 |
www.conserver.com |
|
|
www.conserver.com |
Terminal Server BREAK-off Test Resultshttp://www.conserver.com/consoles/BREAK-off/breakinfo.html
|
|
Why we are doing this testing: We started this project because finding information on whether or not a device has the serial BREAK problem was more like tracking down an urban legend than science. Even with high confidence in a device, you can usually find someone who will suggest that some unidentified third party has had some trouble with that same device in the past. We have posted our test methodologies on a separate web page. |
I do realize that good science relies on larger sample pools. We are continuing to solicit loaner equipment that we can put through our testbed. As we run out of easy access to equipment, we will begin to accept more offers from other teams at universities and other large sites to perform this testing on their own gear using our test methods. We'll review their test results, and add those results to our database as 'external data' submissions, in order to identify which testing we have performed personally. In some cases, the manufacturers are contacting us directly now, which has made it easier to get equipment loaned for testing. (Manufacturers may want to look at the Frequently Asked Questions page for information about how to inquire about testing.
Our initial test run was intended to help prove how many devices had software-related causes versus hardware-related causes, and which devices do not exhibit the problem. (A side effect of the testing was that we could be confident about certain hardware, when suggesting the devices to customers that have BREAK-sensitive equipment.)
|
Devices that work with Conserver, and don't have the BREAK Problem We have not found any add-in cards for Intel-based servers which do not send BREAK when the PC is power-cycled. We have seen mixed results with a combination of in-chassis card with externally-powered interface pods (Cyclades PR series, Digi PortServer). If you have information about hardware that does well, please let us know. Since our initial interest was to find devices that we could use with the Conserver application, and which would be compatible with Sun machines, this has been the main focus of the testing. The devices that we will recommend at this point;
|
|
Here is a list of the vendor and models that we have tested so far. (Basically, we only want to see "yes" in the far-right-hand column for remotely controlling console ports.) Notice that there is some duplication of server models...these are cases where we were able to get more than one of each unit to test. Each line below represents one of the units that we have tested.
|
|
Footnotes to the test results;
|
|
Next hardware I'd like to test...
|
The Lantronix line has been evolving and expanding. (I've only just heard that Lantronix acquired the Lightwaves products last year, which rolls in some good high-security experience to a robust group of hardware folks!) The SCS should have some interesting features, and I want to try their user interface. Wiring for the Lantronix will be easy, since they use the same RJ-45 pinouts as Cisco console ports, so I can use my Cisco adapters on the Lantronix gear as well.
Other equipment we'd like to test...
Please let us know if you have any other suggestions for vendors devices that you think we should test. If you are a vendor, and would like us to test your terminal server, please contact us (email to break-off at conserver.com).
|
Note: You should be aware that the Console and Aux ports on many Cisco products will send serial BREAK when the device power is turned off. If you are connecting these devices to the serial port on a Sun device for control or monitoring, it is possible that the Cisco could cause the Sun device to halt when the Cisco loses power. As for other Asynch lines (other than the Console and Aux ports), these are the the results of testing on some of the Access products; 2509-11 series devices DO send BREAK when the power is turned off. 2600 and 3600 series devices (using the NM-xxA async modules) work fine! 2800 and 3800 series devices (using the HWIC-xxA async modules) work fine! The result of our testing is a very high confidence in the Cisco 3600 series devices (3620, 3640, 3660 tested) and the Cisco 2600 series devices (2621 tested) with the NM-32A modules. They have not sent a detectable BREAK signal under our test conditions. Cisco had an ECO (Engineering Change Order) for the 2511-RJ models only, which inhibited the BREAK when power was turned off. This ECO was not available for the non-RJ units. This was a change to the main board of the units, and was in effect for hardware revisions J and K, but the ECO was later revoked, deeming that the problem only affected a limited number of sites. In one 2511 that we tested, it appeared as though the DC Ground was removed when the power was switched off, resulting in some large potential voltages appearing on other signal lines, as the reference ground between the terminal server and the host was lost. In one test case, the target host was halted, and we actually lost data on the drive, requiring four FSCK passes to get the system back on-line. While I can't attribute this event to the terminal server sending BREAK, I will note that this was the only time that we have seen any data damaged on the target host during the testing. We have found one 2511 which does not send BREAK! I believe that this is attributable to the particular power supply in the unit, but I wasn't able to open it up to get more information on that particular unit. |
