47
Copyright © 2014 DataCore Software Corp. – All Rights Reserved. Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 1 Parallel IO and Hyper- converged Performance Enough Talk, Just Prove It! Sushant Rao, Sr. Director of Product Marketing

DataCore Software Parallel IO and Hyper-converged Performance

Embed Size (px)

Citation preview

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

1

Parallel IO and Hyper-converged PerformanceEnough Talk, Just Prove It!Sushant Rao, Sr. Director of Product Marketing

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 2

The Farce of Hyper-converged Blog Wars – No Real Numbers, Just Talk!

Nutanix appears to be charging a very hefty premium for largely equivalent configurations. Even with aggressive discounting, there’s a vast gulf in relative pricing.

Chuck Hollis, VMware

EVO:RAIL is struggling and close to getting dumped. VSAN has seen a bit more success but it isn’t a substantial part of VMware’s business, or a replacement for EMC’s declining

sales.

Lukas Lundell, Nutanix

Nutanix offers substandard performance running data center workloads when compared to both external storage arrays and especially an equivalently configured VSAN cluster.

Chuck Hollis, VMware

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 3

Criteria for Benchmarking Hyper-converged Performance

CriteriaIndustry Standard

Independently Verified & Audited

Peer Reviewed

Covers different types & generations of technology

Maps to “real world” performance (OLTP databases)

Shows cost for achieving performance level

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Storage (I/O) is the Bottleneck,especially for Virtualized Infrastructure

1990 2000 2010 2020

Performance gap between Compute

& Storage

Compute vs StoragePerformance

Yearly Performance Gains

Compute: 26%

Storage: 2%

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 5

SPC is the Right BenchmarkCriteria SPC BenchmarkIndustry StandardIndependently Verified & AuditedPeer ReviewedCovers different types & generations of technologyMaps to “real world” performance (OLTP databases)Shows cost for achieving performance level

Major Storage Vendors with SPC Benchmarks

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 6

SPC is a Torture Test for Hyper-Converged

Hyper-Converged must be Powerful enough to do Both!

Stresses Enterprise Database (OLTP) Applications + their I/O loads

Hyper-Converged

VMVMVMVM VMVMVMVMVMVMVMVM

Separate Servers & Storage

Separate servers create

I/O load for storage system

Storage system only has to focus on handling I/O

load

Same servers creating I/O load need to

handle the I/O load!

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 7

100K

200K

300K

400K

$50K $100K $150K $200K

SPC-1 IOPS

PRICE

DellCompellent

SC4020

NetAppEF560

AFA

EMCVNX 8000

500K

HP3PAR7400

Major Storage BrandsPrice Performance Leaders

Price & Performance go hand in hand

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 8

100K

200K

300K

400K

$50K $100K $150K $200K

SPC-1 IOPS

PRICE

DellCompellent

SC4020

NetAppEF560

AFA

EMCVNX 8000

455K

DataCore500K

HP3PAR7400

Comparison with Major Storage Brands

Price Performance Leaders

DataCore Price Performance is Off the Curve

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 9

Price Performance is Even Better when All Costs are Included

Hyper-ConvergedSeparate Server & Storage Components

Servers not included in

price

VNX 8000435K SPC IOPS

$177K(partial price)

$36K(all inclusive price)

Nodes include compute &

storage

455K SPC IOPS

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Industry standard, independently verified benchmark for I/O performance

► Can be compared with storage systems as well

DataCore hyper-converged can be used for mission- & business-critical workloads

► Suitable for Oracle, MS SQL Server & SAP HANA► Removing I/O Bottleneck means higher VM density

• More VMs on same hardware for VDI

Other benchmarks (VMmark, IOmark, etc) under consideration10

Importance of SPC for Hyper-converged

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Big Picture: DataCore Parallel I/O Architecture

11

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

IO Gap

12

IO-Starved Virtualized Servers

Increasingly faster Uni-processors

Com

pute

Serial IO

WorkPotential

2010 20202000

CPU clock ratesslow down

More cores per socket

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Blended Enterprise Application WorkloadsDramatically Expose the ‘I/O Straw’ Bottleneck

Hypervisor (VMware, KVM, Hyper-V…)

