Re: OnPlus - NSS300 Series Detection

Kurt Schumacher · ‎05-20-2011

The NSS300 (NSS322/324/326) NAS series is offered and sold in the market as a "Smart Small Business Network Access Storage"

What does OnPlus make out of it? A device type MEDIASERVER, device class ENTERTAINMENT. COmplaining abiut that this multi-thousand US$ device supposed to hold data, backup data, and replicating business data to the external backup servise i.e. by Cisco (Mozy). And OnPlus does - still after almost one year now - still make it a fancy mediasever, despite of the fact that WMicrosoft File Sharing, AFP, ftp, ... and everything is available.

The detecting priorities are simply not business-class. Repeated reports onver time did not change antyhing. Ridiculous, isn't it? Does cisco expect we are going to manually rename and completely reconfiguring all NSS after the automatic detection?

Worse: Even after changing the driver type (which was not detected before...) to NSS300, it's still a MEDIASERVER, shows the TwonkyMedia logo.

How shall we sell this to our customers?!?

Somehow I can understand, the NSS324 here again not detected as a non-Cisco device. If applicable: Anybody please be so nice to tell me the manufacturer ID used in the device MAC address configured in the detection algorithm? No problems for me changing the MAC address - psst.

New face - and the very basic issues still not addressed? Gentlemen and Gentlewomen: This is not good enough for the real-world market!

Michael Holloway · ‎05-20-2011

Hi Kurt,

I have one of these on my network, and I also see the TwonkyMedia server being advertised with it's UPnP icon. You'll probably notice a second device on your network (with the same IP address but a different MAC address) that is being recognised properly as a Cisco NSS324. The OnPlus driver for this series of NAS devices has a bug in it currently, where the multiple MAC addresses aren't properly seen by the agent as the same device. We'll get this fixed in a future release. Thanks for pointing it out to us.

-mike

Kurt Schumacher · ‎05-21-2011

Not at all! The NSS324 here has just a single TCP/IP address, and is connected by a 802.3ad LACP port trunk - so only one MAC address is visible, and the NSS is detected and picked once - isn't it?

So that we can work on the same level, please let me know which manufacturer MAC address bytes are configured for the NSS300 Series driver! The unit here is a beta sample, still carrying the ICP Electronics, Inc. 00:08:9B:BD:91:EA, cisco_device is set to false.

[~] # ifconfig
bond0     Link encap:Ethernet HWaddr 00:08:9B:BD:91:EA
          inet addr:10.10.1.114 Bcast:10.10.1.255 Mask:255.255.255.0
          UP BROADCAST RUNNING ALLMULTI MASTER MULTICAST MTU:9000 Metric:1
          RX packets:3495717 errors:0 dropped:0 overruns:0 frame:0
          TX packets:457188 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:567619933 (541.3 MiB) TX bytes:141664506 (135.1 MiB)

eth0      Link encap:Ethernet HWaddr 00:08:9B:BD:91:EA
          UP BROADCAST RUNNING SLAVE MULTICAST MTU:9000 Metric:1
          RX packets:3322218 errors:0 dropped:0 overruns:0 frame:0
          TX packets:364258 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:545844976 (520.5 MiB) TX bytes:113424626 (108.1 MiB)
          Memory:feae0000-feb00000

eth1      Link encap:Ethernet HWaddr 00:08:9B:BD:91:EA
          UP BROADCAST RUNNING SLAVE MULTICAST MTU:9000 Metric:1
          RX packets:173499 errors:0 dropped:0 overruns:0 frame:0
          TX packets:92930 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:21774957 (20.7 MiB) TX bytes:28239880 (26.9 MiB)
          Memory:fe9e0000-fea00000

Sorry to re-launch the same discussion again: OnPlus is designated to help us offering and providing better services to SOHO and SMB. These are NOT home environments! There are about half a dozen of much more useful fingerprints (and information strings) for a NAS (and many other computers, including Windows systems with the Windows Media Player that are now detected as a computer, but are still tagged as WIndows Media Connect devices...). UPnP has the lowest priority. Long before, we must see the Micorsoft File Share announcements, the Apple FIle Protocol (AFP) announcements, the ftp announements, ....and near to the end UPnP.

