Jump to content

Suitable Hardware


Vodia PBX

Recommended Posts

This topic is a little bit off the mainstream topics, however I feel that it has some practical relevance for most of the customers who need to run the PBX on their premises.

 

For example, Foxconn offers a bare-bone PC (http://www.foxconnchannel.com/ProductDetail.aspx?T=NanoPC&U=en-us0000020#) with a dual-core Celeron.

 

Also, Intel seems to have interesting devices (http://www.intel.com/content/www/us/en/motherboards/desktop-motherboards/desktop-kit-dccp847dye.html) based on Celeron.

 

There are other companies like Impactics (http://www.impactics.com/produkte/d1nu-series/) or Fanlesstech http://www.fanlesstech.com/2013/04/shuttle-ds47-first-pictures.html?m=1 that offer embedded PC as well.

I know this topic is far from complete; if you have more please feel to add other links (please no advertisements, though).

Link to comment
Share on other sites

Hi,

 

I am glad that you brought out this topic. I have been looking for solutions and have already tried a couple that I can share. I already checked the MB-D38 that appears on the following webpage: http://www.ewayco.com/100-low-cost-pc-products-low-cost-systems-embedded-systems-servers-lcd-pc-panel-pc.html I added a 2.5" 32G SSD and 2G of DDR3 RAMto round it up and it does work. Its pros are: It does work, Debian Linux 64 bits runs there with no problems, its small form factor. Its cons, at least to me are: Coming from China to my country, some 45% of cost is shipping and duties. Although it is Atom based, it does have a couple of small fans which in such a reduced form factor makes the solution a bit noisy, which may not be important in some sites, but it may be a problem in others. Something that I do not know if it is a con or a pro, but that grabbed my attention, is that the chassis appears to be made of thick caliber steel. It would not surprise met hat it could stop a bullet. I also tried assembling a pc myself. All the needed parts are available in my town, exception made of a small good looking chassis and corresponding power supply.The closest chassis that I found was a slim mid tower unit. Not so cool looking, but the unit worked and was affordable. That partial success encouraged me to a next project. Import just the chassis and PSU and find everything else at home. The chassis that I will try next is the M350 http://www.mini-box.com/M350-universal-mini-itx-enclosure I will let you guys know. The chassis is still somewhere on the road.

 

Cheers!

 

Link to comment
Share on other sites

Hi,

 

I am glad that you brought out this topic. I have been looking for solutions and have already tried a couple that I can share. I already checked the MB-D38 that appears on the following webpage: http://www.ewayco.com/100-low-cost-pc-products-low-cost-systems-embedded-systems-servers-lcd-pc-panel-pc.html I added a 2.5" 32G SSD and 2G of DDR3 RAMto round it up and it does work. Its pros are: It does work, Debian Linux 64 bits runs there with no problems, its small form factor. Its cons, at least to me are: Coming from China to my country, some 45% of cost is shipping and duties. Although it is Atom based, it does have a couple of small fans which in such a reduced form factor makes the solution a bit noisy, which may not be important in some sites, but it may be a problem in others. Something that I do not know if it is a con or a pro, but that grabbed my attention, is that the chassis appears to be made of thick caliber steel. It would not surprise met hat it could stop a bullet. I also tried assembling a pc myself. All the needed parts are available in my town, exception made of a small good looking chassis and corresponding power supply.The closest chassis that I found was a slim mid tower unit. Not so cool looking, but the unit worked and was affordable. That partial success encouraged me to a next project. Import just the chassis and PSU and find everything else at home. The chassis that I will try next is the M350 http://www.mini-box.com/M350-universal-mini-itx-enclosure I will let you guys know. The chassis is still somewhere on the road.

 

Cheers!

 

 

I am thinking of building an m350 also, looks very promising with all the mounting options.

Link to comment
Share on other sites

Hi

 

Mine just arrived and looks exactly as I was anticipating. Sturdy, compact, well ventilated and yet something that could be "presentable". I will start assembly asap. One interesting thing is that it comes with a couple of USB ports that are hiding behind the front cap. This last feature makes me wonder a couple of things. First, if it is posible to make run an appliance without an internal hard disk, just with a USB Flash memory (in other words, if I could make a Linux installation there), and second, if it is technically possible, if it would be reliable enough. I mean, I know how prone to failure are hard disks, I have had my share of problems with them, but I do not have a feeling yet for usb flash memories, I have not used them that much.

