Optimizing NetApp SnapMirror Data Replication with F5 BIG-IP WAN Optimization Manager

Transcription

1 PARTNER USE CASE Optimizing NetApp SnapMirror Data Replication with F5 BIG-IP WAN Optimization Manager F5 BIG-IP WAN Optimization Manager (WOM) helps organizations optimize data replication with NetApp SnapMirror on clustered Data ONTAP, improving replication performance and increasing bandwidth efficiency. Matt Quill, F5 Networks Renny Shen, F5 Networks Sujith John, NetApp

2 Contents Introduction 3 Benefits of BIG-IP WOM with SnapMirror 3 Accelerating Data Replication 3 Reducing Bandwidth Costs 4 Security Benefits 5 Review of Environment Setup and Deployment 5 Oracle Setup 6 NetApp Configuration 6 BIG-IP WOM Setup 7 Test Parameters and Results 7 Performance Tests 7 Bandwidth Optimization Test 11 Conclusion 12 2

3 Introduction Businesses today operate in an uncertain climate where a number of factors can disrupt normal operations. Whether responding to technology failure, infrastructure degradation, or a natural disaster, IT organizations must ensure that their businesses can continue to run even with the loss of the supporting IT systems or infrastructure. Essential to this responsibility are the abilities to fail over business applications and back up data to a secondary site. This requires having not only the necessary application infrastructure on standby at the secondary site, but also access to the most recent set of business and application data. NetApp SnapMirror provides a solution to that latter requirement, by replicating business and application data from a NetApp storage system at a primary site to one at a secondary site. But because data replication typically occurs over a wide area network (WAN), replication performance may be affected by unpredictable latency. In addition, many organizations are concerned about the cost of bandwidth consumed and the security of transmitting business-sensitive data over the WAN. F5 BIG-IP WAN Optimization Manager (WOM) complements SnapMirror by optimizing data replication over the WAN. BIG-IP WOM employs adaptive compression, deduplication, TCP optimization, and encryption to help organizations accelerate and secure data replication, while reducing the amount of bandwidth consumed. To better quantify these benefits, F5 worked with NetApp to deploy a joint SnapMirror and BIG-IP WOM solution, using an Oracle application environment to simulate a real-world application environment. Benefits of BIG-IP WOM with SnapMirror Growing organizations generate increasing amounts of data that must be replicated. As the total amount of data grows, so does the number of users and applications, which further increases the rate at which new data is generated. This growth in data volume and rate of generation can have a serious impact on both the ability to replicate data within available windows for disaster recovery (DR) purposes, as well as the financial cost of doing so. Accelerating Data Replication Replicating data over the WAN poses inherent performance challenges. Replication performance can be affected by a number of factors, including the distance between sites, network latency, and packet loss. Within an organization s control, there can be 3

4 congested network conditions as different applications compete for limited bandwidth. These factors all contribute to the growing amount of time required to replicate data. BIG-IP WOM helps organizations better manage dynamic network conditions to improve replication performance. BIG-IP WOM provides symmetric adaptive compression and data deduplication to dramatically reduce the amount of data that needs to be transmitted over the WAN. BIG-IP WOM can perform deduplication on very large data sets using solid state drives, for up to a 3x improvement in replication time over comparable memory-based deduplication. Beyond compression and deduplication, BIG-IP WOM also provides several protocol optimizations, including adaptive TCP optimization, to respond to changing network conditions. This minimizes throughput degradation over long-distance and high-latency links, helping to further accelerate replication traffic. In addition, BIG-IP WOM uses L7 QoS rate shaping to prioritize critical or time-sensitive applications. Organizations can use rate shaping to help enable quality of service (QoS) for SnapMirror data replication over the WAN. Reducing Bandwidth Costs Organizations are often concerned about the cost of bandwidth used for data replication purposes, especially with a growing amount of data that needs to be replicated. For organizations nearing the limit of their existing WAN link, upgrading to the next level can be costly. For example, upgrading from an OC-3 connection to an OC-12 typically costs another $1 million a year. Upgrading from OC-12 to OC-48 can cost another $2.4 million a year. Because of this, organizations are often better served by delaying a WAN link upgrade and instead finding ways to use their existing bandwidth more efficiently. BIG-IP WOM can help organizations increase bandwidth efficiency and get more out of their existing WAN link. BIG-IP WOM provides symmetric adaptive compression and data deduplication to reduce the amount of data being transmitted by up to 50 percent. This ensures that organizations can make better use of their existing bandwidth, while postponing any WAN link upgrade until truly needed. 4

