competitive analysis hydrastor vs. emc data domain multi-node configuration december, 2013 nec...

8
Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version 4 software and generation 4 hardware, and information on DataDomain obtained at the time of writing

Upload: asher-smith

Post on 26-Dec-2015

240 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

Competitive AnalysisHYDRAstor vs. EMC Data Domain

Multi-Node Configuration

December, 2013

NEC Confidential

This document is based on HYDRAstor version 4 software and generation 4 hardware, and information on DataDomain obtained at the time of writing

Page 2: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential

HYDRAstor/Data Domain Market Coverage

Scale-Out Model Competitors

Data Domain

Raw Capacity

DD4200

DD4500

DD7200

DD990

DD160 1.1 TB/h(1)

0.67 TB/h

DD620 2.4 TB/h(1)

1.1 TB/h

DD2500

SMB

Med

ium

-Size

d

Enterp

rises

Large/

Very

Large

Enterp

rises

-12

TB

500

TB +

20 -

360

TB

22 TB/h(1)

10.2 TB/h

22 TB/h(1)

10.2 TB/h

26 TB/h(1)

11.9 TB/h

31 TB/h(1)

15 TB/h

13.4 TB/h(1)

5.3 TB/h

720 TB

540 TB

360 TB

240 TB

171 TB

12 TB

6 TB

(Max)

(Max)

(Max)

(Max)

(Max)

(Max)

(Max)

1HN

Single NodeModel

HYDRAstorScale-Out Model

6.48 TB/h

3.96 TB/h

12-48 TB

8-24 TB

7.9 PB (Max) (Max)

1069 TB/h

2

(1) Max. performance using DD Boost

Page 3: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential3

Per

form

ance

(T

B/h

)

1000

6 8

500

Capacity/Performance Comparison

HYDRAstor

4Raw Capacity (PB)

Enlarged on next slide

2

Data Domain

DD990DD7200

DD4200/4500DD2500

Page 4: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential

4

Capacity/Performance Comparison (Enlargement)

Raw Capacity (PB)1

10

Per

form

ance

(T

B/h

)

30

20

DD Extended Retention Software option

0.5

DD990 (with DD Boost)

DD Extended Retention Software option DD990

DD Extended Retention Software option DD4200 / 4500

1.5

Pe

rfo

rma

nc

e

Inc

rea

se

-720 TB -2.53 PB

-720 TB -2.53 PB

-540 TB

-723 TB (DD4200)-240 / -360 TB

DD Extended Retention Software option

-2.14 PB

DD7200

HYDRAstor

DD Extended Retention Software option DD7200 (with DD Boost)

-540 TB -2.14 PB

Pe

rfo

rma

nc

e

Inc

rea

se

-1.39 PB (DD4500)

DD2500

-171 TB

Page 5: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential5

Category ItemHYDRAstor

Scale-Out Model

Data DomainDD2500

Data DomainDD4200

Data DomainDD4500

Data DomainDD7200

Data DomainDD990

Scalability

Raw Capacity 12 TB – 7920 TB[1] >21 TB

- 171 TB30 TB

- 240 TB(~ 723 TB[3])

60 TB- 360 TB

(~1.39 PB[3])

90 TB- 540 TB

(~2.14 PB[3])

16 TB- 720 TB

(~2.53 PB[3])

Maximum Throughput

6.48 TB/h – 1069 TB/h[2]>> 5.3 TB/h 10.2 TB/h 10.2 TB/h 11.9 TB/h 15 TB/h

>> 13.4 TB/h[4] 22 TB/h[4] 22 TB/h[4] 26 TB/h[4] 31 TB/h[4]

Basic Functions

Global Dedup Yes >> No

Supported Protocols

CIFS, NFS, OST < CIFS, NFS, NDMP, OST, VTL

OST[5]

Optimized Duplication, Optimized Synthetics, AIR* (*planned for v4.2)

=Granular Recovery, Optimized Duplication,

Optimized Synthetics

Connection Interfaces

1GbE, 10GbE < 1GbE, 10GbE, 8GFC

Replication Yes (50 remote sites) <Yes (180 remote sites)

