全面分析Netapp的測試指標_第1頁
全面分析Netapp的測試指標_第2頁
全面分析Netapp的測試指標_第3頁
全面分析Netapp的測試指標_第4頁
全文預(yù)覽已結(jié)束

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進行舉報或認領(lǐng)

文檔簡介

1、PowerlinkJanuary 2012No-Compromise Scale-out Storage:Analysis of NetApp 8.1 C-Mode BenchmarksTable of Contents· Background· Claim 1: NetApp has 35% more performance than EMC Isilon· Claim 2: NetApp holds the SPECsfs world record · Claim 3: NetApp has linear scalability · Cla

2、im 4: NetApp has excellent storage efficiency · ConclusionNote: This document arms you with responses to the recent SPECsfs 2008 NFS benchmark results posted by NetApp, which at first glance appear to be a compelling world record that surpasses EMC Isilon. You will be able to clearly identify t

3、he weaknesses, apparent upon closer inspection, in their test and articulate these to your customers. BackgroundOn November 2, NetApp posted new SPECsfs 2008 NFS benchmark results using 8, 12, 16, 20, and 24 node HA-paired FAS6240 filers. And, while the numbers may look compelling, ONTAP 8.1 Cluster

4、-Mode has the same constraints and operational inefficiencies as scale-up 7-mode, failing to address the business and operational challenges organizations seek to solve with scale-out storage.EMC Isilon no-compromise, scale-out storage continues to set the standards for operational efficiency, data

5、protection, storage efficiency and linear scalability of performance, and capacity. There are substantial operational differences between the 24 individual discrete 20 TB file systems employed by NetApp to achieve their benchmarks and the single scalable volume and file system used by Isilon. The EM

6、C Isilon system used in this same benchmark is a no-compromise solution that includes several unique benefits:· Automated balancing of performance and capacity, eliminating data migrations· SmartPools tiering to automatically align the performance requirements and value of data to the most

7、 effective media· FlexProtect data protection that goes beyond RAID to automatically align data protection levels to the value of the data· Enterprise infrastructure applications that benefit the entire clusters file system · 200% better storage utilization (when comparing the benchma

8、rk configurations used by NetApp to those used by Isilon)· N-way, scalable resiliency that improves as the cluster grows Questions raised by NetApps claims based on their SPECsfs 2008 NFS benchmark results are posed below.Claim 1: NetApp has 35% more performance than EMC IsilonAnalysis: NetApp

9、demonstrated impressive performance from 24 filers, 24 discrete 20 TB WAFL file systems, 48 aggregates (24 data, 24 ONTAP OS) and 48 volumes (24 data, 24 ONTAP OS). They claim to provide 35% more performance than Isilon. Isilon 1.1M SPECsfs NFS and 1.6M SPECsfs CIFS benchmarks, on the other hand, we

10、re accomplished with the simplicity of a single file system and single volume. The manageability implications of this are significant.When looking at the NetApp configuration more closely, one question that comes to mind is how performance and capacity are balanced across 24 volumes and aggregates l

11、ocated on 24 different filers. As in ONTAP 7-mode, a file can only reside within a single volume, locked within a single aggregate, locked on a single storage controller. Administrators are still required to manage file placement within the controllers aggregates, and monitor the space and performan

12、ce utilization of each of these 24 discrete volumes. In a real-world environment, there will certainly be more aggregates and many more volumes, amplifying the management tasks.Managing 24 different and independent snapshot and replication policies across 24 volumes on 24 different filers also poses

13、 some unique challenges, such as how to control user quotas on 24 volumes across 24 different filers. As in ONTAP 7-mode, data protection and management infrastructure apps are hard-wired to the filer/volume construct and must be managed and applied to each individual volume on each filer independen

14、tly. C-mode simply clusters scale-up filers and along with them, all of their constraints and operational inefficiencies.Claim 2: NetApp holds the SPECsfs world record Analysis: NetApp claims a world record in SPECsfs performance: 1.5M Ops from 24 FAS6240 filers. Looking more closely, it can be seen

