background image

 

17 

 

Based on the tests run using the ESRP – Storage framework, AMI 
recommends the following best practices to improve storage performance: 

 

 

Consider performance before capacity

RAID10 configurations are 

recommended  for  performance  and  fault-tolerance  mirroring 

(multiples of 2) RAID 5 configurations with more drives in an array 
are to be avoided for key volumes that has small random writes. Log 

devices  are  essentially  sequentially  written.  Write  back  caching  is 
very  helpful  here  as  it  reduces  latencies  from  20ms  to  2ms.  

StorTrends  iTX  with  an  advanced  cache  module  offers  excellent 

performance benefits.

 

 

Invest in high performance disks and spindles

:

 It is imperative to 

make  sure  that  volumes  are  distributed  across  multiple  spindles. 
StorTrends  iTX  software  houses  RAIDed  logical  drives  in  containers 

from  where  capacity  is  allotted  to  various  volumes.    Intelligent 

provisioning  strategies  allow  the  I/Os  to  be  appropriately  striped 
across various logical drives to ensure leadership in performance.

 

 

 

I/O  alignment

:

  In general MS Exchange I/Os are predominantly 4K 

in  size.  It  is  therefore  important  to  ensure  that  these  logical 
transaction  I/Os  are  not  split  into  different  spindles.  Splitting 

increases  physical  I/Os  and  therefore  degrades  performance.  By 

default Windows aligns the partition to 63

rd

 sector. This implies that 

one out of every 16 I/OS will be split, causing performance penalty. 

It is advisable to use Microsoft tool 

diskpart

 to align the volumes to 

4k alignment. 

“diskpart –i 1” 

command will ensure this alignment. 

 

 

Isolate network traffic

:

 Let LAN traffic live on LAN and SAN traffic 

on SAN network. Isolate the Microsoft Exchange database workload 

from other I/O intensive applications or workloads. This ensures the 

highest levels of performance for Microsoft Exchange and makes 

troubleshooting efforts easier in the event of a disk-related 

Microsoft Exchange performance issue. 

 

 

Multiple Paths

: With multiple NICs, Microsoft MPIO should be 

enabled on Exchange servers to ensure highly-available SAN 

connections and load-balancing & failover paths. 

 

 

Network switching

:

 

For best results, use non-blocking Gigabit 

Ethernet switches that support both Jumbo Frames and Flow 

Control. 

 

 

Application Isolation

Do not install Exchange server on a Domain 

Controller or ADS Server

.

 It is best to ensure application isolation 

and thereby improving performance. 

 

Summary of Contents for StorTrends 3200i

Page 1: ...iTX 2 7 600 User iSCSI Storage Solution for Microsoft Exchange Server 2007 Tested with ESRP Storage Version 2 1 Tested Date September 24 2008 American Megatrends Inc AMI 5555 Oakbrook Parkway Suite 200 Norcross GA 30093 770 246 8600 www ami com ...

Page 2: ...ary Storage Disk Configuration Transactional Log Disks 15 Streaming Backup 16 Disk Configuration Streaming Backup to disk 16 Exchange Best Practices 16 Core Storage Best Practices 16 Backup strategy 19 Test result summary 21 Reliability 21 Primary Storage Performance Results 21 Streaming Backup Recovery Performance 22 Database Read only Performance 22 Log Read only Performance 22 Backup to disk Pe...

Page 3: ... issues discussed as of the date of publication Due to changing market conditions it should not be interpreted to be a commitment on the part of American Megatrends and American Megatrends cannot guarantee the accuracy of any information presented after the date of publication Features AMI s StorTrends storage appliances are designed for both Small and Medium Businesses SMBs and enterprise level c...

Page 4: ...trators can simplify backup while ensuring faster more reliable restorations to Shrink or eliminate backup windows without affecting performance Improve Recovery Point Objectives while decreasing Recovery Time Objectives Minimize system and application downtime by seamlessly and non disruptively creating scheduled snapshots Eliminate performance degradation while taking snapshots Reduce if not eli...

