switches require frame routing
Deliver dedicated 100 MB/sec point-to-point virtual circuit bandwidth
Measured 80 MB/sec effective data transfer rates per 100 MB/sec Port
Storage Area Networks
04/19/23
Upper Level Protocol
SCSI IPI-3 HIPPI IP Fc4
Framing Protocol/Flow ControlFc2
8B/10B Encode/Decode Fc1
100MB/s Physical Layer Fc0
Common Services Fc3
When will IP and SCSI co-exist on the same network fabric?iSCSINishanOthers?
Storage Area Networks
04/19/23
Storage Area Networks
FC zoning is used to control access to resources (security)
Two approaches to SAN management:Management functions must migrate to the switch, storage processor, or….
OS must be extended to support FC topologies.
04/19/23
Approaches to building SANs
Fibre Channel-based Storage Area Networks (SANs)SAN appliancesSAN Metadata ControllersSAN Storage Managers
Architecture (and performance) considerations
04/19/23
Approaches to building SANs
Where does the logical device:physical device mapping run?Out-of-band: on the clientIn-band: inside the SAN appliance, transparent to the client
Many industry analysts have focused on this relatively unimportant distinction.
04/19/23
SAN appliances
Conventional storage processors withFibre Channel interfacesFibre Channel support
FC FabricZoningLUN virtualization
04/19/23
SAN Appliance Performance
Same as before, except faster Fibre Channel interfaces
Commodity processors, internal buses, disks, front-end and back-end interfaces
Proprietary storage processor architecture considerations
Internal Bus
Cache Memory
FC
In
terf
aces
FC
Dis
ks
Host
Inte
rfaces
Multiple Processors
04/19/23
SAN appliances
SAN and NAS convergence?Adding Fibre Channel interfaces and Fibre Channel support to a NAS box
SAN-NAS hybrids when SAN appliances are connected via TCP/IP.
Current Issues:Managing multiple boxesProprietary management platforms
04/19/23
SAN Metadata Controller
SAN ClientsSAN Clients
Pooled Storage Resources
Pooled Storage Resources
SANMetadataController
1 3Token
Fibre Channel2
SAN clients acquire an access token from the Metadata Controller (out-of-band)
SAN clients then access disks directly using proprietary distributed file system
04/19/23
SAN Metadata Controller
Performance considerations:MDC latency (low access rate assumed)Additional latency to map client file system request to the distributed file system
Other administrative considerations:Requirement for client-side software is a burden!
04/19/23
SAN ClientsSAN Clients
Pooled Storage Resources
Pooled Storage Resources
Fibre Channel
StorageDomainServers
SAN Storage Manager
Requires all access to pooled disks through the SAN Storage Manager (in-band)!
04/19/23
SAN ClientsSAN Clients
Pooled Storage Resources
Pooled Storage Resources
Fibre Channel
Storage
Domain
Servers
SAN Storage Manager
SAN Storage Manager adds latency to every I/O request
How much latency is involved?
Can this latency be reduced using traditional disk caching strategies?
04/19/23
Architecture of a Storage Domain ServerRuns on an ordinary Win2K Intel server
The SDS intercepts SAN I/O requests, impersonating a SCSI disk
Leverages:Native Device driversDisk managementSecurityNative CIFS support
Fibre Channel HBA Driver
SCSI miniport Driver
Disk Driver
Diskperf (measurement)
Fault Tolerance (Optional)
Data Cache
Fault Tolerance
Initiator/Target Emulation
FC Adaptor Polling Threads
Security
Natives W2K I/O Manager
SANsymphony Storage Domain Server
Client I/O
04/19/23
Sizing the SAN Storage Manager serverIn-band latency is a function of (Intel server) front-end bandwidth:Processor speedNumber of processorsPCI bus bandwidthNumber of HBAs
and performance of the back-end Disk configuration
04/19/23
SAN Storage Manager
Can SAN Storage Manager in-band latency be reduced using traditional disk caching strategies?
Read hitsRead misses
Disk I/O + (2 * data transfer)
Fast Writes to cache (with mirrored caches) 2 * data transfer Write performance ultimately determined by the
disk configuration
04/19/23
SAN Storage Manager
Read hits (16 KB block):Timings from an FC hardware monitor1Gbit/s Interfaces
No bus arbitration delays!
140sec 27sec
StatusFrame
SCSI ReadCommandLength =
400016x1024 Byte Data Frames
04/19/23
SCSI Command Write Setup Data Frames SCSI Status
Fibre Channel Latency (16KB Blocks)
Read vs. Write hits (16 KB block)
04/19/23
SCSI Command Write Setup Data Frames SCSI Status
How is time being spent inside the server?
PCI bus?Host Bus adaptor?Device polling?Software stack?
Decomposing SAN in-band Latency
04/19/23
Memory Bus
64b
it/3
3M
Hz
PC
I
32b
it/3
3M
Hz
PC
I
32b
it/3
3M
Hz
PC
I
4x550MHzXEON
Processors
4-way 550 MHz PCMaximum of three FC interface polling threads
3 PCI buses (528MB/s Total)
1, 4, or 8 QLogic 2200 HBAs
Benchmark Configuration
04/19/23
How is time being spent inside the SDS?PCI bus?Host Bus adaptor?Device polling:
1 CPU is capable of 375,000 unproductive polls/sec 2.66secs per poll
Software stack:3 CPUs are capable of fielding 40,000 Read I/Os per second from cache
73secs per 512-byte I/O
Decomposing SAN in-band Latency
04/19/23
SDS FC Interface Data Transfer
SANsymphony in-band Latency (16KB Blocks)
Decomposing SAN in-band Latency
04/19/23
Impact Of New Technologies
Front-end bandwidth:Different speed ProcessorsDifferent number of processorsFaster PCI BusFaster HBAs
e.g. Next Generation Server2GHz GHz Processors (4x Benchmark System)200MB/sec FC interfaces (2x Benchmark System)4x800MB/s PCI bus (6x Benchmark System)
...