Just to extent the complaint of the detection preferences (which was promised to be addressed and enhanced a long time ago!) - Normal Windows (7, Vista) worstations with WMP11 are also detected as TOYS, and not as a Windows desktop or notebook computer! Sorry, we can not deploy OnPlus to real customers when thier most paert of the valuable business infrastructure (Windows systems, NAS, Windows Home Servers, ...) are detected, typed, and classified as UPnP toys. We are NOT willing to review and edit each and every key device detected, so it shows up correct in the reports - which are supposed to be presented to the customers at the end of the day!

Configuring the NAS (lack of any smart or policy routing on the Linux bas OS) with two interfaces to the same TCP/IP subnet is deprectiated. You will have all the sudden issues - especially when it comes to traffic passing devices with correct working stateful firewalls. The return traffc can and will be returend on a different interface, as the return traffic is not fixed to the incoming interface. This does led to a triangle route then - completely undesired.

When you are looking specifically on Twonky: Unless I'm wrong, Twonky can only bind to one of the two interfaces - the service is not reliable on both networks even if two different TCP/IP subnetworks are configured. Not sure if this was addressed by the OEM in the mentaime, for example broadcast and multicst used to be not available on eth1 - fatal also for other applicaitons when operating the NAS in a dual-network configuration.

Please address this long standing issue, change the appliance detection priorities to business factors, and let us know when we can run a full rescan. Otherwise we see no way to promote and push OnPlus to the real business world.

TIA,

-Kurt.

mrn · ‎05-21-2011

Hi Kurt,

Discovery of devices on OnPlus is complicated by the fact that we have information coming in from something like a dozen sources. The records from multiple sources are merged, and in cases where there is a conflict, an arbitrary priority is used to break the tie.

I am responsible for the discovery code on the NSS300, and I did not test a configuration with any sort of media server running, so I did not see the result you are showing.

The NSS discovery is complicated in addition by the fact that, at least on the firmware I am using, the UPnP advertisements sent by the NSS are broken. (You can observe this on a Windows machine by clicking on the icon in Network Neighborhood - it shold take you to the NSS web page, but instead it does nothing.) Our driver can't resolve the information and returns no UPnP information at all.

In the Device Info tab, can you tell me what value shows for your NSS under "method"? I would like to know whether the Bonjour advertisement was seen by the discovery software.

- Mark

Kurt Schumacher · ‎05-23-2011

Mark,

This issue I'm complaining about sonce the early days of Thunderbolt is not limited to the NSS300 NAS only - but as a Cisco device, the problem is the most obvious here.

As OnPlus does create reports wich are suposed also to be delivered to the end user - our customer - we insist, that industry standard computers running Windows Vista, Windows 7, Windows Server 2003, Windows Server 2008, OS X clients, OS X servers, generic Linux workstations, and generic Linux servers - in virtually any configuration - are listed accordingly in an automated way. And this must be in the first priority following the typical business requirements. There is nothing wrong to have i.e. a DLNA/UPnP server available.

This NSS is on v1.4.0.7 which is actually in the Cisco release process. The issue was the same with all previous firmwares (0.9 from EFT, and the released 1.1, 1.2, and 1.3.0.x) - and all services are enabled and active.

Talking of the NSS300 ... again, when your first priority is taking high-priority information from UPnP ... things will go wrong. Here and on other systems listed above.

There are teo UPnP announcements coming from the NSS - both are valid, and all links are correctly working. There had been a few issues on UPnP during the NSS beta which were related ot the OEM adoption.

Let me guess: In your opinion, the port 49152 is wrong - just because "K.I.S.S" implemented UPnP devices use a reference to the effective Web UI?

Afraimd I have to tell you: The presentaiton URL here is correct - the admin session on the effective port is established on a normal Windows system... (edit: the 404 was a chache issue - I was wrong). Here is the sequence with the JavaScript redirect code:

But now I became curious - why do you think the UPnP behaviour is not correct? The code above certainly works with any kind of browser supporting JavaScript.

But then, when reading the UPnP Device Architecture guide...

