The demand for ultra-low latency (ULL) networks is booming these days as electronic trading has made speed advantages of mere microseconds a competitive differentiator for financial services firms. Ciena is targeting the Electronic Trading market with two new ultra-low latency cards now available for the CN 4200 – part of an expanded ULL portfolio announced today.

 

The 2RS 1GbE and Quad 10-T 10GbE cards on the CN 4200 are specifically designed to squeeze out every possible microsecond of unneeded and undesired latency (a fancy term for network delay) from the optical transport portion of the network, while still providing the high availability and performance monitoring that is critical to the financial industry.

 

That second part of the equation is key, as the race to reduce latency has led some optical vendors to strip out quality of service capabilities such as performance monitoring features from their ULL offerings.  Dimple Amin, Ciena’s VP of Engineering for Metro Transport & Enterprise, says that is a path Ciena has avoided.

 

“Attaining ultra low latency while not sacrificing critical performance monitoring is added value to our customers,” said Amin. “Ciena’s solution uses custom high performance electro-optics that minimizes the latency in the network.” 

 

Ciena’s announcement is just the latest piece of news in a busy year for the topic of ultra-low latency networks.  As I blogged about earlier this year, industry coverage of low latency networks has picked up as providers like Intellifiber and Spread Networks have announced new low latency network routes and offerings.

 

In fact, today at 11:00am ET, Spread Networks Senior VP Brennan Carley and Tabb Group Director of Research Adam Sussman will join Ciena’s Malcolm Loro for a webinar on Ultra-Low Latency Networks & Financial Trading.  The webinar is free to attend, and you can register here.

Thanks to a tweet from Andrew Schmitt of Infonetics, I came across the below chart this morning and thought it was a nice visual on how times have changed in the optical market.  The chart is vintage 2000 from Lightwave Online, and shows the market share of the optical transport market at the time.  Back then it would have been unthinkable that upstart Ciena (which was founded in 1992 and had only just gone public in 1997) would end up acquiring the optical business of 100 year old Nortel.*

 

The mark-up on the chart is courtesy of the Banana Stew blog that posted this bit of optical history.

**Nortel ® is a trademark of Nortel and is used by Ciena under license from Nortel.

One of the topics I’ve been covering on this blog has been the dramatic rise in interest by the industry for low latency optical networks.  What used to be a boring topic that only the engineers were interested in now has the attention of the entire optical industry.  Much of this interest is focused on supporting the needs of the financial services market, where automated trading has made delays of only microseconds the difference between making a trade and having your competitor take the business.

 

One service provider working to take advantage of this new focus on ultra-low latency networks is Spread Networks, who in June launched a new high-speed network connecting the financial hubs of New York and Chicago.

 

On September 1^st, Spread Networks Senior Vice President Brennan Carley will be a featured speaker for a Ciena online webinar focused on Ultra-Low-Latency Network and Financial Trading.  I was able to ask Brennan (who has a nice blog of his own here) a few questions about this growing market in advance of the webcast.  Below is our conversation:

 

=============================

 

Q: Electronic trading and the drive to reduce latency has really started to get attention this year with the media.  Is this a new requirement from financial services firms or has it been growing for some time?

 A: The industry has been moving to electronic trading for a long time, and latency has always mattered.  What has really changed is the units in which latency is measured. As long ago as the 1800’s the introduction of the telegraph replaced the messenger boys who used to run between the trading floors and brokers' offices, so latency was an issue then.  But it was measured in minutes (or days if you consider orders delivered from other cities.)  Now we measure it in microseconds, but it has always been true that speed conveyed a competitive advantage to those who invested in the latest technology. 

 

Q: The focus for low latency in the financial services market is around high-frequency trading.  What exactly is high-frequency trading and why is low latency so important? 

A: High Frequency Trading gets a lot of attention, but really there are a lot of different strategies that get lumped together under HFT.  Low latency is important for a simple reason:  Prices are continuously fluctuating, and prices in one place (e.g. the futures markets in Chicago) influence the direction of prices in other markets (e.g. the cash markets in NY.)  If you can get data back and forth between Chicago and NY before prices have moved, that has value.  To some traders it is a valuable arbitrage opportunity.  To others it is valuable to their risk management or hedging strategy (i.e. you can put a hedge in place before the market moves against you.) 