04/19/23
Impact Of New Technologies
2GHz CPU & New HBAs
2GHz CPU, New HBAs, 2Gbit Switching
1 year ago
04/19/23
Impact Of New Technologies
May 2001 May 2002
4 x 550 MHz,1 Gb FC, PCI-2
2 x 2266 Mhz,2 GB FC, PCI-X
Unproductive (idle) poll 2.66 secs 0.62 secs
CPU time Cache hit (read) 75 secs 25 secs
Throughput (16 KB) 270 MB/sec 952 MB/sec
I/Os per second 40,000 180,000
04/19/23
Sizing the SAN Storage Manager
ScalabilityProcessor speedNumber of processorsPCI bus bandwidth
32bit/33MHz 132MB/sec 64bit/33MHz 267MB/sec 64bit/66MHz 528MB/sec 64bit/100MHz 800MB/s (PCI-X)
NGIO???Number of HBAs
200 MB/sec FC interfaces feature faster internal processors
04/19/23
Sizing the SAN Storage Manager
Entry level system: Dual Processor, single PCI bus, 1 GB RAM
Mid-level departmental system: Dual Processor, dual PCI bus, 2 GB RAM
Enterprise-class system: Quad Processor, triple PCI bus, 4 GB RAM
04/19/23
0
10,000
20,000
30,000
40,000
50,000
0 2 4 6 8 10FC HBA s
Max
IO
s p
er s
eco
nd
(@
512
by
tes)
0
50
100
150
200
250
300
Max
MB
/sec
(@
16 K
B)
Max Read I/ Os Max W rite I/ Os
Max Read Throughput Max W rite Throughput
SAN Storage Manager PC scalability
04/19/23
0
10,000
20,000
30,000
40,000
50,000
0 2 4 6 8 10FC HBA s
Max
IO
s p
er s
eco
nd
(@
512
by
tes)
0
50
100
150
200
250
300
Max
MB
/sec
(@
16 K
B)
Max Read I/ Os Max W rite I/ Os
Max Read Throughput Max W rite Throughput
Entry level
Departmental SAN
Enterprise class
SAN Storage Manager PC scalability
04/19/23
May 2002
SAN Storage Manager PC scalability
SAN Appliance performance
0
200
400
600
800
1000
1200
100 1000 10000 100000 1000000 10000000
Block size
Th
rup
ut
(MB
/se
c)
0
40000
80000
120000
160000
200000
240000
I/O
s p
er
se
co
nd
Thruput (MB/sec) IOs per second
DataCore SPC-1 Results (August 2003)
www.storageperformance.org/Results/SPC-1/DataCore_2003-08-11_SANsymphony/
SPC-1 Submission Identifier: A00015
Tested Storage Configuration (TSC) Name: DataCore SANsymphony Network Edition
Metric Reported Results
SPC-1 IOPs 50,003.55
SPC-1 Price-Performance $6.11/SPC-1 IOPS™
Total ASU Capacity 1,407GB
Data Protection Level Mirroring
SPC-1 LRT 1.68 ms
Total TSC Price (including three-year maintenance)
$305,608
Fujitsu Softek SPC-1 Results (August 2003)
http://www.storageperformance.org/Results/SPC1/Fujitsu_2003_08_11_ETERNUS3000-M600M/2003-08-11_Fujitsu_ETERNUS3000-
M600M_SPC1-FDR.pdfSPC-1 Submission Identifier: A00016
Tested Storage Configuration (TSC) Name: Fujitsu Storage Systems ETERNUS 3000 Model 600M
Metric Reported Results
SPC-1 IOPs 64,249.77
SPC-1 Price-Performance $32.72/SPC-1 IOPS™
Total ASU Capacity 15,609GB
Data Protection Level Mirroring
SPC-1 LRT 2.31 ms
Total TSC Price (including three-year maintenance)
$2,102,147
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
$0$10$20$30$40
Price per SPC-1 IOPS
SP
C -
1 IO
PS
DataCore 3-nodeSun 63203PAR 8-nodeFujitsu ETERNUSIBM ESS 800HP EVA 2C12DLSI E4600
DataCore SANSymphony Cost/Performance
50,003SPC-1 IOPS™
50,003SPC-1 IOPS™
$6.11 perSPC-1 IOPS™
$6.11 perSPC-1 IOPS™
04/19/23
SANsymphony Performance
ConclusionsFC switches provide virtually unlimited bandwidth with exceptionally low latency so long as you do not cascade switches
General purpose Intel PCs are a great source of inexpensive MIPS.
In-band SAN management is not a CPU-bound process.
PCI bandwidth is the most significant bottleneck in the Intel architecture.
FC Interface cards speeds and feeds are also very significant
04/19/23
SAN Disk Subsystem Modeling framework:
ComponentsFC front end interface bandwidthInternal PCI bus bandwidthNumber of Intel processors and their speedCache memoryFibre Channel Disk interfacesSCSI Disks
Internal segmented buffering Seek, Rotational Delay, Data transfer rate
Logical:physical disk mappingLinear mappingDisk stripingRAID mapping
04/19/23
SAN Storage Manager – Next Steps
Cacheability of Unix and NT workloadsDomino, MS ExchangeOracle, SQL Server, Apache, IIS
Given mirrored writes, what is the effect of different physical disk configurations?JBODRAID 0 disk stripingRAID 5 write penalty
Asynchronous disk mirroring over long distances
Backup and Replication (snapshot)
04/19/23
??Questions
www.datacore.com