What is our primary use case?
It's used in switches, servers, and routers. It's pretty fast. We also use it on an onboard airplane for Airforce One. Besides the fiber connection, it's very fast, with a speed of 25 gig. So it works well.
Broadcom's software is very easy to modify and configure, but the license to use it is expensive. Once you get past that, it's easy to modify. The temperature rating for the parts is also very good, so our thermal engineering is able to get some good thermal replay using the parts. The great thing about Broadcom is that it's very high-end, and now it's a very mature part.
What is most valuable?
The most valuable feature is the ability to update the firmware or software for new features. The good thing about the StrataDNX is that the Broadcom family is very interchangeable, so they work well with different optical transceivers.
It's almost plug-and-play, and it's still hard to develop with any of those high-speed chips, but relatively Broadcom provides a nice reference design that you can pretty much copy and paste into your design. Then you modify the software for different features you want to implement in them, like the PCP timing. It helps make it cheaper.
What needs improvement?
There is room for improvement in the pricing model. When I say "make it cheaper," what I'm really talking about is making the software cheaper. For example, Time-Sensitive Networking (TSN) has been developed on it, but it hasn't been implemented as much as it probably could be. It has good Denial of Service (DoS) and Zintag capabilities for IEEE 802.1br port extension. The quality of service, IEEE 802.1p, is good, but it probably could be developed more.
For how long have I used the solution?
I have been using this solution for two years,
What do I think about the stability of the solution?
If we do the design right, it's very stable. We've used these for a lot of filter applications over ten years. The number of products you have with these designs is that the fourth switch probably goes out of stock over time versus the chip. So it's very reliable.
What do I think about the scalability of the solution?
Scalability is good. It's part of it. So the software is like a hardware extraction layer with the software. If you move to a different status part, it's pretty easy. It's almost plug-and-play. You can pop it into your design or buy a new white box and use most of the software in the new WiFi design.
How are customer service and support?
Customer service support depends on money. If you're a white box manufacturer and we're selling thousands of these, you're going to get awesome support, a ten, amazing support. But if you're starting and need only one or two chips, the support is just access to online stuff or whatever's been done. The better support comes with more units bought.
Initially, I would probably rate the support a five out of ten. But then we paid more money, and suddenly, it became a ten, awesome. We had people showing up in the office and walking us through it. But there are also different support models. You can work with them directly or with distributors like Avnet or Arrow, who can provide support. And then there's higher level support with the team in Los Angeles who can provide even more support. It just depends on how many units you're going to buy. The more you buy, the more support you get in the queue.
How would you rate customer service and support?
How was the initial setup?
The initial setup is complex, but if you follow the recommended design, it's not so bad. The hardest part is when people divert from the design and try to do things that it wasn't meant to do. If you try to do that, then it starts to fail. But Broadcom's team that supports it is really good. To be fair, I used to work for AppNet on the Broadcom team to support this part. Our team of support was actually pretty quick and good. It really depends on the two types of people - those who design with the chip and spend the hundred thousand on the license to use the software, and those who buy the white box with the chip in it and then work with them to tweak it. That's kind of the two versions of developers that will do this.
So if you buy the white box, the design is already done, and you're just tweaking the upper-level software on it. But it's a bit more challenging if it's a little white box manufacturer that's actually implemented. Some people think they can get it done in a year if they've never used the Broadcom part, but generally, it takes about two years and a lot of money to make it work. And then when people make that Livebox, it's much easier. If you buy the white box with somebody, then it's already done, and then it's just a matter of working with them to figure out what type of settings you want - PSN or other features. Broadcom sells the software kind of like going to Burger King. How do you want it? Oh, I want this feature, I want this feature, I want this feature. Okay. Cost as much. It's all there because you're on.
What about the implementation team?
For the deployment process, initially, we started with five people. We tried to do the base design, including the electrical design, and then we started on the software. But that was really hard, so we scrapped that and bought a white box design. We were able to get it deployed with the same five people.
It may seem simple, but it's actually much more complex if you try to design it yourself. Many people use their reference design and software, and it can be brought up pretty quickly. But it becomes more complex if you want to make changes and customize it. So the faster way out the door is to buy from companies like Arista, Seldom, or Equifax, who sell pre-designed boxes. If you want to control the design yourself, it could take two to three years.
What's my experience with pricing, setup cost, and licensing?
I would rate the pricing model a five out of ten. It is somewhere in the middle. They're pretty good. The tricky thing is there aren't many choices anymore. It used to be that you could buy Marvell or Qualcomm, but I think Broadcom bought Marvell. Now it's down to Cisco and Broadcom, and there aren't many other options. Cisco is super expensive, while Broadcom is much more affordable and better.
What other advice do I have?
Overall, I would rate the solution a nine out of ten. I would definitely recommend using it for design.