|
The RAS-2000 units MAY send BREAK (depending on which which revision of the serial card hardware you have); they key is the revision of the serial cards in the units, not the chassis! If BREAK is a problem for you, get the Rev. B boards for your chassis. With the older (Rev. A) boards, the ports will send BREAK when power is turned off and turned on. However, We tested a prototype of the Rev. B boards early in 2000, and found that they solved the BREAK problem. We are hoping to get a production Rev. B board once they get their backlog cleared. We also have a version of their IntelliServer to test (good for use at a co-location site when you only need a small number of ports). We are also negotiating to get an RCM24 or RCM48 unit in for testing (one RU tall, and 24 or 48 ports). Note: (Summer of 2002) Computone was acquired by Symbiat, and they no longer sell Console Server hardware. (I believe they will still sell you a hardware support contract for currently deployed Computone hardware...) NEWS: (March 2003) Altura XL has bought the Computone IntelliServer® RAS 2000 line from Symbiat! I have received a production version of the REX-16RJ (rev B card), and it has passed the BREAK tests. They say they are developing 24- and 48-port single rack-unit servers as well! I hope they will hire the technical writer, as I believe the older Computone documentation was probably the easiest to work with, regarding setting up console services. |
|
The AlterPath ACS 16, ACS-1, TS2000, TS800, and the PR3000 models have passed our BREAK tests. The PR-3000 device is actually a combination, stackable unit. There is a Main Unit, and then you can daisy-chain attach a number of multi-port Pods. During our early tests, the PR-3000 was the first device that seemed to have a BREAK problem due to software, versus hardware. We had found that if the main unit *and* the pod lose power at the same time, a BREAK would be sent when the main unit finished booting, and it then initialized the pod. Cycling only one or the other did not send BREAK. We reported our test cases and results to Cyclades, and they turned out a software revision within two weeks. We found that this had solved the problem. They report that this fix will be rolled into the standard code tree, so any newer revisions should include this fix. We have returned our testing unit to Cyclades, at their request. They are phasing this product out of the sales chain, though they will be continuing to sell it in a few markets, and they say they will be maintaining support for units already in the field. We received our TS2000 unit to test, and it has performed in the BEAK test very well. These come in a variety of port densities, from 4 to 48, and they even sell a 1-port (TS100 unit for applications where you really need an ethernet-based serial port to talk to a console, etc.) You may have noticed the Linux Inside icons on the Cyclades web site, and the stickers on their TS products. We now have the Software Developers Kit for the TSnnn family, and we will be trying to port Conserver to the unit directly. (If you are interested in the outcome of this effort, please sign up for the Conserver Announcements email list on the Conserver main page.)We received an ACS 16 in early December 2003 for testing. The unit has a dark blue paint coating, darker than the TS2000 pebble-finish paint job. The other difference that caught my eye on first look was that the status LEDs for each port are now on the RJ-45 jack itself, which is good if you want your LEDs and cables on the same side of the server. Finally, there was an assortment of DB-25 and DB-9 pre-wired adapters in the box, as well a male-RJ45-to-female-RJ45 adapter to connect a Cisco console to the Cyclades AlterPath RJ45 wiring scheme. I'm glad to see this type of hardware available from the vendors. As for the BREAK testing, the unit passed, but this was something I expected. The hardware under the hood is largely an expansion of the TS product line, as far as the serial path hardware is concerned. There is more memory in the unit, and the two PCMCIA slots on the opposite side from the serial and power connections. In the fall of 2003, I also had the chance to test a couple of Cyclades smaller units! The TS800 (8-port unit) is