UPnP Device Architecture http://www.upnp.org/specs/arch/UPnP-arch-DeviceArchitecture-v1.1.pdf

HTTP/1.1 Redirect restrictions

HTTP/1.1 defines OPTIONAL support for redirecting an HTTP request. UPnP 1.1 devices MAY redirect a request, although this is

NOT RECOMMENDED. If a UPnP 1.1 device redirects a request, it MUST respond with a “307 Temporary Redirect” HTTP status

code (see also RFC 2616). UPnP 1.1 devices MUST NOT return any other HTTP/1.1 redirect options. Control points MUST

implement HTTP/1.1 redirect and SHOULD redirect the request upon receiving a “307 Temporary Redirect” HTTP status code

(see also RFC 2616).

Certainly, your UPnP discovery mechanism does not interpret JavaScript. Because according to the text above, it does not have toSomehow, I could start to agree ...

Regards,

-Kurt.

Efrat Noy · ‎05-24-2011

Hi Kurt,

Thank you very much for your participation in the OnPlus trail! Your input on OnPlus Beta Trial is extremely valuable as it helps define the future of OnPlus going forward and towards production, which is scheudled for the US in the Summer of 2011.

Internally, we are actively discussing the feedback you've given us, in particular identifying a fix/plan for teh following items you raised:

- NSS SMB/NetBiOS issue

- discovery of Twonky media server vs. NSS

- SMS target configuration

- view of availbale VLANs on a network.

Once a plan is in place to develop these features/fixes, we will notify you and the rest of the community, of the availabilty of a Beta version so you can review and evaluate these enhancements.

Again, we greatly value your participation in the OnPlus trial, and see you as a valued represntative of the broader population, to give us your field knowledge and insight to problems and solutions we present.

We will be contacting you as well to discuss your concerns and feedback, so we can do that live, and make the communication more productive.

Thank you!

Efrat

Kurt Schumacher · ‎06-01-2011

Dear Efrat,

Thank you for the nice follow-up. What concenrs us is the fact that all the issues are in fact re-discoveries of issues reported in the very early days of Thunderbolt.

Looks like you need some adjustments and arguments:

1. We expect a typical SOHO or SMB network can be scanned, and is delivering a 99% correct and business-wise appropriate detection of the network devices - including NAS; Servers, and Workstations.It must not be required to completely rework almost >90% of the typical hosts detected on suc a business network.Yes, I have a long-term (>25 years) HP OV and IBM Tivovli (based on HPOV code licnese originally) background - honeslty, we have barely seen auch poor and mis-guided detection.

2. The key reason to deploy i.e. Sx300 or more expensive managed switches is the ability to maintain VLAN.in an easy and reliable way. Obvious, in most SOHO and SMB environment,s we have deployed more than a data and a voice VLAN. Otherwise we can go with unmanaged Sx100 or simple Web managed switches like the Sx200. SO either the supported switch range is wrong in OnPlus, or the functionality is simple below basic expectations.

3. I'm concenred the selected SoC platform (Marvell Kirkwood 88F6281 in the SheevaPlug, potentially 88F6282 to gain the performance firmware RD is always looking for in the final appliance design [?]) has massive limitations and drawbacks when it comes to VLANs on the network side.

4. It was announced on the OnPlus conf call that there are no more asymmetric (half) notifications, and any DOWN is followed by an UP event. Afraid this is not the case on the OnPlus Connection Status (for example). I'll post an dedicated entry on this one. Oh: Yes, not new again.

It really looks ot me that there is virtually no ability or willingness to change OnPlus the way the market wants and expects it. No way ot go for the market this Summer in my opinion - simply not good enough yet.

Regards,

-Kurt.

Efrat Noy · ‎06-01-2011

Hi Kurt,

Thank you again for the candid feedback. your points are taken and discussed.

One point to mention regarding VLAN management, and especially on the 300 series managed switches - those are supported in OnPlus Beta.

Therefore, if you provide the credentials to it in its product window, the OnPlus network agent will be able to discover all teh VLANs that are managed in that switch, and show the devices and network topology on those VLANs.

Does thsi address your concern? I just wanted to make sure you're aware of this. If not - maybe we can have a 1:1 so you can explain better what you are looking for.