 

Q: Outside of the performance of optical equipment from vendors like Ciena, how else do you eliminate latency from the network?

A: You have to start by re-thinking the basic design of your network.  Most networks are built to scale to a large number of end-points.  But the network designs that are good for scaling (e.g. implementing hierarchical network topologies) add latency.  So you start by stripping your network down to the fundamentals, which is really about getting data fast from A to B.  You then need to peel away the layers that the carriers add (switches, muxes, etc), which are all put there to enable the carriers to serve many customers, sell many services, and aggregate lots of traffic onto shared backbones.  Those things are great for carriers, but add latency you don’t need.  So now you are down to the raw dark fiber itself, and you need to select a network that has the right geographic routing, the fastest fiber, optimal placement of optical amplification and regeneration facilities, and minimum excess slack (mileage).  Then light it with the right equipment!

 

Q: I read in an article about your network that you literally blasted through mountains in order to take the straightest path between New York and Chicago.  How much was involved in determining the best path and then getting all the access rights to lay the fiber? 

A: Spread Networks spent close to 3 years determining the optimal routing, securing the rights of way, and then constructing the network.  It was a huge task, mostly because nobody had ever done it before.  The carriers hadn’t done it, and it was new to the towns and counties Spread Networks built through.  So it was a big task.

 

Q: The financial services market is obviously your primary target customer, but are there other opportunities with different kinds of customers?

A: Absolutely.  While they aren’t latency sensitive, there are many customers in other industries that value access to a state of the art broadband network between NY and Chicago.  Until we built the network, there was no generally available dark fiber between NY and Chicago.  We think non-financial customers will find that very valuable for applications including data center mirroring, telepresence, and many more.  We also have an affiliated company, NorthEastern ITS, that offers service to the communities that our network passes through.  Many of these are rural communities where local libraries, hospitals, and schools will benefit tremendously from high quality broadband services.

 

Q: Looking into the future of this market, how much faster can you go?  

A: From a pure latency perspective I think we have delivered about as fast a network as can be built, which was our goal.  Our current customers are using a combination of 1G and 10G wavelengths, but I can see that evolving to 40G and even 100G, which won’t reduce latency but will increase capacity.  Beyond that there are some technologies that are in research labs that could reduce the latency of the fiber itself, but I think those are at least 3-4 years out.  The good news is that we designed our network so that when new fiber comes to the market, we can easily deploy it into the field and our customers will benefit from that.

 

============================

 

To hear more from Brennan Carley, as well as Ciena’s Malcolm Loro and Adam Sussman from TABB Group,  be sure to attend the Ciena webinar on September 1^st -- Ultra-Low-Latency Networks and Financial Trading: Fast Networks for Fast Markets

Several years ago I had Verizon’s FiOS service installed in my house, and signed up for their triple play bundle (TV, Internet, phone).  At the time, I would have been considered an early adopter as the fiber-to-the-home (FTTH) market was really just taking off.  But a slew of new reports out over the last few weeks show that FTTH is not only growing, it’s going mainstream.

 

One of the first steps in selling FTTH is people understanding what it is.  In a recent survey by the FTTH Council, 41% of broadband subscribers recognized the term “fiber to the home.”  That’s a huge jump from the previous 28% from a similar survey just a year earlier.

 

Actual FTTH deployments are popping up everywhere.  A recent report on the market pegged global FTTH subscribers at 41 million by the end of 2009, with a huge growth estimate to 306 million by 2014.  In fact, the UAE will go all fiber by 2011, and even the residents of a small village of 60 homes in the UK have decided to build their own FTTH network. 

 

As video becomes an increasing driver for bandwidth, these FTTH home deployments are increasingly focusing on even higher bandwidth speeds. A new Infonetics report says that 10G FTTH technologies are increasingly being tested by service providers around the world (see graph).  

 

And as residential customers adopt FTTH, their high satisfaction levels for the service indicate that they are unlikely to switch back.  The same FTTH Council survey found that 71% of FTTH subscribers said they were “very satisfied” with their Internet service, almost 20 points higher than competitive services.

