Twin Studies

Performance and Monitoring

A direct comparison of benchmark results would not be particularly useful in this case: The server on which I installed the Nexenta Community Edition had two quad-core Xeon processors, and thus more than enough computing power; Gigabit Ethernet provided sufficient network bandwidth. The server also had a generous RAM allocation of 16GB; however, Nexenta recommends 48GB, or even 192GB for deduplication because most of the information for this operation is stored in RAM. With the recommended amount of memory, performance probably would have been faster, but at a higher cost.

In each case, you have to judge what difference the disks make in relation to price. For technical reasons, I could not use identical disks in both storage systems, which is why the measured values are not absolutely comparable. Also, in the test setup, only the Netgear system had an SSD module, which I set up temporarily as a read cache.

Middling Fast

A look at a few numbers might help in understanding orders of magnitude. The Netgear system achieved an average read speed of 60.8Mbps and a write speed of 45.3Mbps on a connected iSCSI RAID-6. The Nexenta installation was slightly faster, reading from a RAID-Z2 volume via NFSv4 at 66.8Mbps. All of these values (determined in each case with IOzone in throughput mode, a 64KB record size, and a 2GB file size) are not bad, but not surprisingly good; they are middling.

On comparison of the monitoring data, Netgear scores right off the bat. First, it knows its own hardware and can thus more easily monitor it – for example, it provides temperatures and fan speeds. Second, it offers RRD charts of the most important parameters: volume and network throughput and space utilization (Figure 3).

Figure 3: The Netgear system clearly has the fancier charts. It can also provide better hardware monitoring, because it knows what it is running on.

The NexentaStor Community Edition, however, not only offers gauges (Figure 4), but charts as well, and you just need to click to put them together (Figure 5). The advantage of Nexenta, again, is the greater flexibility. Admins can choose from around 50 parameters to display individually or together in bar or line charts.

Figure 4: Instead of charts, Nexenta uses gauges for a status overview.

Figure 5: The NexentaStor Community Edition also supports charts. They involve more interaction but are also more flexible.

Conclusions

Netgear's ReadyDATAOS provides an easy-to-use, turnkey solution with a nice look. If you do not want to delve too deeply into the topic of storage technology, you will find this conveniently usable software. Additionally, it gets bonus points for better monitoring of the specific hardware, integrated replication, and hardware and software from a single source.

Compared with the Netgear product, Nexenta exclusively supports RAID-Z3, providing a text console, and in many places offering more granular options for customizing the configuration, The downside is that it also requires more expertise. The free edition of Nexenta is not bound to a fixed hardware, which opens up degrees of freedom but also limits the possible matches between software and factory hardware.

It is not surprising that convenience comes at a price: Netgear's asking price is EUR 5,200 (~US$ 7,000) without disks and around EUR 7,400 (~US$ 10,000) with. A suitable server that could install the community version of Nexenta, including several disks, would likely cost half of the EUR 5,000 price without disks. Outside the private or academic sector, you would need to purchase the Nexenta Enterprise Edition; there, an entry-level configuration with preconfigured hardware would also cost barely less than EUR 5,000.