On your other points:

As mentioned before, we are discussing a resolution to the discovery issue you raised.
What is your concern with teh Marvell chip? The software that runs on it is fully developed internally by Cisco, and it goes thorough quality assurance procedures to assue product readiness.
RObert has addressed yoru concern on assymetric notifications. Hope thsi helps.

I'd again encourage to have a 1:1 discussion so you can share with us your feedback and we can then share our thoughts for teh future as well.

Please let me knwo when woudl be a good time for you, by replying directly to me (enoy@cisco.com).

Thank you,

Efrat

Kurt Schumacher · ‎06-14-2011

Hello Efrat,

Sorry for slightly leaving the original subject topic here.

It can't be, that devices configured to diffeent VLAN are presented as beein in exactly the same broadcast domain like the main resp. the appliance network. This is definitively wrong. It's a different network, and must be presented as such - in an individusl tree, default rooted at the Sx300 or PBX as used to detect them all.

Adjustment? Well, here again: We're selling the customers the need for dedicated VLAN, and start typically with the voice one. Now every vieew and every report is creating a false flat-network view again.

And no - unless I was reading to fast - the asymmetric DOWN and never up is not answered - it's a cheep excuse for poor mangementand/or decision only - see the dedicated thread.

In my opinion, the appliance should be able to scan many networks VLAN directly. This requires a connection to a switch trunk port, the appliance management interface configured into the untagged or a tagged VLAN, and must be able to scan more VLAN - and present in a correct manner again. From what I'm aware of, this is a flaw of the Marvell SoC in place.

Adjsutment again? In this business, you can't expect the customers have the latest and greatest Cisco SB switches in place alreeady... The key intetion is to seel more and better services, and more Cisco SB hardware at the end of the day. None of these targets is reached. Tis comes paired with a potential wrong hardware platform in place...

Nothing of all that is new again.

I'm happy to go into a future discussion again. Please understand, that I am slightly tired riding on the same horse, finding many basic essentials have not changed...you are welcome to come back when a Windows Worksation is shown as a Windows worksation, a Windows Server is shonw as a Windows Server, a Cisco NSS as a Cisco NSS, a Linux System as a Linux System - not as a toy, not as a media server or player - regardless of the services enabled on these systems. then it's time to look into more.

When you reached that step, you will find enough qualified partners worldwide to implement OnPlus.

Before it's a prototpye. Still is - sorry all.

Regards,

-Kurt.

jopenman · ‎06-13-2011

Hi Kurt,

Ive done some checking on this issue and wonder if you have Bonjour CSCO-SB turned on? When this is enabled (should be on by default), the dicovery information we get from the NSS overrides the uPnP discovery info and shows the NSS correctly in OnPlus. When this is turned off, we do not get enough information from uPnP to properly determine the device (productId, ViD is missing) and thus the device displays as an entertainment device (from uPnP).

Please let me know if this changes or fixes what you are seeing.

Regards,

John

Kurt Schumacher · ‎06-14-2011

Hello John,

Thank you for taking care of the feedback.

Yes, the "CSCO-SB (Cisco Small Business enabled devices)" service is enabled in Bonjour. For operational reasons in an OS X and Windows mixed environment, the SAMBA announcement is switched off. This is done to avoidthe OS X users are randmomly using AFP and SAMBA servces, which is leading to all sudden issues (say files opened in AFP appear to be locked whnhen coming back by SAMBA...). Altnrnate approach would be using SAMBA only, but this is not the perfered mehtod for OS X.

Guess oyu realized the Bonjour announcements on the NSS are not following the suggested way - announcing a (single) host and a list of services provided. Instead, all service announcements on the NAS are in fact individual host announcements - thus thee are individual names required. I'm fighting wotu the NSS OEM for a long time to substitute the actual code ie. by Avahi.

Regards,

-Kurt.

jopenman · ‎06-14-2011

Hi Kurt,

Would you mind if we logged into your OnPlus unit to run some tests?

Regards,

John

jopenman · ‎06-21-2011

Hi Kurt,

It appears that with the new load pushed to all the Beta sites this issues has been resolved. Please let me know if you are still seeing the issue. I checked your network and the NSS shows correctly as a NAS.

Regards,

John