Late last week, UK communications regulator Ofcom issued its annual Communications Market report covering the state of the UK communications market.  The report found that while Britons continued to increase their consumption of telecoms services (both fixed and mobile), total operator revenues derived from these services declined year-over-year for the first time ever in 2009, falling 2.6% (see Ofcom’s release, or the full PDF report).

 

Some of the most striking figures in the new report relate to the huge growth in mobile data networks, as Ofcom says that 2009 saw a 240% increase in data traffic over mobile networks.  This growth was driven by an 81% spike in smartphone ownership over the last year.  As of May 2010, 12.8 million people in the UK (26.5%) said they owned a smartphone. These mobile users are also increasingly using 3G connectivity, with nearly one in three mobile connections using 3G technology by the end of 2009.  That’s a 39% increase over the prior year.

 

In spite of this jump in mobile use, mobile services reversed their traditional revenue growth trend (12% a year over the last decade), declining for the first time ever in 2009 by a total of 3.5%.  Figure 5.2 below from the Ofcom report shows the details the trend of telecom revenues by category.

 

And while the Ofcom report shows that total mobile data revenues have increased by 90% over the last two years (more than offset by a 5.3% drop in mobile voice and texting revenues), data use over mobile networks has grown an astounding 2200%, as figure 5.6 from the Ofcon report shows below.

 

Ofcom’s report suggests that the global economy is likely playing a factor in the 2009 dip, with mobile phone spending being the most likely communications service for Britons to reduce spending on.  But while an economic rebound may reignite mobile revenues, it’s obvious that those revenues will not keep pace with mobile bandwidth demand.

 

All of this relates back to my post from a few weeks ago asking why service providers aren’t going broke.  The key data point not detailed in the Ofcom report is the fact that the cost of bandwidth for the service provider and mobile operator continues to fall at a steady pace.  This constantly declining cost-per-bit is what enables UK operators to meet the demanding bandwidth requirements of their customers while keeping their pricing models relatively stable.  Only time will tell how long that model can last.

Unlimited bandwidth.  Everybody would love to have it, but nobody really wants to pay for it.  Even with the constantly declining cost of bandwidth that I’ve talked about previously, the surging amounts of bandwidth required by today’s enterprises can still be cost prohibitive.  

So what’s an Enterprise to do?  One possible solution is to dynamically increase (or decrease) the amount of bandwidth you pay for as demand changes.  In other words, bandwidth on-demand.  But can today’s carrier provide a true bandwidth on-demand service?

To answer that question, I’d like to introduce a new guest blogger.  Jim Morin is Ciena’s Product Line Director in the Managed Services & Enterprise Group and has been blogging on Ciena’s Partner Community for several years now.  Below is Jim’s guest post:

---

How will bandwidth on-demand work? This is the 64 million dollar question – how to make bandwidth dynamic?  (Kudos to anyone who remembers where the phrase “$64M question” originated.   Hint, it started out as the $64 question then.)

This is a complex endeavor for a carrier on many different levels and has been sort of a “holy grail” in the industry for a long time.  We know that change will come incrementally and it will take a while for a truly end-to-end fully dynamic bandwidth environment.  But it is possible to do some of this today, and progress is being made for greater flexibility in the future.


Today’s bandwidth on demand

With intelligent platforms like the CN 2000 and CN 4200 ESOM/EM6 users have the ability to dynamically increase/decrease a client port’s percentage use of a pre-provisioned network connection.  For example, you may want a storage array on port 1 to change from 10% allocation of the bandwidth, to 90% of the bandwidth, so the array’s replication job gets done faster during a peak load period like end-of-month processing. This is a simple dynamic change to achieve an on-demand increase of allocated bandwidth, and works for any WAN – SONET, IP, Carrier Ethernet.

The above example did not add any net new capacity, so what we also need is the ability to turn on additional capacity too.  Certainly this is easy to do by adding more wavelengths to a DWDM to increase the virtual throughput over the fiber.  If the equipment is pre-provisioned, this is a simple software change, either in a private optical network or a carrier delivered wavelength service.  Carrier Ethernet delivered services work in the same manner – where network capacity could start at 100 Mbps and be somewhat dynamically increased up to 1 Gbps or beyond.  But generally neither of these would be on-demand or easily scale bandwidth both ways – up or down.

A lot of the “thin provisioning” models in storage and elsewhere work on this same concept to dynamically expand capacity on pre-provisioned equipment.

Future bandwidth on demand

True on-demand is bandwidth where you need it, when you need it.  It assumes the physical infrastructure is already installed to support this, through the edge, metro and core to the same on the other side.  It might also assume the ability to transverse more than one carrier, and sort out all the provisioning and billing.  This is a very tall order!

So the first steps will be to break down this problem by offering on-demand services between two known locations where pre-provisioning equipment will make sense because of the expected business volumes.  One example might be between mega data centers, where it is a good guess that high bandwidth workload balancing will occur, even though you don’t know exactly when.

Ciena is working on ideas for the software glue to enable the ability to dynamically add, remove or resize bandwidth under software control.  The idea here is that applications like VMware’s vCenter may receive a request for a virtual machine (VM) and storage migration, and Ciena software we’re calling Translation, Arbitration & Control (TAC), can take that request and interface with the Ciena equipment from the edge through the carrier’s infrastructure to make it happen.  Automation and policy enforcement will be important, as well as managing Layer 2 domain extension.

The software interfaces between the virtual environment and the bandwidth infrastructure will define this capability more than the bandwidth type, e.g. wavelength, SONET, IP, Carrier Ethernet.

Benefits from bandwidth on demand

One area where bandwidth on-demand can have a large impact is the leasing of bandwidth between data centers.  Today, service providers and enterprises likely lease required bandwidth from wholesale carriers on a fixed price, 3 or 5-year contract, for a fixed capacity, typically sized around 1.5 times peak load.  This results in paying for bandwidth you're not using a lot of the time.  Virtualizing bandwidth and enabling capacity to be on-demand offers the potential to increase efficiency.

For example, a similar problem occurs with storage and servers which typically operate from 15-40% efficiency.  Virtualization enables companies to drive that efficiency into the 90% range. 

We think virtualized bandwidth could offer the same efficiency benefits if the peak or cyclical loads could be handled with an on-demand service.  We think cloud services and virtualization helps transform the whole computing and networking model into much more of a fluid workload mobility model, with on-demand, pay as you need it type of service environment, at least for the non-mission critical applications.

The efficiency benefits of network virtualization could accrue to both end users and carriers.  For end users, the carrier’s pricing model for on-demand services will hopefully be lower than the fixed rate contract.  For carriers, virtualization should enable them to drive infrastructure efficiencies much higher, lowering their capex and opex costs. 

So bandwidth on-demand for pre-provisioned equipment is really here today, but true dynamic capacity is still a future delivery that is now getting closer to reality.

Ciena on Friday issued this short press release stating that we will announce Ciena’s fiscal year 3Q10 financial results on Wednesday, September 8th.  The press release will be issued in the morning before the U.S. markets open, and a brief audio webcast will be hosted by Ciena management at 8:30am ET to review the results.

You can access the webcast from the Ciena Investor page, and a replay of the webcast will be made available later that day for those not able to attend the live event.

A bill passed by the U.S. House of Representatives this week cut $302 million in federal grants that were targeted to funding broadband expansion and adoption.  The bill reallocates the money as part of a $10 billion effort to help reduce job losses for teachers and other educators.

 

Ciena’s expert on the broadband stimulus program, Jim Gerrity, says this is a good news, bad news situation.

 

“The bad news is that there are now fewer funds dedicated to these types of broadband expansion projects that are being allocated to service providers mostly in rural areas around the U.S.,” said Gerrity. “The somewhat good news is that the original proposed cut was $600 million, but that proposed bill was rejected last month.”

 

Overall, the NTIA and RUS was given $4.7 billion in grant money to stimulate broadband deployment and adoption – part of the larger American Recovery and Reinvestment Act (ARRA) passed in early 2009.  About $1.6 billion has already been dispersed in over 100 different grants, with the remaining amount set to be allocated over the next six weeks.

 

Jim Gerrity and fellow Ciena expert Chris Janson will be detailing the status of the broadband stimulus program in a webcast on August 24^th.   You can register for the webcast here.  During the free one hour event, Jim and Chris will talk about where stimulus fund are being spent for broadband projects, give specific examples of some of the already allocated grants, as well as give an update on the FCC broadband program.

 

For more information on the August 24^th webcast and to register, visit the event page.

I’ve been talking about the booming wireless backhaul market for a while now on this blog, around both the growth of the mobile broadband market and Ciena’s success with wireless backhaul solutions.  However, a new broadband forecast from analyst firm Ovum opened my eyes to how narrowly I had been viewing the topic. 

 

It’s sometimes hard to break out of your own experiences and knowledge with a blog, and in my U.S. centric view, I saw the broadband wireless explosion as purely a mobility boom.  My bandwidth hogging iPhone was all the proof I needed there.  But Ovum’s new forecast highlights the increasing use of wireless broadband as a substitute for fixed broadband lines. 

 

Ovum’s new forecast predicts that by 2015 there will be 3.2 billion mobile broadband subscriptions worldwide, versus only 785 million fixed broadband subscriptions.  That’s four wireless broadband lines for every fixed broadband line.

 

That kind of disparity obviously can’t come from mobile devices alone. Ovum predicts that for most countries where an existing fixed broadband infrastructure isn’t already in place, wireless broadband will command the lion’s share of broadband lines.  In fact, countries like India may have only 10% of broadband access come from fixed lines.

 

Ovum is by no means predicting the end of fixed broadband – they forecast a 7% CARG for fixed broadband, and project that 36% of homes worldwide will have a fixed broadband subscription by 2015.  But much of that base will be in developed countries, where fixed line penetration will be 60% or more.

 

You can read more about Ovum’s new broadband forecast here.

It’s hardly new news that service providers are scrambling to catch up to the bandwidth demands that are being put on their network.  That story, in one version or another, has been part of the industry conversation for most of the 14 years that I’ve been in the industry.  Whether it’s the wireless backhaul portion of the network today, the last mile bottleneck five years ago, or metro and core network demands back during the Internet bubble, there’s always some part of the network that seems to be falling behind demand.

 

And in most cases, the companion message to that story line is the fact that service provider revenues aren’t growing nearly as fast as bandwidth usage.  A new report out this week from ABI Research reinforced that message.  As you can see in Kevin Fitchard’s blog post on Connected Planet, ABI predicts that mobile data usage in the U.S. will grow at a 55% CAGR through 2015, while revenues will grow only 18% over the same period.  This large gap between bandwidth usage growth and revenue growth is something I’ve highlighted before in this previous post.

 

At first glance, this seems like a quick path to losing a ton of money.  But if this story line has been repeating itself for years and years now, why haven’t all the service providers gone broke? 

 

The answer is simple, service providers don’t have to increase their revenues to match the increase in bandwidth usage because the cost of bandwidth is constantly going down.  Just like that 8GB SD card you can now buy for under $20 on Amazon, the cost of bandwidth just isn’t what it used to be.

 

For some parts of the network (for example transoceanic routes), you could argue that supply and demand has driven the drop in bandwidth prices.  But more broadly, bandwidth is getting cheaper because of innovation and technology.  Vendors like Ciena are continuously developing technologies that enable higher capacities over existing network infrastructure, allowing for large jumps in total bandwidth without the service provider having to rebuild the network.  Ciena’s 40G and 100G optical solutions are a perfect example.  As these technologies mature, the cost of deploying them declines, further reducing the cost of new bandwidth.  Just look at the declining cost of deploying a 10G line over the last decade.

 

Analyst firm Ovum as been tracking the cost of backbone DWDM bandwidth over time.  As you can see from Figure 1, the price of bandwidth has been declining dramatically, with the cost per Gbps per kilometer falling to under $1 by 2009 (special thanks to Ovum for the data and chart).

 

Over on the BusinesofVideo.com blog, Dan Rayburn puts the power of declining bandwidth prices into a context we can all relate to, Netflix video streaming:

 

“To put the rate of pricing decline in terms everyone can understand, today Netflix pays about five cents to stream a movie over the Internet. If Netflix tried to do this in 1998, at the same quality they are doing it today, it would of cost them $270 per movie.”

 

So maybe the recurring headlines about the huge gap between service provider revenues and exploding bandwidth demand aren’t as dire as they may seem.  I’m not suggesting that you completely dismiss the revenue growth versus bandwidth growth gap that will undoubtedly continue to be a storyline in our industry.  It’s still an issue that service providers are dealing with every day.  Just keep in mind that the cost of bandwidth is the third piece of the story.