5 Cost of bandwidth (annual) $3.6M $1.2M $40K $120K T3 (45 Mbps) OC-3 (155 Mbps) OC-12 (622 Mbps) OC-48 (2.5 Gbps) Figure 1: Increase in annual bandwidth cost when upgrading WAN links. Security Benefits While not specifically included in the joint testing described here, BIG-IP WOM provides site-to-site data security features. BIG-IP WOM transmits all data over the WAN through encrypted isession network tunnels, ensuring the security of data replicated through SnapMirror. In addition, hardware-accelerated SSL processing is available on all BIG-IP platforms, offloading bulk encryption and decryption from the NetApp storage system. The ability to consolidate replication performance and security features into an integrated WAN optimization platform can reduce upfront and management costs for organizations data replication needs. Review of Environment Setup and Deployment To demonstrate the benefit of optimizing SnapMirror data replication with WOM, F5 and NetApp deployed a test environment at NetApp s Research Triangle Park campus in North Carolina. The test environment reproduced a typical enterprise application infrastructure and consisted of two virtual data centers: a primary data center and a failover site. The two data centers were connected by a simulated WAN link that was configurable with various settings for bandwidth, latency, and packet loss. In the primary data center, an Oracle Database 11g instance generated real-world application data that was stored on NetApp storage systems configured with clustered 5

6 Data ONTAP SnapMirror replicated Oracle data to the secondary site, over the WAN link optimized by BIG-IP WOM. Oracle Setup An Oracle Database 11g instance provided the application data for all the joint replication tests. The database server was run as a virtual machine on VMware ESX Server. For testing purposes, F5 and NetApp provisioned four data volumes for the Oracle database 7 GB, 13 GB, 30 GB, and 1 TB in size and generated Oracle data using Quest Software s DataFactory software tool. Oracle Database 11g Optimized WAN Link Oracle Database 11g NetApp SnapMirror BIG-IP WAN Optimization Manager BIG-IP WAN Optimization Manager NetApp SnapMirror NetApp FAS NetApp FAS Primary Data Center Secondary Data Center Figure 2: Test configuration of BIG-IP WOM with NetApp SnapMirror. NetApp Configuration To ensure high performance, a pair of NetApp FAS 3240 controllers configured with clustered Data ONTAP and 450 GB 10K SAS disk technology provided storage for the test environment. FAS 3240 Software version Clustered Data ONTAP Disk shelf 24 x 450 GB 10K SAS Figure 3: Configuring NetApp storage. 6

7 BIG-IP WOM Setup To support this validation, F5 provided two BIG-IP appliances running BIG-IP WOM and configured with an SSD option. Device model BIG-IP Software version TMOS 11.2 Hard drive 4x 300 GB SSD Figure 4: Configuring BIG-IP WOM. Test Parameters and Results Performance Tests BIG-IP WOM can dramatically reduce the amount of time required to replicate data with SnapMirror. To help quantify this benefit, F5 and NetApp performed a series of joint tests that measured replication times under three scenarios: Replication over metro distances Replication over regional distances Replication over long distances In addition, the tests examined the benefit of BIG-IP WOM over four different WAN links: T3 (45 Mbps), OC-3 (155 Mbps), OC-12 (622 Mbps), and 1 Gbps. Replication over metro distances In this test, F5 and NetApp reproduced conditions typically experienced by enterprises replicating data over metro distances. The simulated WAN link was configured with the following characteristics: Network bandwidth Network latency Network integrity T3 (45 Mbps) Low (5 ms) Lossless (0% packet loss) Figure 5: WAN characteristics for simulating replication over metro distances. 7

8 This test measured the amount of time required to replicate a 7 GB Oracle database and involved the following steps for both the unoptimized and WOM-optimized test cases: 1 Create a 7 GB Oracle data volume at the primary site using the DataFactory tool. 2 In the unoptimized test case, replicate the entire Oracle data volume to the secondary site without BIG-IP WOM. 3 In the optimized test case: a Replicate the entire Oracle data volume to the secondary site with BIG-IP WOM to measure the initial data deduplication impact. b Perform a second replication pass with BIG-IP WOM to measure the additive effects of data deduplication. Replication time over metro distances Minutes WOM Optimized (Pass 2) WOM Optimized (Pass 1) Unoptimized Figure 6: BIG-IP WOM reduced replication time more than 9x when tested over metro distances. In this scenario, the test results demonstrated an improvement in performance of more than 2x in the first pass and more than 9x in subsequent passes. Without BIG-IP WOM, it took SnapMirror 24 minutes, 21 seconds to replicate 7 GB of data. With BIG-IP WOM, symmetric compression, data deduplication, and TCP optimization reduced the amount of time required to 11 minutes, 30 seconds. The second replication pass showed further improvement, reducing the amount of time required to 2 minutes, 39 seconds. 8

9 Replication over regional distances In this test, F5 and NetApp reproduced conditions typically experienced by enterprises replicating data over regional distances. The simulated WAN link was configured with the following characteristics: Network bandwidth Network latency Network integrity 1 Gbps Medium (20 ms) Lossless (0% packet loss) Figure 7: WAN characteristics for simulating replication over regional distances. This scenario measured the amount of time required to replicate a 30 GB Oracle database and involved the following steps for both the unoptimized and WOMoptimized test cases: 1 Create a 30 GB Oracle data volume at the primary site using the DataFactory tool. 2 In the unoptimized test case, replicate the entire Oracle data volume to the secondary site without BIG-IP WOM. 3 In the optimized test case: a Replicate the entire Oracle data volume to the secondary site with BIG-IP WOM to measure the initial data deduplication impact. b Perform a second replication pass with BIG-IP WOM to measure the additive effects of data deduplication. In this use case, the test results demonstrated an improvement in performance of almost 3x in the first pass and 4x in subsequent passes. Without BIG-IP WOM, it took SnapMirror 30 minutes, 5 seconds to replicate 30 GB of data. With BIG-IP WOM, symmetric compression, data deduplication, and TCP optimization reduced the amount of time required to 13 minutes. The second replication pass showed further improvement, reducing the amount of time required to 7 minutes, 36 seconds. 9

