Riverbed WAN Acceleration for EMC Isilon Sync IQ Replication

EMC Isilon Scale-Out NAS storage solutions enable the consolidation of disparate pools of storage into a single highly reliable and scalable storage system. To guarantee the availability of mission critical data, EMC Isilon also provides a host of backup and recovery solutions designed to meet the organization’s business continuance requirements. For disaster recovery purposes, EMC Isilon provides asynchronous replication using SyncIQ software application,. The EMC Isilon OneFS® operating system provides the intelligence behind all Isilon scale-out storage systems.

Riverbed Steelhead complements EMC Isilon’s OneFS operating system providing leading wide area network (WAN) and application acceleration (WAAS) technologies that, deployed along with EMC Isilon Scale-Out NAS storage systems, greatly accelerates data transfers over wide area networks to further consolidate and protect data. Riverbed Steelhead complements EMC Isilon’s SyncIQ replication by accelerating data transfers over wide area networks. This helps SyncIQ deliver high performance, asynchronous replication of data over WAN (Wide Area Networks) to address a broad range of recovery point objectives (RPO) and recovery time objectives (RTO). In addition to disaster recovery, SyncIQ can also facilitate the distribution of content between regional sites as organizations expand geographically. 

Test Setup and Methodology
Riverbed appliance integrates seamlessly with EMC Isilon clusters by connecting to the same local area network the EMC Isilon cluster network is connected to. The data transferred between the two EMC Isilon clusters is accelerated through the Steelhead appliances connected to the WAN.

Riverbed provides TCP streamlining and data streamlining optimization for SyncIQ data replication. In addition to standard WAN acceleration, EMC Isilon data replication jobs are further accelerated when redundant data is sent across.

The test lab network is comprised of two data centers connected over a simulated WAN. Each data center has an EMC Isilon System with SyncIQ policies configured to sync data from cluster 1 to cluster 2. The SyncIQ job is started manually and the replication data goes through the two Steelhead 7050 appliances. 

Network Topology
The figure below shows the network topology used for disaster recovery sync replication testing. A windows client is used to populate test data to EMC Isilon FS Cluster 1 in Data Center 1. There are two Steelhead 7050 appliances running RIOS 6.5.2 installed in the two data centers. The EMC Isilon system is running OneFS 6.5.3. A WAN simulator is connected between the two data centers that simulate a 45 Mbps link with 80ms round-trip delay between the data centers.

Note: The Riverbed Steelhead appliances are powered off during the initial phase of baseline testing, during which performance results are collected for non-optimized topology, referred as “Baseline no optimization” in the performance graphs. 

How to read test results
A ‘Baseline no optimization’ run is where SyncIQ is performing data sync operation to the target cluster Isilon storage array in absence of Steelhead appliances. A ‘Steelhead cold’ run is where Steelhead appliances are configured in the setup and the data is seen by the Steelhead appliances for the very first time. Subsequent runs of the same data between the clients and the Isilon storage array are referred to as ‘Steelhead warm’ runs. First time SyncIQ replication job with Steelhead also referred to as ‘Steelhead cold’ run shows incremental performance and subsequent replication jobs with Steelhead referred to as ‘Steelhead warm’ show dramatic data reduction and performance improvement when compared to the baseline performance. 

Email archive replication test
The chart below shows the Email Archive Replication time (in seconds) taken by SyncIQ to perform data replication from Cluster 1 to target Cluster 2. With Steelheads in the network the file transfers are always faster than when no Steelheads are present. It is important to note here that with warm cache on Steelheads it took 37 seconds to transfer 2.57 GB data over a 45 Mbps 80ms round-trip time WAN link with 0.01% packet loss. The same run took 1582 seconds to complete the same operation without Steelhead optimization. 

The next chart shows the Email Archive Replication Bandwidth usage by SyncIQ to perform data replication from Cluster 1 to target Cluster 2. With Steelheads in the network data transferred over WAN is always less than when no Steelheads are present. It is important to note here that only 9.63 MB of data was transferred over a 45 Mbps 80 ms round-trip time WAN link with 0.01% packet loss with warm cache on Steelheads. The same test required 2.80GB of data transfer to complete the same operation without Steelhead optimization. 

VMware virtual machine replication test
The third chart below shows the VMware Virtual Machine Archive Replication time (in seconds) taken by SyncIQ to perform data replication from Cluster 1 to target Cluster 2. Again, file transfer results with Steelhead optimization were impressive. It is important to note here that with warm cache on Steelheads it took 161 seconds to transfer 10 GB data over a 45 Mbps 80 ms round-trip time WAN link with 0.01% packet loss. The same run took 9997 seconds to complete the same operation without Steelhead optimization. 

The final chart below shows the VMware VM Files Replication Bandwidth usage by SyncIQ to perform data replication from Cluster 1 to target Cluster 2. The data transfer result showed significant bandwidth savings with Steelhead optimization. It is important to note here that only 24.89 MB of data was transferred over a 45 Mbps 80 ms round-trip time WAN link with 0.01% packet loss with warm cache on Steelheads. The same test required 11.00 GB of data transfer to complete the same operation without Steelhead optimization. 

Steelhead data reduction report
The Steelhead Optimization Reports below shows the Total Data Reduction % achieved for SyncIQ to complete data sync from one cluster to the other. The screenshot on the left shows 89% Data Reduction achieved with cold optimization on the Steelhead appliance. The screenshot on the right show 99% Data Reduction achieved with Steelhead appliances using warm optimization. 

The combination of the scalability of EMC Isilon Scale-Out NAS storage solutions with Riverbed WAN acceleration provides the means to maintain an always-on architecture giving remotes users fast access to centralized storage. EMC Isilon SyncIQ with Riverbed WAN acceleration also enables businesses to improve bandwidth utilization, minimize costs. Enterprises can replicate more data over longer distances without investing in costly WAN upgrades. . The results show dramatic improvements in recovery point objectives (RPO) and recovery time objectives (RTO) to replicate data across two Data Centers over a DS3 45 Mbps link with 80ms RTT latency Wide Area Network (WAN): up to 100 times faster replication and using up to 100 times less WAN Bandwidth.