small, easily portable, and doesn't send BREAK. This is small and light enough to carry in your laptop case, giving a field service person the ability to quickly connect 8 serial devices to their laptop for field work. It's also a good low-power device to use with your UPS, to ensure that you can get console access to those few critical devices you have at the core of your critical networks! The other unit was a prototype ACS-1, which is a 1-port version of the ACS (giving you the power of the ACS line, as well as the PCMCIA slots, and the RS-232 port and RS-422/485 ports as well. This unit is a bit larger than the TS-1 (the ACS-1 I have is 1.5" x 6.5" x 4"), and uses a middle-of-the-cord single-voltage switch-mode power supply. This is a beefy little computer, with a few interfaces, and a single serial port to connect to an adjacent device, and will probably lend itself well to Home Automation. NOTE: Cyclades has produced a good booklet, supporting remote console management, and comparing it to other system management options. You can get a free copy from their website, on most of the product pages you'll find a link to "request your free copy" of their Console Port Management booklet. |
|
DEC Network Products Group (http://www.dnpg.com/) makes a wide array of network connectivity products. The DECServer product line still supports DEC protocols, but now also include a reverst-TCP capability as well. If you have a DEC environment, the DECServers will give you console access as well as providing DEC protocol support for serial printers and more. The DECServer 90+M line is small device, that can be used as a stand-alone unit, or as a module in a larger chassis, to add serial conenctions with Ethernet connections. |
|
Digi (http://www.digi.com/) makes a wide array of serial connectivity products, including the PortServer TS line, and the "Secure" PortServer CM lines. The PortServer TS line is small (front-to-back), and lightweight, making it easy for one person to install it. Configuration was also easy, out of the box. The one part I had trouble spotting was that the ports needs to be configured as 'printers' to allow the reverse-TCP connections from Conserver. Use the command; set port range-1-16 dev=prn There are a few LEDs on the front of these units, but the functions are basic. If you like LEDs to see the status of your serial lines, take a look at their CM series before you by a TS unit. I tested a CM 8, and it performed very well in the BREAK tests. But, I was also impressed with their documentation. The steps in the manual follow the structure of their Command Line menu tree (which is also reflected in their Web-based config GUI), but they also have a great 'getting started' overview poster, which will help get admins working quickly. The CM software also supports PCMCIA cards, and the system is geared to easily configure various PC Card devices and interfaces including memory drives, NICs, and wireless NICs. I found this aspect of the configuration to be much simpler on the Digi CM, because of the software and the manual combined. I also tried testing the integration with other power controllers, but the firmware I was testing only recognized a BayTech-DCD device. (I was surprised that the software didn't list a selection for the Digi RPM 8 power controller. Perhaps that is an OEM'd product, or maybe I have old firmware? I will look into firmware upgrades this fall.) |
|
Made by Chase Research, the IOLAN Rack+ devices DO send BREAK when the power is turned off. These are rack-mounted units, meant for data center access. It seems that I was mistaken, when I posted here that the old Chase IOLAN line was being discontinued. The products were picked up by Perle Systems. The rackmount servers still send BREAK, but we've tested the model 104 and it does not have a serial BREAK problem. This is probably the 4-port unit that ASP Technologies told me about in December 2000. These units have an external, single-voltage power pack. There is a model 108 as well. |
|
iTouch Communications picked up the legacy Xyplex line from N-Base, and they are continuing to support the older units. They have also been working on designing newer units to expand the family line, and they have been designing their new hardware to be BREAK-free, including their replacement series for the Xyplex 1600 (look at the 8000 and 9000 product lines.) We have tested two In-Reach Manager devices. The IR-MGRM-101AC devices DO send BREAK. The IR-MGRM-111AC devices, however, work fine! (Look for the "Sun Certified" symbols on the front of the units.) We tested an IR-8020 unit, and it also performed well. (These seem to be the non-BREAK Heirs to the Xyplex 1600 series, since the RJ-45 serial pinouts are the same. In mid-2003, MRV Communications, the parent of N-Base and iTouch, decided to start branding the console products themselves, including a color change to black. They really like Flash animations on their website, and it's hard to skip the intro, so here is a link to their In-Reach products main page. |