10 Replication time over regional distances Minutes WOM Optimized (Pass 2) WOM Optimized (Pass 1) Unoptimized Figure 8: BIG-IP WOM reduced replication time more than 4x when tested over regional distances. Replication over long distances In this test, F5 and NetApp reproduced conditions typically experienced by enterprises replicating data over long distances. The simulated WAN link was configured with the following characteristics: Network bandwidth Network latency Network integrity OC-3 (155 Mbps) High (80 ms) Lossy (0.1% packet loss) Figure 9: WAN characteristics for simulating replication over long distances. This scenario measured the amount of time required to replicate a 7 GB Oracle database and involved the following steps for both the unoptimized and WOMoptimized test cases: 1 Create a 7 GB Oracle data volume at the primary site using the DataFactory tool. 2 In the unoptimized test case, replicate the entire Oracle data volume to the secondary site without BIG-IP WOM. 3 In the optimized test case: a Replicate the entire Oracle data volume to the secondary site with BIG-IP WOM to measure the initial data deduplication impact. 10

11 b Perform a second replication pass with BIG-IP WOM to measure the additive effects of data deduplication. In this use case, the test results demonstrated an improvement in performance of more than 12x in the first pass and 18x in subsequent passes. Without BIG-IP WOM, it took SnapMirror 63 minutes, 26 seconds to replicate 7 GB of data. With BIG-IP WOM, symmetric compression, data deduplication, and TCP optimization reduced the amount of time required to 5 minutes, 24 seconds. The second replication pass showed further improvement, reducing the amount of time required to 3 minutes, 34 seconds. Replication time over long distances Minutes WOM Optimized (Pass 2) WOM Optimized (Pass 1) Unoptimized Figure 10: BIG-IP WOM reduced replication time more than 18x when tested over long distances. Bandwidth Optimization Test In addition to increasing performance, BIG-IP WOM also enables organizations to use available bandwidth more efficiently. F5 and NetApp performed a final test to quantify this benefit. The test configuration reproduced a typical configuration used for data protection over metro distances, using a simulated WAN link with the following characteristics: Network bandwidth Network latency Network integrity OC-12 (622 Mbps) Low (5 ms) Lossless (0% packet loss) FIgure 11: WAN characteristics used during the bandwidth optimization test. 11

12 This test measured the amount of data transmitted in replicating changes made to the 1 TB Oracle database. The test involved the following steps for both the unoptimized and WOM-optimized test cases: 1 Create a 1 TB Oracle data volume at the primary site using the DataFactory tool. 2 Perform an initial SnapMirror replication synchronization of the Oracle database with the secondary site. 3 Generate an additional 59 GB of data to the Oracle database at the primary site. 4 In the unoptimized case, replicate the additional changes to the secondary site without BIG-IP WOM. 5 In the WOM-optimized case, replicate the additional changes to the secondary site with BIG-IP WOM to measure the data deduplication impact. The test results demonstrated a significant reduction in the amount of data transmitted. When optimized by BIG-IP WOM, symmetric compression and data deduplication reduced the amount of data transmitted from 59 GB to 18.6 GB. Amount of data transmitted Minutes WOM Optimized Unoptimized Figure 12: BIG-IP WOM improved replication efficiency by more than 3x. Conclusion Optimizing SnapMirror data replication can enhance an organization s ability to replicate data for DR purposes. BIG-IP WOM offers capabilities that help decrease replication times under a variety of network conditions. In addition, BIG-IP WOM helps organizations make better use of their existing bandwidth and delay costly WAN link 12

13 upgrades. Consolidating increased performance, reduced costs, and secure data replication into a single appliance offers a simple and cost-effective solution for optimizing SnapMirror data replication. For more information about unified application and data delivery solutions from F5 and NetApp, visit f5.com/products/technology/netapp or contact NetApp@f5.com. F5 Networks, Inc. 401 Elliott Avenue West, Seattle, WA F5 Networks, Inc. Corporate Headquarters info@f5.com F5 Networks Asia-Pacific apacinfo@f5.com F5 Networks Ltd. Europe/Middle-East/Africa emeainfo@f5.com F5 Networks Japan K.K. f5j-info@f5.com 2013 F5 Networks, Inc. All rights reserved. F5, F5 Networks, and the F5 logo are trademarks of F5 Networks, Inc. in the U.S. and in certain other countries. Other F5 trademarks are identified at f5.com. Any other products, services, or company names referenced herein may be trademarks of their respective owners with no endorsement or affiliation, express or implied, claimed by F5. CS