We assessed the Netgear ReadyNAS 3100 RNRP4420 in terms of usability, features, and NFS and iSCSI I/O performance.
Usability and features testing consisted of setting up the device on Network Test's production, first to function as an NFS server, and then again as an iSCSI network-attached storage (NAS) device. In the NAS case, we created Windows and Linux virtual machines using VMware vSphere 4 on VMware ESX 4.0 hosts, and used the ReadyNAS as the datastore.
We also assessed the ReadyNAS device's ability to conduct other common management tasks, such as configuration of administrator rights; software upgrades; and connection to uninterruptible power supply (UPS) systems.
For performance testing, we again used NFS and iSCSI, but this time on an isolated test network. For NFS testing, we used the Mu Test Suite tools from Mu Dynamics to generate a sequence of events between NFS client and server.
In the NFS sequence, a client mounts an NFS volume on the ReadyNAS device; creates a directory; writes a 10KB file to that directory; deletes the file; deletes the directory; and unmounts the drive. All this was captured during an actual exchange between an Apple MacBook Pro and the ReadyNAS device.
The Mu tools then manipulated the packet capture so that repetitions of the sequence used different source and destination ports and directory names each time; in effect, each session looked to the ReadyNAS like a unique event. This is a vast improvement over capture/replay tools that simply repeat the same sequence, often overloading the device under test due to socket contention issues.
The Mu Test Suite ran a script that measured response time, failures, and timeouts for various numbers of concurrent sequences, ranging from 1 to 512 concurrent sessions. In all cases, the test duration was a 60-second ramp-up phase followed by a 300-second steady-state phase. The measurements for response time and transaction and error count presented here are taken only from the steady-state phase.
The actual NFS script used for this project is available for download from Mu's pcapr community site.
To measure iSCSI input/output rates, we used IOzone, an open-source filesystem benchmarking tool. The goal of these I/O tests was to compare virtual machine disk performance between local storage (the virtual machine and datastore resided on a VMware 4.0 ESX host) and iSCSI storage (the datastore resided on the ReadyNAS 3100, acting as an iSCSI target). We used 64-bit versions of CentOS 5.5 and Microsoft Windows Server 2008 R2 for both tests.
IOzone measures various disk operations for file sizes between 64 bytes and 4GB, using record sizes between 4KB and 16MB. While IOzone tests for 13 types of disk operations, we focused on four of the most common: reads, writes, rereads, and rewrites. Performance may be higher for reread and rewrites because the filesystem may cache data that it previously has seen (in the reread case) or works with metadata that already exists (in the rewrite case).