I last discussed the deployment of Network Attached Storage (NAS) in June last year, where I suggested the adoption of NAS to meet the unique needs of small- and mid-sized businesses. I also talked about a number of features such as deduplication, cloud synchronization and hardware-accelerated encryption to watch for when buying NAS.
A recent article by George Crump on InformationWeek, called "What To Look For In SMB NAS," got me thinking about the changes in technology since then, and its impact on NAS. Like me, Crump looks beyond the acquisition of NAS in terms of merely garnering sufficient storage to meet current needs. While vendors have implemented various strategies to improve the performance of current NAS with the use of SSDs (solid-state drives) and capabilities such as auto tiering, Crump argues that such moves will only help performance by "only so much."
Casting an eye on the development of Ethernet networking, he identified the limitations of Gigabit Ethernet as an issue that could result in a performance bottleneck on the scalability of storage in the near- to mid-term. Crump noted that:
You need to make sure that NAS can scale to meet tomorrow's capacity needs without having to re-buy your NAS every few weeks. There is also the increasingly real potential of a performance bottleneck caused by many users overwhelming a single 1-Gbps Ethernet connection.
While the most direct solution would be to upgrade to 10Gbps networks, the less obvious fact is that 10Gbps Ethernet is not likely to be cost effective for most small- and mid-sized businesses. For sure, there are projects that implement 10GbE; I recently chatted with a network engineer friend who was involved in one such deployment. This particular project entails setting up a high-speed computer trading platform for a Stock Exchange, however, which is hardly SMB-level stuff.
The solution, points out Crump, would be to trunk multiple 1Gbps Ethernet connections together using a less-talked about feature called "link aggregation." To put things into perspective, it would be far less cost prohibitive to acquire Gigabit Ethernet switches with link aggregation capability than to roll out 10GbE infrastructure today that you may not even need.
Larger systems such as SANs with Ethernet ports tend to have link aggregation built in, though NAS servers that support it are still limited, necessitating a more careful acquisition process. This fact is not lost on vendors either, and is likely the reason why storage specialist Synology recently launched the xs series of NAS servers, which sport a generous four Gigabit Ethernet ports per appliance. Moreover, they can also be upgraded with another dual 10GbE ports via a PCI-E add-on card for future expansion. In fairness, I must point out that older Synology NAS such as its DS1511+ already comes with dual-port link aggregation.
Interestingly, I mounted a search for network switches capable of link-aggregation a couple of months back and am looking at one such device from Hewlett-Packard right now. I shall be presenting a review of it next week, so stay tuned.
In the meantime, have you implemented an NAS solution for your SMB recently? What were your experiences like and the primary considerations in terms of future growth? Feel free to add a comment below, or drop me an email.