Documente Academic
Documente Profesional
Documente Cultură
2 3 4
HOW DOES FLASH CACHE IMPROVE PERFORMANCE? ................................................................ 4 HOW DOES FLASH CACHE IMPROVE DATA CENTER EFFICIENCY? ............................................ 5 TESTED CONFIGURATIONS ............................................................................................................... 5
4.1 4.2 4.3 BASELINE SUBMISSION: FAS3160, 224 FC 15,000 RPM DRIVES ...........................................................................6 SECOND SUBMISSION: FAS3160, 56 FC DRIVES WITH FLASH CACHE 256GB ...................................................7 THIRD SUBMISSION: FAS3160, 96 SATA DRIVES WITH FLASH CACHE 256GB ...................................................8
5 6
LINKS TO SPECSFS2008 RESULTS ................................................................................................... 9 WHY YOU CANT COMPARE SPECSFS2008 WITH SPECSFS97 ..................................................... 9
LIST OF TABLES Table 1) Flash Cache creates equivalent or better performance for 96-drive SATA and 56-drive Fibre Channel SPECsfs2008 submissions. ...........................................................................................................4
1
1.1
NetApp submitted three results for SPECsfs2008 over NFS, each of which was for the FAS3160 storage system. The first submission is based on a FAS3160 controller with 16 disk shelves of 15,000 RPM Fibre Channel drives, totaling 224 disk drives. In the next configuration we used the same drive types but added a Flash Cache (formerly named PAM II) to the solution, enabling the reduction of half the drives (to 56 disks). Also, a configuration with four disk shelves of 96 SATA disk drives was submitted. Each of the configurations showed very similar performance characteristics in the results, demonstrating the ability of Flash Cache (PAM II) to provide acceleration while reducing required drive counts. While these configurations demonstrate performance improvements with NFS and the SPECsfs2008 benchmark, the Flash Cache (PAM II) product is designed to accelerate workloads from all protocols: NFS, CIFS, FCP, and iSCSI.
1.2
SPECSFS2008 PERFORMANCE
The results are described in the SPECsfs2008 users guide as follows: SPECsfs2008 results summarize the servers capabilities with respect to the number of operations that can be handled per second, as well as the overall latency of the operations. A minimal SPECsfs2008 result will contain ten load points and the results of the server at each of the requested load points. Each load point presents a requested number of operations per second. So, when looking at throughput numbers, higher is better. In each of the submitted configurations the throughput was consistent, differing by less than 200 operations from the baseline figure of 60409 SPECsfs2008_nfs.v3 ops/sec. Also mentioned in the quote, latency is measured along with throughput at each load point. The load point latencies are then summarized in the results as the overall response time, or ORT. When looking at an ORT, lower is better; the requests are being responded to more quickly. In the results, the ORT of the baseline system was 2.18ms. The system with the Flash Cache card and one-quarter the Fibre Channel disk shelves achieved a better result at 1.58ms and the SATA-based submission was the same at 2.18ms. In addition to these figures, we also include reference points in the table below for storage capacity, electricity usage, and rack space.
SPECSFS2008 SUBMISSIONS AND RESULTS Table 1) Flash Cache creates equivalent or better performance for 96-drive SATA and 56-drive Fibre Channel SPECsfs2008 submissions.
Baseline Submission: No Flash Cache, 224 FC Disks FAS3160 controller; 16 x 300GB 15k RPM FC shelves (224 spindles)
Second Submission: Flash Cache and 56 FC Disks FAS3160 controller; 4 x 300GB 15k RPM FC shelves (56 spindles); 1 Flash Cache (PAM II) card per controller ~ Same (60,507)
Third Submission: Flash Cache and 96 SATA Disks FAS3160 controller; 4 x 1TB SATA shelves (96 spindles); 1 Flash Cache (PAM II) card per controller ~ Same (60,389)
Configuration
Throughput SPECsfs2008_nfs.v3 60,409 ops/sec Overall response time Raw capacity Electricity usage Rack space 2.18ms ~64TB Reference point Reference point
By decoupling performance requirements from capacity requirements, Flash Cache can allow you to implement only the amount of hardware required for either performance or capacity. The third example in the table above demonstrates this point. When reducing the latency of operations, Flash Cache enables a system with SATA disk drives to match the performance of a system with Fibre Channel drives in a SPECsfs2008 submission. This is accomplished while also reducing to one-quarter the number of drives in the test. The large capacity of SATA drives then brings the potential additional benefit of 50% more storage capacity, an approximate 66% reduction in electrical power requirements, and an estimated 59% less rack space for the solution. This is particularly significant because of the technology behind SATA. Running at 7,200 RPMs instead of the 15,000 RPMs of Fibre Channel enterprise drives, SATA draws less power and has a higher spatial density. Alone, this is a good combination, but the slower rotational speed means less performance. Flash Cache removes the performance barrier in the third submission, making SATA viable for both its storage capacity and reduced power.
TESTED CONFIGURATIONS
In this section, we look at the specifics of the tested configurations. In each of them, the system was based on the NetApp FAS3160; identical clients and networks were also utilized. The only variables were the reduction of drives from the baseline, the addition of Flash Cache to the solution, and the change to SATA drives in the third submission. The baseline submission is based on a FAS3160 controller and 224 Fibre Channel disk drives and does not include Flash Cache. The SPECsfs2008 results for this configuration are 60,409 SPECsfs2008_nfs.v3 ops/sec with an overall response time of 2.18ms.