 

Can someone (or snomeone) shed some light on this idea?

 

 

Regards,

Link to comment
Share on other sites

  • 3 weeks later...

Hi,

 

I just built the first unit using the 350 as the chassis, an Atom D2500 as the core. and a small SSD for storage. It is exactly what I was anticipating. Does the job, looks presentable, it is affordable and allows you to say the catchy phrase: All solid state, no moving parts.

 

Now then. Here at my hometown I could also get (for about the same price) an ECS NM-70 I2 mother board that has a Celeron dual core 847 processor (same form factor). It would be more speed everywhere (processor, RAM and even access to the hard drive). Being price about "equal" the pros and cons I see are the following: In favor of the Atom; 1) Three years warranty as opposed to only one, 2) Fan-less processor (one less thing that could go wrong) and in favor of the Celeron: 1) Raw speed (The processor scores 1042 against 407 on pass-mark benchmark, RAM runs at 1333 MHz as opposed to 1066 and Access to the SSD is at 6 Gb per second instead of only 3)

 

Being cost practically the same, it is only a thing of valuing reliability against speed. One way of thinking would be that since electronic parts tend to either fail pretty soon or last for years, perhaps I should favor the faster solution and stress test the units before sending them to a final user. The other way of thinking would be to use the Atom for users without plans for growth and the Celeron for the ones that are either large already or have plans for growth. The problem here is that I do not have an idea of how many concurrent calls would each system be able to handle.

 

I would appreciate your thoughts and insights.

 

Hi

 

Mine just arrived and looks exactly as I was anticipating. Sturdy, compact, well ventilated and yet something that could be "presentable". I will start assembly asap. One interesting thing is that it comes with a couple of USB ports that are hiding behind the front cap. This last feature makes me wonder a couple of things. First, if it is posible to make run an appliance without an internal hard disk, just with a USB Flash memory (in other words, if I could make a Linux installation there), and second, if it is technically possible, if it would be reliable enough. I mean, I know how prone to failure are hard disks, I have had my share of problems with them, but I do not have a feeling yet for usb flash memories, I have not used them that much.

 

Can someone (or snomeone) shed some light on this idea?

 

 

Regards,

Link to comment
Share on other sites

Hi,

 

I just built the first unit using the 350 as the chassis, an Atom D2500 as the core. and a small SSD for storage. It is exactly what I was anticipating. Does the job, looks presentable, it is affordable and allows you to say the catchy phrase: All solid state, no moving parts.

 

Now then. Here at my hometown I could also get (for about the same price) an ECS NM-70 I2 mother board that has a Celeron dual core 847 processor (same form factor). It would be more speed everywhere (processor, RAM and even access to the hard drive). Being price about "equal" the pros and cons I see are the following: In favor of the Atom; 1) Three years warranty as opposed to only one, 2) Fan-less processor (one less thing that could go wrong) and in favor of the Celeron: 1) Raw speed (The processor scores 1042 against 407 on pass-mark benchmark, RAM runs at 1333 MHz as opposed to 1066 and Access to the SSD is at 6 Gb per second instead of only 3)

 

Being cost practically the same, it is only a thing of valuing reliability against speed. One way of thinking would be that since electronic parts tend to either fail pretty soon or last for years, perhaps I should favor the faster solution and stress test the units before sending them to a final user. The other way of thinking would be to use the Atom for users without plans for growth and the Celeron for the ones that are either large already or have plans for growth. The problem here is that I do not have an idea of how many concurrent calls would each system be able to handle.

 

I would appreciate your thoughts and insights.

 

The Celeron 847 processor is only 1.1 ghz vs the atom's 1.87 ghz. I see the front side bus difference but what makes the cerleron faster? I am not being sarcastic, I really want to know.

Link to comment
Share on other sites

Steve,

 

You are right. I am not even sure if one is faster that the other. I am relying on a benchmark test where we do not even know exactly what kind of processes they put the chips to do so they can grade them. The key process that we are interested on in our application, would be the ones that allow a higher number of concurrent calls. That parameter would really be the one to use to judge them. With the information that I have at the moment, I would intuitively suspect that the Celeron would have more "throughput" thanks to its memory management capabilities despite of its slower clock frequency. Just as a big fan can move more air than a little one, despite the fact that the little one may have more RPMs, but I am guessing and again, relying on a benchmark test which may be scoring processes that may be irrelevant to our application (what it takes to be able to handle more concurrent calls).

 

I hope that someone who has had hands on experience with these two processors or who knows of their inner workings will give us more info.

 

Regards!

Link to comment
Share on other sites

  • 3 weeks later...

Hi,

 

After being quoting and installing around 20+ users systems, the first larger beast is rearing its head. I have just been asked to quote a system for 120 users. It will have to coexist with a digital E1 to interact with a carrier and with some 10 or so FXO lines, as well as be prepared to one day use SIP trunks. Reliability is the number one characteristic required and, of course, it should have enough processing power to take care of the 120 users generated traffic. Now I wonder, it makes sense to have redundant power supplies, redundant fans, but, how about redundant hard drives? Would that actually work? I mean, license wise I suppose that if the hard drive where the system is residing "dies", it would not help to have another one instantly attempting to do the job, it would lack the proper license, or not?

 

What would be the recommended way to address the issue of achieving the highest levels of reliability and resiliency?

 

I went for the ZOTAC ZBOX ID81 running CentOS with 64Gb SSD and 2Gb RAM.

Runs well with no issues.

Link to comment
Share on other sites

  • 2 weeks later...

Hi,

 

After being quoting and installing around 20+ users systems, the first larger beast is rearing its head. I have just been asked to quote a system for 120 users. It will have to coexist with a digital E1 to interact with a carrier and with some 10 or so FXO lines, as well as be prepared to one day use SIP trunks. Reliability is the number one characteristic required and, of course, it should have enough processing power to take care of the 120 users generated traffic. Now I wonder, it makes sense to have redundant power supplies, redundant fans, but, how about redundant hard drives? Would that actually work? I mean, license wise I suppose that if the hard drive where the system is residing "dies", it would not help to have another one instantly attempting to do the job, it would lack the proper license, or not?

 

What would be the recommended way to address the issue of achieving the highest levels of reliability and resiliency?

 

 

Sorry to throw out another question, as opposed to offering a concrete recommendation: In the scenario described by Carlos, does anyone view the cloud as a viable redundancy option?

Link to comment
Share on other sites

Sorry to throw out another question, as opposed to offering a concrete recommendation: In the scenario described by Carlos, does anyone view the cloud as a viable redundancy option?

 

I think that depends where the cloud is. If your system is in Boston, your backup server in the cloud should not be in Taiwan as this would cause an immense delay (might still be better than no service, though).

Link to comment
Share on other sites

I think that depends where the cloud is. If your system is in Boston, your backup server in the cloud should not be in Taiwan as this would cause an immense delay (might still be better than no service, though).

 

On that note, is there a way to measure how far is "too far?" If so, are we looking to have the cloud server w/in a certain # of miles; or below a certain ping time; etc.?

I ask b/c I recently stumbled upon a 'younger' cloud-hosting provider: DigitalOcean (aff. link). They're based out of New York, but also have data centers in San Francisco and Amsterdam. I'm located in Wisconsin, but I've been playing around with them b/c of their competitive rates (compared to Amazon, Linode, Rackspace, and others). Right now, I have a VPS w/1GB RAM and 30GB SSD for $10/mo. They also offer nodes @ other attractive price-points: $5/mo., $20/mo. (w/2GBs of RAM, 2 cores & 40GB SSD), $40/mo., and up.

At those rates, putting as many customers up in the cloud as possible might be great; so long as quality isn't noticeably reduced. Netcraft released a recent report that details the history of DigitalOcean's (non-aff. link) rapid growth in comparison to other cloud hosting providers.

