Scribd
Upload
750 views

SAN Problems: Application IOPS and RAID Configurations

The document discusses input/output operations per second (IOPS) and how they are impacted by different RAID configurations. It provides write penalties for RAID 1 (2x), RAID 10 (2x), RAID 5 (4x), and RAID 6 (6x). These penalties determine how many disk operations are required for each write operation based on the RAID level. The document also includes two example problems calculating disk loads and number of disks needed for applications with given IOPS, read/write ratios, and RAID configurations.

Uploaded by

Arsalan Ozil
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
750 views

SAN Problems: Application IOPS and RAID Configurations

The document discusses input/output operations per second (IOPS) and how they are impacted by different RAID configurations. It provides write penalties for RAID 1 (2x), RAID 10 (2x), RAID 5 (4x), and RAID 6 (6x). These penalties determine how many disk operations are required for each write operation based on the RAID level. The document also includes two example problems calculating disk loads and number of disks needed for applications with given IOPS, read/write ratios, and RAID configurations.

Uploaded by

Arsalan Ozil
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

SAN Problems

Application IOPS and RAID Configurations

 When deciding the number of disks required for an application, it is important to


consider the impact of RAID based on IOPS (Input / Output Per Second) generated
by the application.
 The total disk load should be computed by considering the type of RAID
configuration and the ratio of read compared to write from the host.
Write penalty in RAID 1 is 2
Write penalty in RAID 1/0 is 2
Write penalty in RAID 5 is 4
Write penalty in RAID 6 is 6
***************************************************************************
Example Problem 01:
Consider an application that generates 5,200 IOPS, with 60 percent of them being reads.
a. Calculate the disk load in RAID 5, RAID 1 and RAID 6.
b. Calculate the number of disks required for the application. HDD (hard Disk Drive)
with a specification of a maximum 180 IOPS is used.
Solution:
According to the given problem statement among 5200 IOPS 60% is read. Than 40% is write.
i.e 60% of 5200 = 0.6 * 5200 = 3120 reads
40% of 5200 = 0.4 * 5200 = 2080 writes
Disk Load in RAID 5:
With respect to write:
In RAID 5 the write penalty is 4. i.e every 1 write instruction results in 4 write operations.
With respect to read:
In RAID 5 1 read instruction results in 1 read operation.
So, disk load in RAID 5 is = read load on disk + write load on disk
read load on disk = 0.6 * 5200 = 3120
write load on disk = 4 * (0.4 * 5200) = 8320
Disk load in RAID 5 is 3120 + 8320 = 11400 IOPS.
The computed disk load determines the number of disks required for the application. The
number of disks required to meet the workload for the RAID configuration would be as
follows:
RAID 5: 11,440 / 180 = 64 disks
Disk Load in RAID 1:
With respect to write:
In RAID 1 the write penalty is 2. i.e every 1 write instruction results in 2 write operations.
With respect to read:
In RAID 1, 1 read instruction results in 1 read operation.
So, disk load in RAID 1 is = read load on disk + write load on disk
read load on disk = 0.6 * 5200 = 3120
write load on disk = 2 * (0.4 * 5200) = 4160
Disk load in RAID 5 is 3120 + 4160 = 7280 IOPS.
The computed disk load determines the number of disks required for the application. The
number of disks required to meet the workload for the RAID configuration would be as
follows:
RAID 1: 7280 / 180 = 40.444 disks, rounded off to 42 disks (approximated to the nearest
even number)

Disk Load in RAID 6:


With respect to write:
In RAID 6 the write penalty is 6. i.e every 1 write instruction results in 6 write operations.
With respect to read:
In RAID 6 1 read instruction results in 1 read operation.
So, disk load in RAID 6 is = read load on disk + write load on disk
read load on disk = 0.6 * 5200 = 3120
write load on disk = 6 * (0.4 * 5200) = 12480
Disk load in RAID 6 is 3120 + 12480 = 15600 IOPS.
The computed disk load determines the number of disks required for the application. The
number of disks required to meet the workload for the RAID configuration would be as
follows:
RAID 6: 15600 / 180 = 86.666 disks, rounded off to 88 disks (approximated to the nearest
even number)
Example Problem 02
An application has 1,000 heavy users at a peak of 2 IOPS each and 2,000 typical users at a
peak of 1 IOPS each, with a read/write ratio of 2:1. It is estimated that the application also
experiences an overhead of 10 percent for other workloads. Calculate the IOPS requirement
for RAID 1, RAID 5 and RAID 6. compute the number of drives required to support the
application in different RAID environments if HDD with specification of 130 IOPS is used.
Solution:
The application has 1000 heavy users with 2 IOPS each. i.e 1000*2=2000 IOPS and
the application has 2000 typical users with 1 IOPS each i.e 2000*1 = 2000 IOPS.
So totally 2000 + 2000 = 4000 IOPS. Also the given problem states that application
experiences an overhead of 10%. i.e 10% overhead above 4000 IOPS.
So 10% of 4000 IOPS is over head.
0.1*4000 = 400. So overall 4000 + 400 = 4400 IOPS the application generates.
Now among 4400 IOPS how many are write and how many are read?
According to problem statement read/write ratio is 2:1. So Ratio 2:1 implies that
approximately 66.6% are reads and 33.3% are writes.
read load on disk is 66.6% of 4400 = 0.66 * 4400 = 2904
write load on disk is 33.3% of 4400 = 1452
RAID 1:
In RAID 1 the write penalty is 2. i.e every 1 write instruction results in 2 write operations.
So, disk load in RAID 1 is = read load on disk + write load on disk
read load on disk = 0.66 * 4400 = 2904
write load on disk = 2 * (0.33 * 4400) = 2904
Disk load in RAID 1 is 2904 + 2904 = 5808 IOPS.
The number of HDD required is 5808 / 130 = 44.67 round off to 46 disks.
RAID 5:
In RAID 5 the write penalty is 4. i.e every 1 write instruction results in 4 write operations.
So, disk load in RAID 5 is = read load on disk + write load on disk
read load on disk = 0.66* 4400 = 2904
write load on disk = 4 * (0.33 * 4400) = 5808
Disk load in RAID 5 is 2904 + 5808 = 8712 IOPS
The number of HDD required is 8712 / 130 = 67.05 round off to 68 disks
RAID 6:
In RAID 6 the write penalty is 6. i.e every 1 write instruction results in 6 write operations.
So, disk load in RAID 5 is = read load on disk + write load on disk
read load on disk = 0.66* 4400 = 2904
write load on disk = 6 * (0.33 * 4400) = 8712
Disk load in RAID 6 is 2904 + 8712 = 11616 IOPS
The number of HDD required is 11616 / 130 = 89.38 round off to 90 disks.

PTO
Application Require ments and Disk Performance
Little’s Law is a fundamental law describing the relationship between the number of requests
in a queue and the response time. The law states the following relation
N=a×R
where
“N” is the total number of requests in the queuing system (requests in the queue + requests in
the I/O controller)
“a” is the arrival rate, or the number of I/O requests that arrive to the system per unit of time
“R” is the average response time or the turnaround time for an I/O request — the total time
from arrival to departure from the system
Utilization law is another important law that defines the I/O controller utilization. This law
states the relation:
U = a × RS
where
“U” is the I/O controller utilization
“Rs ” is the service time, or the average time spent by a request on the controller. 1/Rs is the
service rate.
From the arrival rate “a”, the average inter-arrival time, Ra, can be computed as:
Ra = 1/a
Consequently, utilization can be defined as the ratio of the service time to the average inter-
arrival time, and is expressed as:
U = R s /Ra
Example Problem 03
Consider a disk I/O system in which an I/O request arrives at the rate of 80 IOPS. The disk
service time is 6 ms. Compute the following:
i. Utilization of I/O controller ii. Total response time
iii. Average queue size iv. Total time spent by a request in a queue
Solution:
i. Arrival rate a = 80 IOPS (Input / Output Pe r Second).
Service time (Rs) = 6ms
Utilization of I/O controller U =?
U = a × Rs
Rs service time is given in milliseconds and arrival rate ‘a’ is given in seconds so convert a to
milliseconds.
a = 80 IOPS => 0.08 I/O per millisecond
U = 6 * 0.08
=0.48 or 48% utilization of I/O Controller.
ii. Total response time
Total Response Time (R) is given by:
R = Rs / (1-U)
= 6 / (1-0.48)
= 11.5 ms.
iii. Average queue size
Average Queue Size = (U*U) / (1-U)
=> (.48*.48) / (1-0.48)
=> (.2304) / (.52)
=> 0.44
iv. Total time spent by a request in a queue
Total Time Spent by a request in a Queue = U * R
=> 0.48 * 11.5
=> 5.52 milliseconds.

You might also like