virtual infrastructure optimization san transparency and performance from reactive to proactive alex...
Post on 20-Dec-2015
219 views
TRANSCRIPT
Virtual Infrastructure Optimization
SAN Transparency and Performance
From Reactive to Proactive
Alex D’Anna
Director, Solutions Consulting, EMEA
November 9, 2010
Agenda
SAN & Virtualization Challenges
Virtual Infrastructure Optimization
Application Views and Risk Reduction
Customer Examples and Deployment
About Virtual Instruments
• Focus on optimizing Fibre channel
• Leader in Virtual Infrastructure Optimization
• Private equity spinout from Finisar: June 2008
• Virtual Instruments Leadership− John Thompson, former CEO of Symantec and
Director of IBM Americas
− Barry Cooks, Engineering of VMware
− Former Siebel Leadership
− Key Finisar Engineering
• Key partnerships: Brocade, HDS, VMware, IBM, LB Systems, MEN@NET
• Growing 2X Year over Year
• In EMEA: Nov. 2009 2 Dec. 2010 17
San Jose, CA Headquarters
About Virtual Instruments
• Where to find us?
• LB Systems and MEN@NET!!!
• Full lab, demo and offer the services and capabilities to deploy
• Where on the Web?
• LinkedIn Group: Virtual Instruments SAN Storage and Virtualization Forum
• Twitter: virtual_inst, virtual_wisdom, virtual_io
• YouTube: SNW Europe 2010 or http://www.youtube.com/user/sos4sans#p/a/u/0/1dnhEHKnWLE
San Jose, CA Headquarters
The Industry Challenge…The Industry Challenge…
...the “perfect storm”...the “perfect storm”
I/O
I/O
The Virtualization Challenge
1. The SAN has lacked any real I/O systems-level performance
– Original FC spec was designed for 32 “storage channels”
– Not designed as a “network”
– Lacks self-health, diagnostics and transparency to the I/O
FC Fabric
There’s a “perfect storm” happening in data management today…
Servers & Virtual
Machines
StorageArrays
SAN Cloud
Servers & Virtual
Machines
1. The SAN has lacked any real I/O systems-level performance
2. Data growth at an unprecedented rate (average 30-60% CAGR)
– A 200TB shop in ‘05 growing 50% is now 1PB & will be about 8 PB in 5 years
– A net-new 7 PB of storage; how much will it cost, and where will it be deployed?
There’s a “perfect storm” happening in data management today…
SAN Cloud
The Virtualization Challenge
1. The SAN has been a “black box”, lacking any real I/O systems-level performance, so it’s heavily over-provisioned as a result
2. Data growth at an unprecedented rate (average 30-60% CAGR)
3. More “abstraction” being added
– Further limits I/O visibility
– Challenges performance
– Slows deployment of cloud infrastructures
There’s a “perfect storm” happening in data management today…
Virtual Server Cloud
SAN Cloud
Storage Virtualization Cloud
The Virtualization Challenge
Common Large-scale SAN Challenges
• Explaining/avoiding application outages & slowdowns
• Identifying SAN problems
• Identifying physical layer problems
• Reducing vendor finger-pointing
• Tracking SLAs & compliance
• Over-provisioning and consolidation
• Storage tiering
• Environmental costs (avoiding new data centers)
• Capacity planning
• Containing rising costs of storage/SAN w/ flat budget
• Explaining/avoiding application outages & slowdowns
• Increasing server consolidation ratios
• Reducing vendor finger-pointing
• Tracking SLAs & compliance
• I/O subsystem troubleshooting
• Deploying Tier 1 mission critical applications
• Showing adherence to performance standards
• Isolating workload peaks that cause resource conflicts and bottlenecks
Common Virtual Infrastructure Challenges
The primary virtual infrastructure challenge
We have found greater than 90 percent of the
VMware-related performance issues
encountered by our customers are due to the
storage tier.
Scott Drummonds,Performance SpecialistVMware
Virtual Server Market Share 2008-2012
~ 55M vms~ 10M vms
Process and Tech Standard Phase
• “VM 1st” Policy
Heavy-Use Phase• Mission Critical • More than just
Servers
Light-Use Phase• “Virtualization-
Lite”Pilot Phase
• Play
TIME
NU
MB
ER
OF
VM
s
Are You Here?
Phases of VMware Infrastructure
Stuck due to:•Lack of “know-how”•Lack of Tier 1 app confidence•Lack of client virtualization maturity
Why Do Customers STOP Here??