Page 5: ...osts Our advanced data replication capability standard in each preinstalled StorTrends appliance provides point in time periodic replication to multiple remote sites This method easy to set up and operate delivers multiple restore points and application consistent copies that can be restored in minutes High Availability A high availability configuration can be created with a replicated secondary a...

Page 6: ...mally complex time consuming tasks Using StorTrends storage appliances to centralize storage and ManageTrends for centralized management of data not only reduces hardware and operating costs but also reduces the risk of having compromised or redundant information StorTrends uses iSCSI technology thereby allowing existing network personnel manage the new storage environment There is no need for ext...

Page 7: ...ID level migration High Availability HA Grouping Load Balancing Active Active Configuration Universal UPS Support Network Teaming Encryption Up to 16 iSNS servers are supported Compatible with MS iSNS Server v3 0 and later iSNS client supporting Draft 22 of iSNS specification Security ACL security implementation supports Local users Windows NT 2000 Domain users Windows 2003 Active Directory users ...

Page 8: ...s products are easy to implement and maintain and offer excellent TOC Total Operating Cost in comparison with Fibre Channel and other IP Storage solutions currently in the market AMI s StorTrends delivers and interoperates with all the above technologies of Microsoft and this document describes an approach that can be used to configure Exchange Solutions around AMI s StorTrends IP SAN series 1300i...

Page 9: ... which is used to simplify sizing and configuration when using a low number of disks to ensure the highest performance while remaining fault tolerant This unit of measure or building block is designed to be scalable and expandable into larger building block deployments based on customer I O and latency requirements The following configuration tables and diagram illustrates the building block layou...

Page 10: ...on data 300 GB Datatgt WriteBack Direct Read None log 10 GB Logtgt WriteBack Direct Read None backup 400 GB Backuptgt WriteBack Direct Read None Exchange JetStress Configuration Users Size MB Database Size GB IOPS User Thread Count 600 200 120 0 45 8 Storage Array Configuration single storage controller iSCSI Array Database Store File System Disk Settings Storage Controller settings Block Size K R...

Page 11: ... 2 RAID1 2 Disk Drives 500GB SATA 7200 RPM RAID Disk Group 3 RAID 5 3 Disk Drives 500GB SATA 7200 RPM Dell R905 Windows 2003 server x64bit Microsoft Exchange Jetstress 2007 600 Users 200MB Mailbox 0 45 IOPS StorTrends 3200i iTX 2 7 single storage controller Data LUN 300 GB 8 DISKS 1 SPARE Log LUN 10 GB 2 DISKS 1 SPARE Backup LUN 400 GB 3 DISKS 1 SPARE NIC 1 NIC 2 NIC 2 NIC 1 Multi Pathing I O Giga...

Page 12: ...must address The storage subsystem can sometimes be a complex area to understand and can require extensive domain expertise to get the best out of it A proper understanding and sizing of the storage subsystem with use of vendor specified best practice guidelines can thus help alleviate one of the most critical and common reasons for sub optimal Exchange performance This solution provides tested be...

Page 13: ...rage design However storage is not the only factor to take into consideration when designing a scale up Exchange solution Other factors which affect the server scalability are server processor utilization server physical and virtual memory limitations resource requirements for other applications directory and network service latencies network infrastructure limitations replication and recovery req...

Page 14: ...rmance characteristics change based on the percentage utilization of the individual disks Tests that use a small percentage of the storage 25 may exhibit reduced throughput if the storage capacity utilization is significantly increased beyond what is tested in this paper Primary Storage Hardware Storage Connectivity Fiber Channel SAS SATA iSCSI ISCSI Storage model and OS firmware revision StorTren...

Page 15: ...08 01 0240 005 Replication solution name version N A Primary Storage Disk Configuration Mailbox Store Disks Disk type speed and firmware revision SATA 3Gbps 7200 RPM Seagate Barracuda Rev SN04 Raw capacity per disk GB 465 GB Number of physical disks in test 8 1 spare total raw storage capacity GB 1858 92 GB Disk slice size GB N A Number of slices per LUN or number of disks per LUN N A Raid level R...

