Midrange Storage Performance Planning: Identifying the Environment + Characterization of Workloads

Question 1

What is Live Optics?

Answer 1

a performance automation platform that enables you to gather analyze and view details of a system environment

Question 2

What are the three components of Live Optics?

Answer 2

Collector

Project

Web Portal

Question 3

What is the Live Optics Collector?

Answer 3

collects performance data from hosts/OS

allow array based performance data to be uploaded

Question 4

What is the Windows Collector called in Live Optics?

Answer 4

Optical Prime

Question 5

What is a Live Optics Project?

Answer 5

a container that gathers information from one or more collectors and organizes them into a hierarchy of objects

Question 6

What is the Live Optics Web Portal?

Answer 6

Where you can view Projects

always viewable by the customer

Question 7

How does an Optical Prime LO Capture work?

Answer 7

pulls performance and other data from the server/OS

supports up to 256 hosts per instance

Question 8

What is generated once a LO project is completed?

Answer 8

Project ID

Question 9

How does a Storage LO Capture work?

Answer 9

downloads performance archive files from the array

Question 10

What is the single most important performance consideration for storage sizing?

Answer 10

workload characterization

describes different types of workloads being pushed to the system

Question 11

What are the workload attributes?

Answer 11

IO Size
Read vs Write
Random vs Sequential
Working Set Size
Skew
Concurrency

Question 12

What is IO Size?

Answer 12

IO Request/Transfer Size

amount of data in each IO request by host

Question 13

What are typical IO sizes in environments?

Answer 13

8KB for typical file system/Oracle DB

32KB for Exchange

64KB for backup/restore process

256KB for video streaming

Question 14

What effect does IO size have on storage performance?

Answer 14

throughput

larger the IO size the higher the bandwidth

larger IOs take longer to transmit/process but some of the overhead to run IO is fixed

Question 15

What is true about an environment with a large IO size?

Answer 15

if data exists in larger chunks it’s more efficient to transfer larger blocks

host can move more data faster by using larger IOs than smaller ones

Question 16

How is response time related to IO size?

Answer 16

response time of each large IO transfer takes longer than single small transfer

combined service time of all small transfers is more than single large IO transfer that contains same amount of data

Question 17

What is read vs write?

Answer 17

read/write ratio

important to know which access is the majority for each workload

reads consume less resources than writes

Question 18

What type of reads consume the fewest resources in storage?

Answer 18

sequential reads that pull from cache

also have the highest throughput

reads not found in cache (typically in random access) have much lower throughput/response time since its grabbing data from the disk

Question 19

Why are writes typically slower and use more resources than reads?

Answer 19

because protection is typically added to new data

all writes must be cached, mirrored and acknowledged

calls for larger cache size for buffering writes

Question 20

What is random vs sequential?

Answer 20

two access patterns hosts use

Question 21

What is random access?

Answer 21

OLTP workloads

where data reads and modifications are made in a scattered manner across an entire dataset

Question 22

What is a random workload?

Answer 22

workload where reads and writes are distributed across the relevant address space

Question 23

Why are HDDs typically slower with random workloads?

Answer 23

requires drives to seek data across the platters which involves a slowed head movement

Question 24

What is sequential access?

Answer 24

successive reads or writes that are logically contiguous within the relevant address space

Question 25

When is sequential access typically done?

Answer 25

during backup and restore operations

event logging

Question 26

What do intelligent storage systems do to improve performance in sequential workloads?

Answer 26

detect patterns and pre-fetch data into cache

Question 27

How can storage systems improve performance on sequential writes?

Answer 27

can be coalesced when many smaller writes are combined into a few larger transfers to disk

Question 28

What is working set size?

Answer 28

portion of the total data space an application uses at a certain time - also known as active data

Question 29

What is skew?

Answer 29

when a percentage of the total storage capacity in a storage system is the target area for most IOPS

locality of active data within the storage system

Question 30

What is concurrency?

Answer 30

the ability to execute more than one application IO task/thread simultaneously

midrange systems require IO concurrency to deliver best performance

Question 31

What is required for a storage system to hit high random IO rates?

Answer 31

high thread counts per drive group

sequential access requires multiple threads as well

Question 32

What is the System Designer path in Midrange Sizer?

Answer 32

enables customer configurations for AF and Hybrid arrays

Question 33

What does System Design in Midrange Sizer use to build a solution?

Answer 33

workload types

block size

IOPS requirements

storage capacity requirements (drives/data reduction)

Question 34

What is the performance prediction calculated from on System Design on Midrange Sizer?

Answer 34

computes maximum drive IOPS based on classic drive IOPS (sum of all drives) + RAID overhead

displays the lowest results and assumes ports/buses are not a limit nor response time

Question 35

What is the Live Optics/NAR path in Midrange Sizer?

Answer 35

enables the creation of a performance solution for a system

uses historical performance archive files (NAR) + LO

Question 36

What is the output in the Live Optics/NAR path in Midrange Sizer based on?

Answer 36

which LUNs work together in the pool

LUNs grouped based on IO Size, access patterns and performance rates

Question 37

What is the Simple Performance path in Midrange Sizer?

Answer 37

enables the sizing of applications with minimal input at early stages before complete workload details are available

predicts the decay of data access over time that is based on the inputs

Question 38

How does the Simple Performance path in Midrange Sizer calculate a solution?

Answer 38

computes a skew value and determines how best to build a pool to efficiently capture that data

determines IOPS rate achievable based on the drives being no more than 70% busy

Question 39

What are the application detail options in Simple Performance path in Midrange Sizer?

Answer 39

• Small read/write mix – Application has 67% read and 33% write mix with 8 KB I/O size.
• Medium read/write – Application has 67% read and 33% write along with 64 KB I/O size.
• Small read heavy – Application has 80% read and 20% write with 8 KB I/O size.

Question 40

What are the RAID options in Simple Performance path in Midrange Sizer for different drive types?

Answer 40

Flash - RAID 5 (8+1)

SAS - RAID 5 (4+1)

NL-SAS - RAID 6 (6+2)

Question 41

What is the Application-Oriented path in Midrange Sizer?

Answer 41

enables the sizing of configurations that are based on the workloads of the applications

enables you to define multiple applications and workloads within a single configuration

Question 42

What is the Advanced Performance path in Midrange Sizer?

Answer 42

enables the sizing of a Dell Unity system by creating pools which contain manually defined loads

models single tiered and multi tiered pools and can estimate system saturation and response time for those loads