|
Moxa Technologies (http://www.moxa.com/) manufactures a wide variety of computing products and accessories. They have a variety of devices that work well to provide remote serial ports under Microsoft OSs, as well as a few devices that will provide remote serial ports for UNIX and Linux platforms. Their Terminal Server family has basic dial-up and remote printing capabilities, but they also support a RTELNET that serves a a Reverse-TCP service, which can work with a console management server. I tested the CN2510, and found that it passed our BREAK testing. The programming of the device was done using a Command Line Interface, using cursor keys to navigate. |
|
OpenGear (http://www.opengear.com/) is a new player in this field, and they are using an Open Source model for code development. It will be interesting to see how well this does, in the area of addressing any SSH vulnerabilities that may turn up in the future. (Their pricing is certainly lower than most of their competition!) I tested a CM4116 server, and it passed the BREAK tests well. The setup was easy, even though I didn't have the quick-start guide handy. The only complication I had was due to their device RS232 console speed being 115.2 Kbps. (This was noted in the quick-start guide, and it's one of only 3 devices I've found which used 115.2 Kbps as a default speed.) After finding the correct speed, the set-up was very intuitive (to me). |
|
Perle Systems (http://www.perle.com/) markets a number of network products. They also bought up the Chase IOLAN product line. The Perle CS9000 is designed to be a Console Server first, although it also has plenty of dialup/terminal server features. One handy example is that the CS9000 starts off with ports set for reverse-telnet service, right out of the box. It will try to configure itself using DHCP, but it also has a dedicated "admin" (RS-232) serial port for configuring the unit manually. Another fine feature is the way the hardware was designed to avoid the BREAK problem. (The unit I tested was probably the cleanest power cycle I observed at the serial port.) And the units boot quickly in stand-alone mode. One feature that I really like is that the console server ports use an ethernet NIC-style RJ-45 jack, with an LED on either side of the locking tab of the jack...one LED indicates input traffic on that port, while the other LED indicates outbound traffic. Having these LEDs on the same side of the units as the cables is a good thing, in my opinion. Using these illuminated jacks is a great idea, as the LEDs are literally part of the jack, so you don't have to think hard to compare a bank of LEDs to a bank of cables. The IOLAN model 102 and 104 (2-port and 4-port desktop units, respectively) use a single-voltage DC-based power pack, and they do not have the BREAK problem. These may work well in small sites, where you'll never need more than 4 ports, and you definitely need something that will not send BREAK unless you tell it to. |
|
OpenGear (http://www.opengear.com/) is a new player in this field, and they are using an Open Source model for code development. It will be interesting to see how well this does, in the area of addressing any SSH vulnerabilities that may turn up in the future. (Their pricing is certainly lower than most of their competition!) I tested a CM4116 server, and it passed the BREAK tests well. The setup was easy, even though I didn't have the quick-start guide handy. The only complication I had was due to their device RS232 console speed being 115.2 Kbps. (This was noted in the quick-start guide, and it's one of only 3 devices I've found which used 115.2 Kbps as a default speed.) After finding the correct speed, the set-up was very intuitive (to me). |
|
(Also see the iTouch information below! The Xyplex, iTouch, and MRV Communications products share the same RJ-45 serial wiring pinouts.) We have tested the 1600, 1620, and 1800 units. These devices DO send BREAK when the power is turned off. We are expecting the loan of a 1640, to see if there is any difference with that model (but we are expecting that it will also send BREAK). The folks at ASP Technologies make a special power control modules that can be installed inside the case of the Xyplex units listed above. These require a soldering iron to install. You can find more information about these dongles from their hardware website. (The dongles were designed for use with the Digi SCSI-attached devices, but have been used successfully in Xyplex units.) |
|
NOTICE: Most of the pages, articles, and tutorials on this website are copyrighted works. You may make 'deep links' to various pages. (If you let me know which page(s) you are linking to, I'll let you know if I move the page(s) during updates.) Please send me email if you wish to republish any material, or use it on your own website. |