VM 2 VM 3 VM 4VM1

I/O througha straw

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 14

Serial vs. Parallel Processing

Time to finish1 worker(Serial)

Pile of work

5workers

(Parallel)

Load 1

Load 2

Load 3

Load 4

Load 5

Load 1 Load 2 Load 3 Load 4 Load 5

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 15

Modern Multi-core CPUs

Worker1

Worker2

Worker3

Worker4

Worker5

Worker6

Worker7

Worker8

Worker9

Worker10

Multiple “workers” capable of simultaneously handling compute, networking and I/O loads

10-cores

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 16

Standard use of Multi-core CPUsin Virtual Servers

VM1

VM2

VM3

VM4

VM5

idle I/Oidle idle idle

ParallelCompute

Serial I/O

VM = Virtual Machine

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

idle

17

Serial I/O Bottleneck in Virtualized Server

idleidleidleidle I/O

Compute

I/O

Compute waits on I/O CPU cores are wasted Very little work gets done

Workload

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 18

Impact: Many servers needed to spread I/O

Workload

Server 2 Server 3 Server 4 Server 5Server 1

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 19

I/OI/OI/OI/O

Turbo-Charge through Parallel I/O

Compute

I/O

Workload I/O keeps pace with

compute demands CPU cores are fully used Lots of work gets done in

very little timeI/O

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Worker1

Worker2

Worker3

Worker4

Worker5

Worker6

Worker7

Worker8

Worker9

Worker10

DataCore’s Adaptive use of Multi-core CPUs in Virtual Servers

VM1

VM2

VM3

VM4

VM5

I/O

ParallelCompute

Parallel I/O

VM = Virtual Machine

I/OI/O I/O I/O

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Translations:

21

Work completes in 1/5th the time

2 machines can do the work of 10

\ˈper-ə-ˌlel, ˈpa-rə-, -ləl\ \ˈī-(ˌ)ō\

PARALLEL I/O

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Work completes in 1/5 the time 2 machines can do work of 10 5X lower overall solution cost All-inclusive simplicity

► Compute & storage services combined

22

DataCore Parallel I/O Breakthrough

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

23

Adaptive Parallel I/O

Workload

ResponseTime

(millisec)IOPS

No more load

400,000 IOPS< 1 millisec

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 24

100K

200K

300K

400K

$50K $100K $150K $200K

SPC-1 IOPS

PRICE

DellCompellentSC4020SSD / SAS

NetAppEF560 All Flash Array

EMCVNX 8000SSD / SAS

455KDataCore

500K

HP3PAR 7400SSD / SAS

The Major Storage BrandsPrice Performance Leaders

Objective Comparisons

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 25

Point of Reference

Workloads (Servers)

StorageSystem

SAN

OnlyPartialPrice

NotIncludedin Price

Traditional SPC-1

$177K435K IOPS

Industry-recognized I/O SPC-1 benchmark

Public results available for numerous products

Pricing included

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 26

No Match

All-inclusivePrice

I/O

Hyperconverged SPC-1

Workloads

VM VM VM VM

$36K

455K IOPS

Workloads (Servers)

I/O

SAN

OnlyPartialPrice

NotIncludedin Price

Traditional SPC-1

$177K

435K IOPS >

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Hyper-converged: It’s more than performance

27

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 28

High Availability: 3 vs 2 nodes

VMware VSAN Cluster* DataCore Virtual SAN

* Minimum requirement; Recommendation is 4 nodes

Lower Hardware Investment

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Room B

Building B

City B

29

High Availability: Stretch ClustersVMware VSAN Cluster* DataCore Virtual SAN

* Under development

Room A

Building A

City AUp to 100 KM

Better Availability & Resiliency

Virtualized Shared Storage

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Expand storage capacity by adding servers► Compute, memory, storage capacity & storage performance all tied together► Is your data growing at the same rate as your compute?► What if you don’t need more compute / memory?