4.1
Gigabit Ethernet
FAS3160
e0a e0b
0a 0b 0c 0d
FAS3160
e0a
e0b
0a
0b
0c
0d
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
FC
Power
FC
Power
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
System
Shelf ID
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
Figure 1) Each storage controller has eight disk shelves. The shelves are connected with both primary and secondary paths to the controllers. There are a total of 224 disks in the solution.
4 Gbit/s FC-AL
6
4 Gbit/s FC-AL
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
DS14
FC
Power
MK4
FC
Power
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
System
System
Shelf ID
Shelf ID
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
FC
Power
FC
Power
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
System
System
Shelf ID
Shelf ID
4.2
Gigabit Ethernet
Flash Cache 256GB
FAS3160
e0a e0b
0a 0b 0c 0d
FAS3160
e0a
e0b
0a
0b
0c
0d
Figure 2) Each storage controller has two disk shelves. The shelves are connected with both primary and secondary paths to the controllers. There are a total of 56 disks in the solution.
The second submission used the same FAS3160 controller. It included one-quarter the number of drives as the baseline at 56 disks and included two Flash Cache 256GB, one per controller. The SPECsfs2008 results for throughput slightly improved from 60,409 to 60,507 SPECsfs2008_nfs.v3 ops/sec, while the overall response time improved by 37.6% from the baseline of 2.18ms to 1.58ms. As can be seen in comparing the two diagrams, the reduction in drives resulted in a direct reduction in disk shelves, saving approximately up to 67% on electricity and reducing rack space up to 67%.
4 Gbit/s FC-AL
4 Gbit/s FC-AL
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
DS14
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
MK4
FC
Power
FC
Power
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance NetworkAppliance
Fault
Loop A
Loop B
System
Shelf ID
DS14
FC
Power
MK4
DS14
FC
Power
MK4
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
System
14 300GB FC Drives
72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F 72F
Fault
Loop A
Loop B
System
Shelf ID
Shelf ID
4.3
Gigabit Ethernet
Flash Cache 256GB
FAS3160
e0a e0b
0a 0b 0c 0d
Figure 3) The configuration uses four disk shelves. The shelves are connected with both primary and secondary paths to the controllers. There are a total of 96 disks in the solution.
The third submission used the same FAS3160 controller. The disk drives in the solution changed from Fibre Channel to SATA and less than half the number of drives as the baseline at 96 disks. It included
3 Gbit/s SAS
450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB 450GB
FAS3160
e0a
3 Gbit/s SAS
e0b
0a
0b
0c
0d
450GB
450GB
450GB
DS4243
450GB
450GB
450GB
450GB
DS4243
450GB
450GB
450GB
450GB
DS4243
450GB
450GB
450GB
450GB
450GB
450GB
450GB
450GB
DS4243
450GB
450GB
450GB
450GB
450GB
450GB
450GB
450GB
DS4243
450GB
450GB
450GB
450GB
450GB
two Flash Cache 256GB modules, one per controller. The SPECsfs2008 results for throughput were consistent at 60,389 SPECsfs2008_nfs.v3 ops/sec, while the overall response time stayed the same at 2.18ms. As can be seen in comparing the diagrams, the reduction in drives resulted in a direct reduction in disk shelves, saving up to 66% on electricity and reducing rack space by up to 59%. The use of the larger capacity SATA drives also increased storage capacity by up to 50% over the baseline.
The results and other information about SPECsfs2008 can be found at www.spec.org. For each specific result: Baseline submission: FAS3160 with 224 Fibre Channel drives www.spec.org/sfs2008/results/res2009q3/sfs2008-20090727-00127.html
Second submission: FAS3160 with 56 Fibre Channel drives plus Flash Cache www.spec.org/sfs2008/results/res2009q3/sfs2008-20090727-00126.html Third submission: FAS3160 with 96 SATA drives plus Flash Cache www.spec.org/sfs2008/results/res2009q3/sfs2008-20090727-00128.html
SPECsfs2008 includes a number of changes to the benchmark that make it unrealistic to compare or extrapolate performance from the older versions. Its documentation includes the following guidance: The SPECsfs2008 release of the benchmark includes major workload and functionality changes, as well as clarification of run rules. The code changes compared to earlier SFS versions were NOT performance neutral, therefore comparing SPECsfs2008 results with SFS 3.0 results is NOT allowed. In accordance with this guidance, NetApp recommends only comparing SPECsfs2008 to other SPECsfs2008 submissions. Any other comparisons will not prove useful.
NetApp provides no representations or warranties regarding the accuracy, reliability or serviceability of any information or recommendations provided in this publication, or with respect to any results that may be obtained by the use of the information or observance of any recommendations provided herein. The information in this document is distributed AS IS, and the use of this information or the implementation of any recommendations or techniques herein is a customers responsibility and depends on the customers ability to evaluate and integrate them into the customers operational environment. This document and the information contained herein may be used solely in connection with the NetApp products discussed in this document.
Copyright 2011 NetApp, Inc. All rights reserved. No portions of this document may be reproduced without prior written consent of NetApp, Inc. Specifications are subject to change without notice. NetApp, the NetApp logo, Go further, faster, xxx, and xxx are trademarks or registered trademarks of NetApp, Inc. in the United States and/or other countries. <<Insert third-party trademark notices here.>> All other brands or products are trademarks or registered trademarks of their respective holders and should be treated as such.
10