Page 16: ...intensive application Based on the testing run using the ESRP framework we would recommend the following to improve the storage performance For Exchange 2007 best practices on storage design please visit http technet microsoft com en us library bb124518 aspx Core Storage Best Practices In a MS Exchange environment where thousands of email transactions happen in the blink of an eye performance beco...

Page 17: ...is therefore important to ensure that these logical transaction I Os are not split into different spindles Splitting increases physical I Os and therefore degrades performance By default Windows aligns the partition to 63rd sector This implies that one out of every 16 I OS will be split causing performance penalty It is advisable to use Microsoft tool diskpart to align the volumes to 4k alignment ...

Page 18: ... monitoring and tuning offers the best storage solution StorTrends iTX stack also has built in automatic performance tuning load balancing and I O plugging capabilities enabling it to yield extremely high I O performance In such deployments to guard against data loss snapshots are extensively used The special Redirect on write algorithm used by iTX stack gives it a differentiating edge over the mu...

Page 19: ...additional degradation with the usage of multiple snapshots Provision tracking feature of iTX ensures that the snapshots are extremely space efficient StorTrends iTX VSS scheduler allows up to 8 different snapshot schedules For example one can have 60 snapshots minutes apart 24 hourly snapshots 30 daily snapshots and 24 monthly snapshots Figure 5 shows how with ease such schedules can be defined A...

Page 20: ...a mailbox store onto the running production Exchange 2003 server Data can be recovered within the restored mailbox store whilst the current store is still running Use of the RSG on a production server does not interfere with the users as the RSG is logically isolated users cannot log into it and mail cannot be delivered to it Now let us consider some common recovery scenarios Let us assume that so...

Page 21: ...will be much smaller than the restored production database Test result summary This section provides a high level summary of the test data from ESRP and the link to the detailed html reports which are generated by ESRP testing framework Please click on the underlined headings below to view the html report for each test Reliability A number of tests in the framework are to check Reliability tests r...

Page 22: ...only streaming backup type is supported for testing in the framework There are two tests in this section First one is to measure the read IO performance metrics by running checksum on all the databases and log files The second test is to measure the end to end performance when the databases are backed up to disks Database Read only Performance The test is to measure the maximum rate at which datab...

Page 23: ...ses to validate the storage design for a specific customer environment ESRP program is not designed to be a benchmarking program tests are not designed to getting the maximum throughput for a giving solution Rather it is focused on producing recommendations from vendors for Exchange application So the data presented in this document should not be used for direct comparisons among the solutions Con...

Page 24: ...tstress StorTrends_performance DBChe cksum_2008_9_24_12_54_24 blg Database Sizing and Throughput Achieved I O per Second 331 463 Target I O per Second 270 Initial database size 138868178944 Final database size 139755274240 Database files count 1 Jetstress System Parameters Thread count 8 per storage group Log buffers 9000 Minimum database cache 32 0 MB Maximum database cache 256 0 MB Insert operat...

Page 25: ...xchange Database JetstressCmd Database Cache Size Last 242450400 0 lower bound 241591900 0 upper bound none 9 24 2008 12 54 24 PM Performance logging ends 9 24 2008 12 54 24 PM JetInterop batch transaction stats 37922 9 24 2008 12 54 24 PM Dispatching transactions ends 9 24 2008 12 54 24 PM Shutting down databases 9 24 2008 12 54 24 PM Instance2284 1 complete 9 24 2008 12 54 25 PM Performance logg...

Page 26: ...0 000 0 000 0 000 0 007 82 555 Host System Performance Counter Average Minimum Maximum Processor Time 9 807 3 678 12 780 Available MBytes 8920 774 8835 000 9094 000 Free System Page Table Entries 16757673 000 16757673 000 16757673 000 Transition Pages RePurposed sec 2653 995 0 000 10140 790 Pool Nonpaged Bytes 77380770 670 73146368 000 79728640 000 Pool Paged Bytes 265029465 819 71245824 000 35367...

