Size Clariion

Hello,

There is a chapter on 'Storage System Sizing and Performance Planning' in EMC CLARiiON Best Practices for Performance and Availability - Release 29.0 Firmware Update available on Powerlink. You may also want to read up EMC CLARiiON Fundamentals for Performance and Availability for basic concepts on storage fundamentals.

Hi I am posting a cut out from a document that i prepared comprising of best practices for clariionb boxes. Hope you find it suitable.

Disk Drive Considerations

·         Drives with higher rpm (revolutions per minute) provide overall random access throughput and shorter response times , therefore for higher performance we should use higher rpm drives.·         Incase of a database it is always recommended that FC Drives should be used for storing data files and online redo logs.

·         SATA drives has lower rpm and less expensive  as compared to FC drives so SATA drives are the best option  for  storing archived redo logs and the flashback recovery area.

I/O Size

Small Block random access applications like Oracle , it is important to use IOPS ratings as an important factor as they are disk based , not bandwidth ratings that are bus based.

Concurrency and Asynchronous I/O

For Oracle higher concurrency is better. The storage system gains performance , high throughput when there is lots of concurrency. Using Asynchronous I/O is a common technique to achieve high concurrency rate  .

File System Fragmentation

Fragmented File Systems decreases the opportunity for sequential I/O , which reduces overall throughput therefore the file systems should be defragmented at a fixed interval of time may be a month using host utilities. Note if the file system is a NTFS one then the file system cannot be formatted at anything but default extent size. Performing a file level copy i.e. to another LUN or by executing a backup and restore of the file system.File System Alignment affects the performance in two ways:-·         Misalignment causes disk crossings , i.e. an I/O broken across two disk drives.·         Misalignment makes it hard to stripe-align large un cached writes.

It is recommended that operating system disk utility should be used to adjust partitions. For  Oracle and OLTP applications , the volume manager stripe element should be set to the Clariion Stripe Size typically 128 KB or 512 KB.

High Availability and OCFS2 Filesystem

The storage processor failover capability is a key feature unique to Clariion CX3 storage systems. This feature offers redundancy at the storage array level.Unfortunately, the default parameters in the OCFS2 file system are incorrect . If these settings are not corrected this can result into the complete hang of the Oracle Cluster in case of SP failover. Therefore we must modify the following parameter in case you use OCFS2.O2CB_HEARTBEAT_THRESHHOLD = 60

Jumbo frames for the RAC interconnect

Maximum Transfer Unit (MTU) sizes of greater than 1,500 bytes are referred to as   jumbo frames  . Jumbo frames require gigabit Ethernet across the entire network infrastructure—server,   switches, and database servers. EMC recommends the use of jumbo frames on all legs of the RAC interconnect networks. For Oracle RAC 10  g   installations, jumbo frames are recommended for the private RAC interconnect to boost the throughput as well as to possibly lower the CPU utilization due to the software overhead of the bonding devices. Jumbo frames increase the device MTU size   to a larger value (typically 9,000 bytes). Oracle database environments transfer data in 8 KB and 32 KB block sizes, which require multiple 1,500 frames per database I/O, while using an MTU size of 1,500. Using jumbo frames reduces the number of frames needed for every large I/O request and thus reduces the host CPU needed to   generate a large number of interrupts for each application I/O. The benefit of   jumbo frames is primarily a complex function of the  workload I/O sizes, network utilization, and Oracle database server CPU utilization, and hence not easy to predict.

Recommendations for Control file and log files

Control files It is recommended  that when you create the control   file, allow for growth by setting MAXINSTANCES, MAXDATAFILES, MAXLOGFILES, and MAXLOGMEMBERS to high values. It is also recommended that the database has a minimum of two control files located on separate physical ASM diskgroups. One way to multiplex the control files is to store a control file copy on every diskgroup that stores members of the redo log groups. Online and archived redo log files EMC recommends that you run a mission-critical, production database in ARCHIVELOG mode. EMC also recommends that you multiplex your redo log files for these databases. Loss of online redo log files could result in failure of the database being able to recover. The best practice to multiplex your online redo log files is to place members of a redo log group on different ASM diskgroups.

Performance Planning

Determining the Workload

The application is expected to execute random I/O’s. The profile is·         50 % Random Read Requests .·         50% Random Write Requests.·         Predominant I/O size is 8 KB·         Total Host IOPS is 1200.

