Post a Comment
Feature
Mini-NAS: an unfinished masterpiece?
By Brian Dipert, Senior Technical Editor -- EDN, 10/26/2006
A premature hard-drive failure is a compelling motivation to pursue a Prying Eyes project, it seems. Such was the case with the Ximeta NetDisk I dissected a few months back, and it was also the case with this issue's Prying Eyes "patient." I'll get to my theories surrounding the HDD's demise in a minute. Let's begin with the overview.
D-Link's DSM-604H Central Home Drive (Google search, press release) was, as its name implied, one of the first NAS (network-attached storage) devices that targeted home and SOHO users. Its compact chassis befitted D-Link's stated long-term ambition to append the NAS function to a router, an integration aspiration that's as yet unfulfilled within the company's product line (and within those of other companies, as well). What's under the DSM-604H's hood, and what clues to D-Link's possible future plans does it reveal?
At first, when researching this piece, I thought the iXP420 was included in the technology, product, and headcount package that Intel sold to Marvell a few months back, but Marvell's PR person set me straight. Indicative of Intel's continued commitment, the company rolled out a new family of storage processors at its most recent Developer Forum.
|
The other important NAS memory, of course, is its magnetic rotating storage. The Central Home Drive family came in 20-Gbyte (DSM-602H) and 40-Gbyte (DSM-604H) variants, implemented using 2.5-in., 4200-RPM, parallel-ATA HDDs (whose standoffs are visible in the picture) and thereby explaining the unit's svelte size. An Acard ATP865-A PCI-to-IDE controller connects the HDD to the iXP420. Because the ATP865-A handles two PATA channels (each with primary and secondary drive slots), D-Link conceivably could have developed a multi-drive variant with RAID (redundant arrays of inexpensive disks) and JBOD (just a bunch of disks) capabilities.
The Central Home Drive is sparsely ornamented. Four front-panel LEDs alert the user to active power and device status, along with the presence of active LAN and HDD traffic. The back panel encompasses a power plug (5V, 2.5A), a 10/100-Mbps wired Ethernet connnector, and a reset button. Micrel Semiconductor's Kendin KS8721B PHY (physical layer) chip implements the Ethernet interface in conjunction with the MAC (media access control) transceiver built into the iXP420.
Note the prominent, unpopulated IC footprint next to the IDE controller, along with the numerous test points scattered across both sides of the board (especially along one edge of the topside). Did D-Link have additional plans for this hardware design, beyond the initial DSM-604H implementation? Was this particular board an initial-production prototype that got cost-reduced down the road? Did D-Link take directly to production an Intel reference design? Or is there another explanation for the hardware exorbitance?
Now about that hard disk. My DSM-604H contained a 40-Gbyte, 4200-RPM Toshiba HDD. At first, I'd occasionally notice that the NAS was "hung" when I awoke in the morning or after a weekend away from the office. Eventually, a reset (or later, more drastically, a power cycle) didn't succeed in reviving the unit, and an audible "click of death" (subsequently confirmed by pulling the drive, installing it in an enclosure and attempting to access it from my computer) confirmed the HDD as the culprit.
Overheating, as I previously encountered with the Ximeta NetDisk, may be partially to blame. The DSM-604H doesn't contain a fan, and used to get quite warm, even with plenty of unobstructed airflow around it. But fundamentally I tend to point the finger of blame at D-Link's decision to embed a hard drive originally intended for an infrequently used notebook PC in a NAS that's up and running 24/7. At the time D-Link released its NAS, industrial-class 2.5-in. HDDs, such as the SAS-interface drives offered by companies like Fujitsu and Seagate, weren't yet available. And I acknowledge that the 2.5-in. form factor enabled D-Link to make the DSM-604H smaller than would have been possible with a 3.5-in. HDD. However, the follow-on DSM-G600's 3.5-in. HDD suggests that the company realized after-the-fact that its DSM-604H size-over-reliability prioritization was flawed.
I mentioned earlier that D-Link's long-term aspiration was to embed the NAS function within a router. Has any other company succeeded in this endeavor? About the only example I can think of, aside from homebrew generic PC hardware running open-source software such as FreeNAS and Vyatta's router code, is the now-obsolete Toshiba Magnia Internet Appliance series (my example of which, come to think of it, ironically also suffered from imminent 2.5-in. HDD failure).
Alternatively, some routers don't embed HDDs inside but instead harness USB 2.0 ports to which you can tether external HDDs. Examples include ASUSTeK's WL-700gE and US Robotics' USR8200. The concept is pretty cool; not only can you treat these devices as NASs across your LAN, you can also (after navigating obviously necessary security safeguards) read from and write to them over the Internet via protocols such as FTP, HTTP, and BitTorrent.