Yes (270 remote sites)

Application MarkerFiltering

NetBackup, Networker, DataProtector, TSM, Commvault, RMAN

(Configurable per filesystem)

=NetBackup, Networker, DataProtector,

TSM, Commvault, RMAN(NBU is automatically detected. Configurable per system)

[1] 165 node configuration, [2] 165HN configuration, [3] With DD Extended Retention, [4] Using DD Boost[5] HYDRAstor’s unique Express I/O OST Plug-in does not directly correspond to any specific NBU function, but does leverage multiple capabilities of the OpenStorage API

HYDRAstor/Data Domain Comparison 1/2

Page 6: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential6

Category ItemHYDRAstor

Scale-Out Model

Data Domain DD2500, DD4200, DD4500, DD7200, DD990

Archival

WORM Yes = Yes

Encryption(disk)

Yes (Configurable per filesystem) > Yes (Configurable per system)

Encryption (replication)

Yes (Configurable per node) = Yes (Configurable per replication pair)

Data ShreddingYes (Readable/Writeable while in

progress)>

Yes (Read-only while in progress)

Upgrade/Migration

Data Migration to Next-Gen HW

Automatic >> Manual

Reliability

Maximum Controllers

1 - 165 >> 1

ClusteringActive-Active (Readable/writeable on

controller failure)>> No

HDD Failure 6 concurrent HDD failure tolerance >> 2 concurrent HDD failure tolerance (RAID6)

HDD Rebuild Time

5 minutes – 20 hours >> Several hours – Several days

Shelf/SN FailureReadable/Writeable up to 6 SN

failures> Not readable on shelf failure

HYDRAstor/Data Domain Comparison(continued)

2/2

Page 7: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential7

HYDRAstor – Strengths HYDRA’s Maximum Performance and Capacity completely overwhelm the

competition from Data Domain Massive Scalability in both Performance and Capacity

Can scale from the minimum all the way to the maximum configuration by adding nodes. Functions as a single storage pool, even after the addition of nodes. After adding nodes, automatically redistributes data evenly over the entire system,

balancing capacity and load. HYDRAstor can scale freely without any impact on deduplication ratio. → Data Domain cannot scale in performance. When a model reaches its capacity limit, you

must replace it or purchase an additional unit. Operating multiple units in parallel lowers the overall deduplication ratio and complicates management.

High Availability Redundant configurations with no SPOF are possible (in systems equal or greater to

2HN). Fast HDD rebuild times after a failure.→ With RAID6, one 4 TB HDD failure takes from several hours to several days to rebuild.

HYDRAstor (DRD) takes from as little as five minutes to up to twenty hours. Up to 6 HDD/node failure resiliency with less capacity overhead than traditional RAID.

Automatic Migration HW can be upgraded by sequentially replacing old nodes with new nodes

High Physical Compression Ratio HYDRA uses a high-compression engine and achieved a compression ratio 10% higher

than Data Domain when benchmarked with user data (Oracle DB).

Page 8: Competitive Analysis HYDRAstor vs. EMC Data Domain Multi-Node Configuration December, 2013 NEC Confidential This document is based on HYDRAstor version

© NEC Corporation 2013 NEC Confidential8

Selling Against Data Domain

As DD Boost simply performs deduplication at the backup server, using it does not mean that DD performance will significantly increase.

The value for logical (effective) capacity used in the data sheet assumes a 50:1 data reduction ratio (HYDRA assumes a 20:1 data reduction ratio)Additionally, while DD only handles two concurrent failures with RAID6, HYDRAstor’s capacity numbers are based on three concurrent failure resiliency.

GDA has been removed from the DD product lineup likely due to a lack of technical capability. It appears that functionally GDA was an incomplete clustering product that was

abandoned after not being able to strengthen the features.

DD Extended Retention does not increase standard backup/restore capacity; it adds long-retention storage capacity, which cannot be directly restored. This introduces a performance penalty for restore operations.