Determining the required I/O disk load.

For RAID 1/0 (recommended ) Read I/0*IOPS + 2(Write I/O*IOPS) = 0.5*1200 + 2(0.5*1200) = 1800 IOPSFor RAID 5Read I/0*IOPS + 2(Write I/O*IOPS) = 0.5*1200 + 4(0.5*1200) = 3000 IOPS

Determining the number of drives required.

No of Disk Drives =1800/140  =13 Disk Drives + 20% overhead = 16 Drives (RAID 1/0)No of Disk Drives = 3000/140  =21 Disk Drives + 20% overhead = 26 Drives (RAID 5)(Assuming 300 GB 10000 K RPM Drive with 140 IOPS /disk.)

Storage Allocation

RAID 1/0 Usable Capacity 16 Drives = 2400 GBRAID 5 Usable Capacity 26 Drives (RAID 15+1) & (RAID 9+1) = 7200 GB(Assuming 300 GB 10000 K RPM Disk Drive with 140 IOPS /disk.)

Why RAID 1/0

RAID 1/0 provides the best performance on workloads that use very small, random, and write intensive I/O . A write intensive workloads operations consist of greater than 30 percent random writes. A RAID 1/0 also offers performance advantages during certain degraded modes .These conditions are when write cache is disabled or when a disk has failed in a RAID Group. It combines the performance of RAID 0 and protection of RAID 1. A RAID 1/0 group provides the best balance of performance and availabi

I posted this on another forum.  I just made a chart to give me a quick view when planning.  If you have an existing server with local disks for this, take a look at perfmon to see how many write/read/total IOPS are going to that LUN.

If you have a server already on a CLARiiON, just look at Analyzer (if you have a license, if not perfmon again)

If this is a brand, spanking new server, I'd hit up the vendor for their requirements/recommendations.

If you really have no clue how many IOPS you need, I'd make an educated guess, overprovision spindles, then analyze it over the course of a couple of weeks, then migrate that LUN to a more appropriately sized RAID group.

I assume 15k = 180 IOPS

10k = 130 IOPS

	Fibre Channel 15k								Fibre Channel 10k
Disk Count	R5 100% Write Available IOPS	R5 100% Read Available IOPS	R5 20% Write 80% Read Available IOPS	R5 33% Write 67% Read Available IOPS	R1+0 100% Write Available IOPS	R1+0 100% Read Available IOPS	R1+0 20% Write 80% Read Available IOPS	R1+0 33% Write 67% Read Available IOPS	R5 100% Write Available IOPS	R5 100% Read Available IOPS	R5 20% Write 80% Read Available IOPS	R5 33% Write 67% Read Available IOPS	R1+0 100% Write Available IOPS	R1+0 100% Read Available IOPS	R1+0 20% Write 80% Read Available IOPS	R1+0 33% Write 67% Read Available IOPS
2					180	360	300.00	270.68					130	260	216.67	195.49
3	135	540	337.50	271.36					97.5	390	243.75	195.98
4	180	720	450.00	361.81	360	720	600.00	541.35	130	520	325.00	261.31	260	520	433.33	390.98
5	225	900	562.50	452.26					162.5	650	406.25	326.63
6	270	1080	675.00	542.71	540	1080	900.00	812.03	195	780	487.50	391.96	390	780	650.00	586.47
7	315	1260	787.50	633.17					227.5	910	568.75	457.29
8	360	1440	900.00	723.62	720	1440	1200.00	1082.71	260	1040	650.00	522.61	520	1040	866.67	781.95
9	405	1620	1012.50	814.07					292.5	1170	731.25	587.94
10	450	1800	1125.00	904.52	900	1800	1500.00	1353.38	325	1300	812.50	653.27	650	1300	1083.33	977.44
11	495	1980	1237.50	994.97					357.5	1430	893.75	718.59
12	540	2160	1350.00	1085.43	1080	2160	1800.00	1624.06	390	1560	975.00	783.92	780	1560	1300.00	1172.93
13	585	2340	1462.50	1175.88					422.5	1690	1056.25	849.25
14	630	2520	1575.00	1266.33	1260	2520	2100.00	1894.74	455	1820	1137.50	914.57	910	1820	1516.67	1368.42