Yet another Storage silo► Each cluster is separate from other clusters► Virtualized and Physical hosts are separate data islands► Doesn’t work with SAN/NAS

30

Issues Scaling Hyper-converged

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 31

Scale out Storage Capacity: Add Nodes (compute & storage)

VMware VSAN Cluster

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 32

Independently Scale out Storage Capacity: Leverage External SAN with Auto-Tiering

ExternalSAN

ColdWarmBurst Hot

Lower Hardware Costs; Capacity added as needed

VirtualSAN

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Integrated Platform across all Storage

DataCore Software-defined Storage Platform

HYPER-CONVERGED SAN CLOUD

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

DataCore Software-defined Storage Platform

For both Virtual Hosts & Physical Servers

VIRTUALIZED HOSTS PHYSICAL SERVERS

HYPER-CONVERGED SAN CLOUD

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

DataCore Software-defined Storage Platform

AVAILABILITY PERFORMANCE EFFICIENCY

MANAGEMENT

Services for Availability, Performance, Efficiency & Management

VIRTUALIZED HOSTS PHYSICAL SERVERS

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

DataCore Software-defined Storage Platform

Full Stack of Storage Services

VIRTUALIZED HOSTS PHYSICAL SERVERS

AVAILABILITY PERFORMANCE EFFICIENCYSynchronous Mirroring

Snapshots / BackupsCDPAsynchronous Replication

Caching

Random Write AcceleratorAuto-tiering

Quality of Service (QoS)

Thin ProvisioningStorage Pooling

Data MigrationDeduplication/Compression

NAS/SAN (Unified Storage) Centralized Management Analysis & ReportingCloud Integration

MANAGEMENT

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 37

A Seamless Storage Infrastructure

DataCore Software-defined Storage PlatformData Center

Application Clusters

Cloud

External SAN Remote Sites

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 38

Comparison of Options

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Key Criteria DataCore Virtual SANRAM for I/O Acceleration

2 nodes for High Availability

2 nodes for Stretch Cluster

Scale out storage capacity independent of compute

One management platform across storage infrastructure

One set of services across all storage devices

Support for Multi-hypervisor & Non-virtual environments

Hardware independent

39

DataCore TCO Summary

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Background Large restaurant chain with over 1,000 locations All key applications run locally

► Point of sale, order scheduling, etc Application downtime meant site closure (temporary)

► Loss of revenue and poor customer satisfaction

Requirement Lowest cost infrastructure for high availability

40

Case Study 1 – ROBO

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Lowest TCO Only 2 servers for HA per location RAM provides I/O acceleration so Flash not needed Runs natively in Windows, no extra license

Easy Management Multi-site automated deployment with simple wizards Integrates with Microsoft System Center Extensive instrumentation for centralized monitoring

41

Reasons for SelectingDataCore Virtual SAN

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Background Mid-sized Hospital Virtualizing PBX (voice communications), which was 12 physical servers

Requirements Reliable performance, as voice communication is a Tier 1 application Physical storage and compute footprint across 2 separate buildings

(geographically separated) for high availability

42

Case Study 2 – Application Cluster

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 43

VMware VSAN Requires 4 hosts

Only works at single site

Requires Flash on servers

DataCore Virtual SAN Only requires 2 hosts

Stretch cluster with only 2 nodes

Flash is optional (RAM is faster)

Comparing VMware VSAN and DataCore Virtual SAN

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 44

VMware VSAN Requires 4 hosts

Only works at single site

Requires Flash on servers

DataCore Virtual SAN Only requires 2 hosts

Stretch cluster with only 2 nodes

Flash is optional RAM is faster)

Comparing VMware VSAN and DataCore Virtual SAN

TCO of DataCore Virtual SAN was just 50% of VMware VSAN

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 45

75% 10x 4x 100% 90%reduction in storage

costs

performance increase

capacity utilization

reduction in storage-related

downtime

decrease in time spent on routine

storage tasks

Surveyed Customers Report Up To:

www.techvalidate.com

DataCore Benefits at a Glance

Copyright © 2014 DataCore Software Corp. – All Rights Reserved. 46

25,000+ Deployments Worldwide

10,000+ Customers 10th Gen Product

Companies in all Industries & Sizes

Market: Software-defined Storage

Technology: Storage Virtualization

Proven. Globally.

Copyright © 2014 DataCore Software Corp. – All Rights Reserved.

Thank You!

www.datacore.com