Page 27: ...logging ends 9 26 2008 3 45 50 PM Instance2440 1 100 processed 9 26 2008 3 45 50 PM C Program Files Exchange Jetstress StorTrends_StreamBackup StreamingBackup_2008_9_26_12_6_33 blg has 875 samples 9 26 2008 3 45 50 PM Creating test report ...

Page 28: ...0 000 0 000 Test Log 9 26 2008 10 25 58 AM Command Line JetstressCmd exe c Jetstress600_SoftRecovery xml 9 26 2008 10 25 58 AM Jetstress testing begins 9 26 2008 10 25 58 AM Prepare testing begins 9 26 2008 10 25 59 AM Attaching databases 9 26 2008 10 25 59 AM Prepare testing ends 9 26 2008 10 25 59 AM Dispatching transactions begins 9 26 2008 10 25 59 AM Database cache settings minimum 32 0 MB ma...

Page 29: ...1 15 07 AM Test has 0 Maximum Database Page Fault Stalls sec 9 26 2008 11 15 07 AM Test has 0 Database Page Fault Stalls sec samples higher than 0 9 26 2008 11 15 07 AM C Program Files Exchange Jetstress StorTrends_SoftRecovery Performance_2008_9_26_10_26_0 xml has 195 samples queried 9 26 2008 11 15 09 AM C Program Files Exchange Jetstress StorTrends_SoftRecovery Performance_2008_9_26_10_26_0 htm...

Page 30: ...08_9_26_10_26_0 blg Database Sizing and Throughput Achieved I O per Second 309 313 Target I O per Second 270 Initial database size 148630421504 Final database size 148888371200 Database files count 1 Jetstress System Parameters Thread count 8 per storage group Log buffers 9000 Minimum database cache 32 0 MB Maximum database cache 256 0 MB Insert operations 40 Delete operations 30 Replace operation...

Page 31: ...ncy thresholds average 0 01 seconds write maximum 0 05 seconds write 9 26 2008 10 26 01 AM Operation mix Sessions 8 Inserts 40 Deletes 30 Replaces 5 Reads 25 Lazy Commits 55 9 26 2008 10 26 01 AM Performance logging begins interval 15000 ms 9 26 2008 10 26 01 AM Generating log files 9 26 2008 11 15 03 AM E 100 2 generated 9 26 2008 11 15 04 AM Performance logging ends 9 26 2008 11 15 04 AM JetInte...

Page 32: ...ss StorTrends_stress DBChecksum _2008_9_25_22_7_22 blg Database Sizing and Throughput Achieved I O per Second 313 468 Target I O per Second 270 Initial database size 139755274240 Final database size 148630421504 Database files count 1 Jetstress System Parameters Thread count 8 per storage group Log buffers 9000 Minimum database cache 32 0 MB Maximum database cache 256 0 MB Insert operations 40 Del...

Page 33: ...s 30 Replaces 5 Reads 25 Lazy Commits 55 9 24 2008 10 01 08 PM Performance logging begins interval 15000 ms 9 24 2008 10 01 08 PM Attaining prerequisites 9 24 2008 10 07 20 PM MSExchange Database JetstressCmd Database Cache Size Last 242065400 0 lower bound 241591900 0 upper bound none 9 25 2008 10 07 21 PM Performance logging ends 9 25 2008 10 07 21 PM JetInterop batch transaction stats 393063 9 ...

Page 34: ...12 PM Test has 23 1860171334684 Maximum Database Page Fault Stalls sec 9 25 2008 10 22 12 PM Test has 0 Database Page Fault Stalls sec samples higher than 0 9 25 2008 10 22 12 PM C Program Files Exchange Jetstress StorTrends_stress Stress_2008_9_24_22_1_7 xml has 5756 samples queried ...

Reviews: