I like reading Frank's news and so will post all of his installments on this thread. To get us talking about the right things....
Frank's News « back to Frank's News
Fry’s and fiber optics
April 15, 2000
by Frank Levinson
Here in Silicon Valley we have a unique shopping experience – Frys Electronics. The place was started by the sons of one of the prominent Northern California supermarket owners.Fry’s is a place where you can buy Sony Walkman and Playstations, computers and printers, software and test equipment, even surface mount resistors, and a quick meal or stop at the expresso bar.
One time a customer came to see Finisar, and after dinner he specifically requested to go to Frys.He wanted to purchase different color electrical cable tie wraps.No kidding.He got 100 each of the purple, florescent green and even a pink set (for the wife or girl friend, I suppose). He was estatic.
So if you come visit and want to shop a Frys … what can you get? They now have (for more than a year!) fiber optic components. No other proof is required that fiber optics is ready for home deployment. Just go to Aisle 8, you can find ST connectorized jumpers, network hubs with fiber extension ports on Aisle 9, patch panels that support mixed 10baseT and duplex SC on Aisle 14.The place is much more like a Safeway or Krogers than Circuit City.
One rule that stands out at Frys is that an impulse buy is certainly less than $500 with it closer to $300 as the upper limit. So adapter cards for PCs are only available below this level. No high-end raid controllers here. So if fiber optics is now at Frys where is that taking us? Does it mean that we can expect to see fiber optic Gigabit Ethernet or Fibre Channel adapters there sometime soon? In fact, it has already happened.On March 12, 1999 Fry’s carried its first Gigabit Ethernet host bus adapter (HBA) card.
Let’s run the numbers!
Since adapters need to cost approximately $300 to be a solid impulse buy, we will use this as our starting place.Now, the manufacturer of the adapter needs to be able to build it for about 2/3 of that figure in order to make enough money for both a little profit and future R&D.What’s on that card? Well, a fiber optic transceiver, of course. Being fiber optic engineers, we are glad about that … real volume if you can sell at Frys … but … oh, yes there is a SERDES and probably a CMOS chip and then some RAM … and then the printed circuit board. Did you forget about the software on that accompanying diskette or CD and what about the snazzy box with colorful graphics and the instruction manual and the hot line to answer problems.
All of that stuff needs to cost less than $200. So how much do we get for the fiber optic part? The answer is probably not more than $75 and less is better.
Most adapter card companies do not make fiber optic transceivers. Ahhh, now comes the higher math. The transceiver makers also have to do the same dance that the adapter card vendor just went through. And the answer is ….
That $75 transceiver needs to be made for $40 and that means that the laser, photodiode, 2 connector structures, receiver preamp, receiver postamp, signal detect logic, transmitter driver, electrical connector pins … all must cost less than that.This part needs to be build in large volume but it will be assumed to have the same reliability as the chips on the board which means that it will be more reliable than submarine cable parts of only a decade ago.
This is where the field of fiber optic components for data communications is going … to the sales floor of Frys Electronics.
Ready?
Finisar is.
......
Optical Component Margins - Response to Charlie Willhoit
May 15, 2000
by Frank Levinson
Telecom vrs Datacom
In a recent column, analyst Charlie Willhoit discussed what should be appropriate gross margins for optical component manufacturers. He noted that high margins were costly to the system manufacturers and how this might be inhibiting some markets that would otherwise accelerate.
The actual way margins decline in optical components is more dynamic than envisioned by this analyst. Finisar has participated in one of the most dramatic price declines and market empowerments in the last 10 years … so perhaps we can shed some light on what really causes margin declines in this space.
The “poster child” for the new wave of optics is JDS Uniphase. Through acquisitions this company now dominates the landscape that just a couple of years ago any analyst would have thought could only belong to Lucent, Nortel, Alcatel, Siemens or Sumitomo. They have put together an array of products under one corporate banner to supply the needs of nearly all telecom fiber optic components – lasers, transmitters, photodiodes, receivers, couplers, switches, modulators, WDM devices, pump lasers and EDFAs.
But the actual market where the largest long-term growth can occur does not have to be telecom fiber optic components. It could be driven by datacom and when this happens (note the lack of the term “if”) the price erosion will not be measured in single digit percentages but rather in terms of factors of 10x or 100x!
The reason for this lies in the fact that datacom drives unit levels of sales that are so very much greater than telecom. This higher level of sales empowers the manufacturers to employ much greater levels of automation, cost reduction engineering, and device integration. The combination of these tools drastically reduces costs and prices.
An example of this is a transceiver module operating at 2.5 Gb/s over single mode fiber. From a telecom perspective, this device is sold in the $5-10K range. It has a laser-based transmitter, PIN-preamp, a clock data recovery IC, then a set of multiplexer chips and finally a set of SONET/ATM chips that manipulate the line data for the local terminal.
A counter example to this part would be an adapter card for 2x Fibre Channel. This year this card will sell for about $500-1000. It can be configured with a laser-based (singlemode) transmitter, a PIN-preamp, a SERDES which performs clock data recovery and all serial to parallel conversions including framing and an interface IC to take the data and present it to the host computer or storage array. A fully equivalent part with fully equivalent features for ~10x lower price.
The integration in the IC area of this adapter card is vastly higher than the equivalent SONET parts (a single IC is possible compared to a board full of ICs) and the size of the fiber optic transceiver and its cost effectiveness is similar to the ICs.
A History of Telecom vrs Datacom
In 1991, Finisar introduced a 1 Gb/s optical transceiver that sold for $600 quantity 1. The specifications of that part together with work done by IBM was offered to the Fibre Channel standards committee as the basis for a new standard at this bit rate. Within 9 months this new standard was unanimously adopted.
At that time similar bit telecom parts sold for more than $6000 (10x again) and had been in the market for at least 5-7 years. These telecom parts were already selling in reasonable volumes. By the time that the datacom parts were actually selling in quantities similar to the telecom parts (around 1995) the telecom parts had been in the market for approximately 10 years. Today the datacom parts sell in quantities that are more than 10x that of telecom parts; perhaps even higher.
Now let’s look at the same case for 2.5 Gb/s and 10 Gb/s. CY 2000 will be the year that datacom 2.5 Gb/s parts sell in volume substantially greater than their telecom equivalent. CY2002 is likely to be the year when 10 Gb/s datacom parts sell in volume substantially greather than their telecom equivalent. What is happening is that the time gap between telecom and datacom technology is decreasing and is likely to cross over around 2005 or so!
When this happens, all of the arguments about margins, margin pressure, network construction, etc. will stop. The lowest cost possible will be achieved rather broadly across fiber optic markets. The only untouched market for higher cost products will be long haul telecom – transmission equipment used in distances over 40 km. This leaves substantially all of the Metropolitan Area Networks (MANs), Storage Area Networks (SANs) and Local Area Networks (LANs) to be dominated by datacom economics.
An even bolder view of this trend shows that it is likely that long haul telecom will never deploy super high bit rate systems such as 160 Gb/s but that datacom will do so because the fiber that limits technology deployment in long haul will not do so for the short haul! Today the papers at OFC (the premier fiber optic conference) that report the “hero” system experiments are all from the telecom side; in a few years it is possible now to predict that most of these may be from the datacom side as these opportunities will have the most esoteric and interesting technical innovations.
JDS Uniphase Compared to Finisar
Today, JDS Uniphase, the darling of independent telecom optical component suppliers has sales approximately 10x that of Finisar, no profits and market capitalization approximately 30x that of Finisar (a leading datacom optical component supplier). What is interesting to note is that the 2 companies sell approximately the same number of units!
What accounts for this similarity in unit sales?
JDS Uniphase average sell price for each unit sold is approximately $1000. Note that this is the number quoted by Willhoit for their lasers, pumps are more, couplers and isolators a little less.
Finisar’s average sell price is around $100 per unit sold.
This is the reality of datacom – larger volumes but smaller ASPs. What this enforces to the companies that choose to embrace this market is a discipline that should carry them forward as datacom economics accelerates it technical penetration into more and more markets traditionally reserved for telecom.
One additional interesting comparision. Finisar’s annual sales growth is just below 100% per year and has been this way for more than 6 years. Finisar’s annual unit growth is even higher than this as prices are falling while sales continue to double each year. Others in this same space are seeing similar growth.
So, while JDS Uniphase sees growth numbers of 30-40% per year in units and sales. The datacom companies see unit growth of 100-200% per year and sales growth of 100% per year! This remarkable growth rate makes the collision of these historically separate markets inevitable and the exponential growth rate of each should make it most interesting.
The chart in Figure 1 tries to encapsulate this coming collision. It shows that at10 Gb/s technology will likely reach the sweet spot of the sales curve for datacom parts about 1 year after the telecom parts reach the same sales point.
Figure 1. Relative market windows for 1, 2.5 and 10 Gb/s technology both for telecom (green) markets and datacom (red) markets.
What this figure tries to represent is that for 1 Gb/s fiber optic systems datacom lagged telecom by up to 8 years. When the deployment of 2.5 Gb/s systems occurred this lag was only 4 years. Now we are moving towards the sweet spot of 10 Gb/s and in this case the telecom and datacom markets may mature only 1 year out of step! In fact, it is possible that as 10 Gb/s is the adopted standard for Metropolitan area networks and as these networks primarily are deployed using Ethernet technology instead of SONET then the datacom version of 10 Gb/s data links will be the dominate market version.
So, a final word about margins…
As this revolution continues therefore it is the datacom economics that will come to the forefront. The issue of margins raised by Charlie Willhoit and other analysts needs to be recast in terms of markets and appropriate technologies. Then the nature of the products that serve these market and their natural or native margins will be clear.
.......
George Gilder, All Optical Networks, And Finisar
June 15, 2000
by Frank Levinson
George Gilder is a widely published columnist and pundit. His thoughts about man’s technological future are widely respected. Lately, he has been evangelizing a vision of an all-optical Internet, plentiful bandwidth and the demise of many traditional Internet technologies like Cisco routers. Some of this vision is underpinned with assumptions that are not applicable and as clarity in assumptions is critical for any future vision, I’d like to use this forum to work through these.
First of all let me quote Gilder directly so that my jumble of opinions do not dilute his message:
“… the Internet is a computer on a planet. Like a computer on a motherboard, it faces sever problems of memory access. The Internet communications depend on ingenious hierarchical memory management, analogous to a computer’s registers, buffers and latches, its three tiers of speculative caches, its bulk troves of archives, its garbage management systems to filter and weed out redundant or dated data, and it direct-memory access controllers to bypass congested nodes.
In a world of bandwidth abundance, an ever-increasing share of roundtrip delay for a message is attributable to speed of light latency. No matter how capacious the transmission pipes, how large in numbers of bits per second the data stream, the first bit in the message cannot move from source to terminal any faster than light speed allows, plus the time waiting in queues and buffers at all the switches or other nodes along the way.”
He then makes the case for strong caching and dispersed assets … here are some of the supporting assertions:
“Even with no hops or other delays, the light-speed limit alone means that Internet users outside the North American continent are at lest 200 milliseconds away from the vast majority of websites.” (80% of web hosts are in continental USA)
“To fetch a web object using the Internet protocol - whether a frame, image, logo, or banner - takes two to seven round trips between the end user and the Web server. With each page comprising as many as twenty-five objects, those round-trip-speed-of-light milliseconds keep adding up even for entirely static material.”
Perhaps a little table will help us keep track of what physics foundation Gilder is operating from.
Description
value
units
speed of light
300000
km/sec
diameter of earth
12500
kilometers
circumference of earth
40000
kilometers
max intercontinental hop
30000
kilometers
round trip time
200
milliseconds
typical intercontinental hop
12000
kilometers
round trip time
80
milliseconds
typical US only hop
2500
kilometers
round trip time
17
milliseconds
So, his assertion of 200 ms for the worst case hop just based on speed of light arguments is well justified. The figure hinges on the following underlying assumptions worst-case distance is
• 50% of earth’s circumferenc
• 50% longer than even this due to routing inefficiencies
• really 2x this because a web page fetch involves a request and an answer so the distance is traversed twice.
This worst case figure is really exaggerated for the point of his article … a more typical intercontinental distance is likely to be 12000 km and then the time drops to 80 ms. And for normal intra-continental distances of 3000 km this is then 17 ms. Actually most traffic will be of this latter category since each citizen/user will most of the time be fetching information localized for their country.
But then his assertion that this is the root of all our problems does not wash. He assumes that web pages will always be transmitted as individually requested objects on a page. This assumption is weak … a much simpler solution is for HTTP and its cousins to migrate towards a single transmission request for all information and for the sender to gather together the information and send it just one time as a single group or stream. No doubt this is better for the server and the Internet infrastructure to see a larger single flow instead of so many little (25) pieces. Again, it is the multiplication of so many separate requests to assemble a single web page that causes the problem.
In a sense, such progress assumes that Internet protocols will continue to evolve … sounds reasonable. Moreover, once the streaming has started the end user will not notice the effects of distance provided there is sufficient bandwidth between the sites and buffer at the receive end to smooth out any potholes in the transmission
Which is the better approach?
Reasonably local/short distance widely dispersed caching or more aggregated page serving?
Let’s use Gilder’s assumption of 25 objects per web page for our starting place. We shall assume the cache resides at an effective distance of 250 km for a round trip time of 1.7 ms and therefore 25 objects will take 43 ms. This is substantially slower than the direct sending of the page over typical intra-continental distances of 2500 km but sent as a single object (17 ms).
The point of this is that direct bundled sending is substantially better than local caching of pages like Akamai or Inktomi can provide.
As our agents and web servers get smarter, we will find that their ability to coordinate the information flow to us so as to minimize our waits and keep their responses more in line with our normal cadence of work will remain acceptable.
So where is the problem of Internet scaling? It is primarily in the core of the network where the global computer must continue to scale to packet per second forwarding levels measured in terms of 1012 and then 1015 packet per second over the next few years. Today the biggest, baddest layer 3 switches or routers have ability to do about 108 packets per second. Collections of these cannot do any better than this because they are ultimately placed into chains and the chain is only as good as its weakest link. The problem with Internet scaling is not with typical web pages or the migration of television or even telephones to the web; it is with distributed computing. As we seek to create larger clusters of computing, especially as clusters are formed and broken down dynamically as their user bases require this distance and light travel time strongly affect the ability to do cluster computing.
Interestingly enough the problem in trying to create a world wide computer system (like that envisioned in the movie, The Matrix) is that the messages will only propagate at the speed of light and the network in between cannot beat this limitation. Responses that require inputs or outputs from disparate points in the system will not respond in computer time but in earth time … milliseconds not nanoseconds.
.........
Frank's News « back to Frank's News
How Finisar Got Its Name!
August 15, 2000
by Frank Levinson
I have been asked many times about the origin of the name Finisar. Well as with many corporate names ultimately it came because it passed the test of time and was found not to mean anything bad in any known language on Earth.
So just how much truth can you stand? Remember, from this point on, reading is at your own peril.
One of the problems with large companies today is they learned early on that engineers are best when contributing in there area of expertise. Thus when a project is defined, it is quickly broken down into the smallest pieces possible and experts are assigned to each piece. This results in a number of bad side effects, but here are a few—
• System problems often have solutions that end up overly specified towards the primary discipline of the system designer.
• In other cases, thoroughbred horses are specified and camels are delivered as pieces are well designed but the fitting into an excellent system falters.
• Engineers do not keep current or are not allowed to follow problems far enough to learn their fields more broadly.
• System trade offs between among the pieces of engineering are not done and simplification is not achieved.
So, the truth is that since starting my engineering career in fiber optics at Bell Labs in 1980, I was never allowed to finish things. Consequently, Finisar was named after my desire to see projects from start to finish.
Ah, but what about the -ar portion of the name? Now, before 1980 I was in graduate school for a PhD in Astronomy. And at that time pulsars and quasars were the rage so the –ar was taken from these exotic astronomy objects and the root of finish was taken for the base name.
There was a competing name – Refinet. For refined networking. But after having a few people try to pronounce it, a distinctly French interpretation emerged. (ref – in – nea) Sounded like a cheap wine with aspirations to have come from the Napa valley but actually grew up in Pittsburgh.
So, do we let engineers finish project here at Finisar? You bet. There are times when our engineers think that their project may never end but overall we relish the idea of doing even complex projects with distinctly small teams. Where the message to the team is - this is your project, others can be called into help but you have the responsibility, you have the freedom. And you must finish it!
Finisar has NO engineering budgets. Engineers who have worked for other larger companies come here and always ask - what is my budget? To which we answer, we want to discuss the time and complexity of the project you are doing. Then either we approve the schedule and functional specification or not. If it is approved, there is no budget, just be on time.
In this context, we encourage our engineers to:
• Purchase the best equipment, do not scrimp.
• Tell us how to staff the project; what other skills are required for success?
• Find ways to move the project faster.
• Always do rapid board or IC turns, fast prototyping, etc.
The point is that finishing projects is where real satisfaction occurs. If you only get to see a piece, or do only a part of the project, you never see the wider impact of your ideas. Finisar engineers usually talk to customers after products ship, are even involved in training so that they can directly get customer feedback.
It turns out that Finisar was the second company that I was able to name. The first company was called Netek. Obvious what was the focus for that company.
Netek was started with the help of Bruce Elmblad (a founder of Prime computer) and AMP back when they also started Lytel. Netek began with mucho bucks (for that day and time), many people, plenty of nice space …
… and only had problems. First of all it had an absolutely terrible CEO (me!) and then we tried to define projects from our own experience without customer feedback. And then we hired a CFO who thought he knew fiber optics and that was truly dangerous! All in all, the result was that the CEO of AMP flew into Morristown, NJ on Jan 31, 1985 and fired me. He then got back into his corporate plane and flew home.
I, on the other hand, got back into my well-worn Chevy station wagon and slugged it home through the largest snowstorm of that year to my wife who was pregnant with our third child. Talk about different perspectives.
It all turned out OK. We sued AMP, they realized that the CFO was the bad egg who had set me up and I was hired by Jerry Rawls (Finisar’s CEO) at the same salary that I had at Netek. Now the story is much more twisty turny and it has other elements that are part of Finisar but that is for another time.
In start contrast, at Finisar, we started out by renting a small Quonset hut, brought in a computer and furniture from our home, purchased a small copier and some very used test equipment. Then we began immediately enlisting customers and listening to their needs. We were able to start by doing contract-engineering work for these early customers from the day we opened our doors and so were immediately profitable and grew slowly as our means enabled us.
That is by far the best way to start a company. Take your time, use your own money and work for others from the beginning. As you do this, you develop a business culture that is always customer focused and always thinking about the top and bottom lines.
And best of all, you and all of the employees have a shot at owning so much more of the company. Today, Finisar is still majority owned by the people who work for us on a daily basis. And that allows us to all work hard knowing that it is we ourselves who are will be benefited most.
Aside: I have been writing these pieces now for about 6 months. Some have sparked email back and this is just great. Please write (flevinson@finisar.com) if you have comments or questions. And especially if you have a subject you want me to write about.
« back to Frank's News
WHy should anyone listen to Frank.....
Dr. Levinson is a prominent physicist with more than 15 years of experience and 20 patents in the areas of fiber optics, lasers, networks, and telephony. Read his column.
I think I will buy some stock in Franks company, 'cause I like his essays? |