Link to comment
Share on other sites

Hi,

 

To compound the problem, I believe that one should take into account not only the distance from the base system and the cloud hosted service, but also where the sip trunk provider is located. The majority of my customers are in the northeastern part of Mexico and the SIP trunk provider is in Central Mexico. It could work that the cloud be in Dallas, I believe it does, but one should run a pilot test because with IP traffic, the actual flow of packets not necessarily follows a straight line. Many moons ago, I negotiated a "peering" arrangement between the local incumbent carrier and a new competitive one. If one customer of the competitive carrier wanted to visit a web page hosted with the incumbent, the packets were to go first to the USA (where the incumbent carrier linked with a tier 1 Internet provider) and then back to the country, to the incumbent. After the peering arrangement, the packets followed a more logical path. So, things like "latency" (the time it takes to the packets to travel from server to server and back), variability of such latency (jitter) and packet loss, are the main parameters to watch. And to make things even more complicated, some measurements, like packet loss, can be tolerable with one codec, but not with another. This is because, proportionally, more data is lost when a high compression (and hence efficient) codec looses a packet. In short, one has to do tests. If it works, great! And here there is a characteristic that snom phones have and I feel that is has not been touted enough yet. At user agent level, the IP phone, has failover between "identities", so, one identity can be registered to the local pbx, and a failover identity registered to a PBX in the cloud. If no backups had ever been done, the users may not have access to their stored voice mail, directory and things like that, but they would be able to continue making and receiving calls.

 

Coming back to "how far is to far", latency is the keyword. The number of milliseconds that a packet takes to make a full round trip. In my short experience, below 50 milliseconds, you are going to be praised for the quality of the voice, from 50 to 150, most likely everybody is going to be just happy, above 150 you will probably start to have some complaints, first by the more discerning users and as latency approaches 250 ms by probably everybody. Above 250, probably registration will become a problem in itself.

 

 

On that note, is there a way to measure how far is "too far?" If so, are we looking to have the cloud server w/in a certain # of miles; or below a certain ping time; etc.?

 

I ask b/c I recently stumbled upon a 'younger' cloud-hosting provider: DigitalOcean (aff. link). They're based out of New York, but also have data centers in San Francisco and Amsterdam. I'm located in Wisconsin, but I've been playing around with them b/c of their competitive rates (compared to Amazon, Linode, Rackspace, and others). Right now, I have a VPS w/1GB RAM and 30GB SSD for $10/mo. They also offer nodes @ other attractive price-points: $5/mo., $20/mo. (w/2GBs of RAM, 2 cores & 40GB SSD), $40/mo., and up.

 

At those rates, putting as many customers up in the cloud as possible might be great; so long as quality isn't noticeably reduced. Netcraft released a recent report that details the history of DigitalOcean's (non-aff. link) rapid growth in comparison to other cloud hosting providers.

Link to comment
Share on other sites

This topic is a little bit off the mainstream topics, however I feel that it has some practical relevance for most of the customers who need to run the PBX on their premises.

 

...

 

I know this topic is far from complete; if you have more please feel to add other links (please no advertisements, though).

 

Does anyone have a reason to think that either of these might not hold up in a production environment; for, say, 5, 10 or 20-extension deployments?

 

The Raspberry Pi (Model B'): $35

- Broadcom BCM2835 700MHz ARM1176JZFS processor with FPU and Videocore 4 GPU

- 512MB RAM

- Boots from SD card, running a version of Linux

- 2 USB 2.0 sockets

- 10/100 BaseT Ethernet socket

BeagleBone Black: $45

- 1GHz AM3359 ARM Cortex-A8 processor

- 512MB DDR3 (800MHz x 16)

- 2GB on-board storage using eMMC

- microSD card slot for additional storage

- HS USB 2.0 Client Port, LS/FS/HS USB 2.0 Host Port

- Ethernet

- microHDMI

- 2x 46 pin headers

Link to comment
Share on other sites

We bought a Raspberry Pi recently... Yes that indeed looks very promising. The CPU will be good for 10 calls, no doubt about that. My primary concern is the connectors. The power supply using USB is very clever, however it can get out of the device very easily and then a 24/7 service will be offline... Also the way the connections are placed on the PCB are challenging for a good housing (3D printing). We thought about making a new PCB layout; however for the volume we are looking at it is probably not worth the effort.

 

The BeagleBoard seems to solve the connection problem in a better way it seems. IMHO it looks much more suitable for a CPE PBX. I just ordered one! To be continued.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...