VISIBILITY….of I/O
Identify / fix physical & virtual infrastructure problems before they occur
Ensure no loss of revenue/ productivityReduce Risk
Optimize IT asset utilization and personnelReduce Costs
Tier 1 apps meet performance SLAsImprove
Performance
What is needed…
Create “Predictability”
ProbeV• Identifies low overall SAN utilization via real-time dashboard• Identifies individual port utilization• Enables verification of historical utilization trends to verify loads over time• Enables intelligent load balancing to avoid expensive purchases
Avoiding Over-provisioning of Links
90% of ports used less than 10%
Improving SAN Utilization and Mitigating Risk
• SAN utilization < 2%
• Some links hitting 100%
• Traffic on ISL’s causing contention
• SFP low-light levels & flopping HBA’s causing CRC issues
Categorization Summary Count % of LinksBalanced 1228 69%
Passive 85 5%
Active 85 5%
Imbalanced 228 13%
Single (not redundant) 143 8%
ProbeV Software Audit
Record and play back metric recordings of intermittent
problems before they build up and disrupt the SAN
Faster Troubleshooting & Root Cause Analysis
ProbeFCX
• Continuously monitors and filters in real-time
• Calculates statistics based on measuring all fibre channel frame traffic
• Automatically notifies staff based on exceeded policy thresholds
Real-time root-cause analysis
Avoiding Performance Problems
ProbeFCX• Identifies potential application slow-down causes
• Recommends corrective action before the slowdown
• Enables fixes before application owner is aware of the problem
Provides visibility into Queue depths, CRC errors, physical link errors, protocol errors, code violations, etc
Optimizing Application Performance
ProbeFCX• Measures all network statistics
• Proactively alerts administrator based on policies
• Enables real-time tuning for maximum performance
Expanding VMware to Mission-critical Applications
ProbeVM• Monitors CPU, memory & SAN utilization and I/O response time
• Identifies performance bottlenecks & recommends vMotion transfers
• Enables “what if” load balancing simulations
• Proves consolidation ratios can be improved
w/out performance degradation
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
OS
APP
&Hosts
FC Switches
StorageArrays
ProbeV(SNMP data)
ProbeVM(VMware vCenter)
VirtualWisdom Deployment
ProbeV (software)
TAPs Probe FCX
ProbeVM (software)
ProbeFCX: (Real-time latency via FC headers) Traffic Access Point (TAP) Patch Panel
(Out-of-band copy of FC traffic)
Solution Example: Virtual Instruments
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
Server, GUI, Dashboards
&Hosts
StorageArrays
Solution Deployment
Comprehensive I/O Visibility is Essential
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
FCTAPs
SAN switches
Representative infrastructure
&Hosts
StorageArrays
Solution Deployment
Extract CPU, Memory data from
vCenter
Phase 1: Virtual Server Monitoring
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
FCTAPs
SAN switches
&Hosts
StorageArrays
Solution Deployment
Extract CPU, memory data from
vCenter
Extractdata from
FC switches
Phase 2: SAN Switch Monitoring
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
FCTAPs
SAN switches
&Hosts
StorageArrays
VirtualWisdom Deployment
Extract CPU, memory data from vCenter
Extract data from
FC switches
Extractdata from
FC frames
Phase 3: Fibre Channel Link Monitoring
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
FCTAPs
SAN switches
Everyone will TAP at Some Point
Traffic Access Points (TAPs):• Have been widely deployed in IP networks (LANs, WANs) for 20+ years
• Provide direct access to all levels of fiber traffic data on SAN/storage performance, utilization, and transmission errors
• “If I could make 1 Recommendation, it’s TAP every Storage Array you deploy”
– IBM Global Escalation Engineer
Faster problem identification & resolution
Proactively find problems before users
Maximize application performance
Other Options for TAPping
TAPping Integrated into the Cabling
&Hosts
StorageArrays
Solution Deployment
Virtual Server Monitoring
SAN Switch Monitoring
FC Physical Layer Monitoring
ConsolidatedView
Comprehensive I/O Visibility: VM to the LUN
Guests
OS
APP
OS
APPOS
APP
OS
APPOS
APP
OS
APP
SAN switches
VM to LUN Correlation
FCTAPs
Customer Example
SAN & Virtualization Challenges
Virtual Infrastructure Optimization
Application Views and Risk Reduction
Customer Examples and Deployment
Installed in 1.5 hours… on March 15, 2010
Multipath Verification • Verification including all Nicknames. The single HBA should be investigated.
Multipath Verification • MP after removing nicknames including the word TAPE . The single HBAs should be investigated.
• Increasing production virtual server deployments
• Application performance degradation• Inability to agree on root causes between
storage/server admins & vendors• Additional storage capacity/bandwidth
failed to resolve problemsSolutions
Results
Challenge
• Implemented VIO solution across server & storage tiersChallengeSolutionsResults • Detection of VMware configuration problems
• Diagnosis of storage I/O latency
• Identification of overloaded “hot” ports
• Correlation between VMware vMotion and performance degradation
Medium Bank 250 VM’s on 24 ESX Servers
Customer Success Story
Summary
• Comprehensive I/O visibility enables
– Real-time performance optimization
– Proactive re-balancing of applications/VMs
– Faster troubleshooting
– Higher infrastructure availability
– Confidence to deploy VMware with I/O-intensive Tier 1 business-critical applications
The Leader In SAN & Virtual Infrastructure Optimization
THANK YOU