15、 that you will incur a 34% performance penalty if you choose to run FAS6240s in C-Mode. In 7-mode, a 2-node HA pair of FAS6240s provided 190,675 SPECsfs NFS Ops. In C-Mode, 24 2-node HA pairs of FAS6240s delivered an aggregate of 1.5M SPECsfs NFS Ops or 126,065 SPECsfs NFS Ops per node. It is diffic

16、ult to see the benefits that make this 34% performance tax worth it. The C-Mode performance penalty is evident even in an ideal benchmark lab situationa perfect environment with homogenous nodes and disks without SnapMirrors, SnapVaults, volume movement, block reclamation scanning, dedupe, compressi

17、on or drive rebuilds executing, all of which would further degrade cluster performance. The practical, real-world performance tax for C-Mode will most likely be much greater in a production environment.Claim 3: NetApp has linear scalability Analysis: In order not to impact benchmark performance, Net

18、App identified the capacity/performance sweet spots for six different numbers of filers in a cluster and artificially limited. For example, the 24-node cluster had a useable, exported capacity of only 288 TBs. Although a FAS6240 can support 4.3 PBs of capacity, each FAS6240 in the benchmark configur

19、ation was limited to a single, 20 TB raw aggregate. With such a low, unrealistic capacity/performance sweet spot, and a 34% filer performance penalty in C-Mode, it is a challenge to justify using NetApp for scale-out, as noted on the NetApp blog: “This is the fastest 24 6240 nodes go with this bench

20、mark. Increasing the number of spindles wouldnt change it, either. It would stay similar just with more space.”Although a FAS6240 in C-Mode now supports 100 TB aggregates, the use of aggregates greater than 20 TBs will have a negative impact on the cluster performance. For this benchmark, NetApp C-M

21、ode could not demonstrate scalable performance beyond 288 TBs of exported, useable capacity. Once that capacity threshold is reached, performance flattens out. And, it can be assumed that as capacity is scaled up, performance will then begin to decrease as the WAFL software RAID performance overhead

22、 increases proportionally with the size of the aggregate.Claim 4: NetApp has excellent storage efficiency Analysis: The total capacity of the NetApp cluster used in the benchmark was 574 TBs. The amount exported for use was 288 TBs from 24 file systems and 24 volumes50% of its useable capacity. The

23、total capacity of the EMC Isilon 140-node SPECsfs benchmark configuration was 864 TBs. The amount exported for use was 864 TBs from its single file system and single volume100% of its usable capacity. The NetApp benchmark configuration had a total raw capacity of 777 TB. 26% was lost to RAID and hot

24、 spare overhead, leaving a useable capacity of 574 TBs, out of which only half (288 TBs) was exported for use. For every dollar spent per TB on this configuration, you would receive only 37 cents worth of capacity, effectively wasting 63 cents per TB. With only half of the NetApp clusters capacity a

25、vailable for use driving up the cost per useable TB, is it worth getting 34% less performance and 50% less capacity utilization for the capital cost of each filer when using them in C-Mode?ConclusionPlaying with Words Another note of interest can be seen on the web site, where the number

26、 of file systems for NetApp is stated as “Single Namespace.” NetApp had initially intended to claim a single file system for the C-Mode benchmarks. However, the committee would not allow that claim. “Single Namespace” was the compromised verbiage for the submission, which was actually compr

27、ised of 24 file systems, 48 aggregates, and 48 volumes.At best, Cluster-Mode is an example of clustered scale-up, and cannot be considered true scale-out. EMC Isilon provides organizations with a no-compromise, scale-out solution that offers five uniquebut relatedbenefits:· Ease of use and ease of management from multiple nodes presenting a single, scalable file system and volume · New levels of data protection and high availability provided by N-w

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負責。
  • 6. 下載文件中如有侵權(quán)或不適當內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論