Archive for the ‘Opinion’ Category

Sources of Latency in Optical Systems Used by Financial Customers

Tuesday, August 24th, 2010
RSS Feed Subscribe to EtherNEWS Bookmark and Share
Sources of Latency in Optical Systems Used by Financial Customers
Every optical-transport function that is intended to contribute to the successful beaming of traffic across fiber connections can inject small amounts of latency into the process. Today’s financial network manager must be cognizant of all of them.
Color conversion (“Transponding” and “muxponding” ) are two common sources of delay. When traffic is readied for transport across a Wavelength Division Multiplexing (WDM) optical network, it’s converted to a color of light to be carried across the glass fiber. This is transponding.
Additionally, in many cases, lower-speed traffic, such as a 1 Gbps information feed, is aggregated into a higher-speed signal such as a 10 Gbps transport link. When multiple feeds are muxponded, significant latency can occur
Optical amplification also can produce latency in a financial network.  The “Erbium-Doped Fiber Amplifier” (EDFA) can introduce microseconds of latency, so financial network managers should instead demand amplifier designs that have been specifically engineered for the contemporary low-latency challenge.
Dispersion compensation is another potential source of latency. At higher speeds, such as 10 Gbps, optical data signals sometimes smear into a rainbow of colors within a fiber cable. It’s an effect called “chromatic dispersion” and its effects can include an increasing degradation of the data signal over distance.
There are fiber Bragg gratings (FBGs) that have been designed to offset dispersion while introducing only negligible delay. In some cases the alternative is to rely on hundreds of kilometers of dispersion-compensating fiber (DCF) that is ill-suited for low-latency applications.
Electrical signal regeneration likewise introduces additional latency. Conventional regeneration techniques may yield hundreds of microseconds of delay. State-of-the-art low-latency techniques will introduce latency of only nanoseconds.
http://communication-solutions.tmcnet.com/topics/optical-access/articles/96342-wringing-latency-from-financial-networks.htm

Every optical-transport function that is intended to contribute to the successful beaming of traffic across fiber connections can inject small amounts of latency into the process. Today’s financial network manager must be cognizant of all of them.

Color conversion (“Transponding” and “muxponding” ) are two common sources of delay. When traffic is readied for transport across a Wavelength Division Multiplexing (WDM) optical network, it’s converted to a color of light to be carried across the glass fiber. This is transponding.

Additionally, in many cases, lower-speed traffic, such as a 1 Gbps information feed, is aggregated into a higher-speed signal such as a 10 Gbps transport link. When multiple feeds are muxponded, significant latency can occur

Optical amplification also can produce latency in a financial network.  The “Erbium-Doped Fiber Amplifier” (EDFA) can introduce microseconds of latency, so financial network managers should instead demand amplifier designs that have been specifically engineered for the contemporary low-latency challenge.

Dispersion compensation is another potential source of latency. At higher speeds, such as 10 Gbps, optical data signals sometimes smear into a rainbow of colors within a fiber cable. It’s an effect called “chromatic dispersion” and its effects can include an increasing degradation of the data signal over distance.

There are fiber Bragg gratings (FBGs) that have been designed to offset dispersion while introducing only negligible delay. In some cases the alternative is to rely on hundreds of kilometers of dispersion-compensating fiber (DCF) that is ill-suited for low-latency applications.

Electrical signal regeneration likewise introduces additional latency. Conventional regeneration techniques may yield hundreds of microseconds of delay. State-of-the-art low-latency techniques will introduce latency of only nanoseconds.

http://communication-solutions.tmcnet.com/topics/optical-access/articles/96342-wringing-latency-from-financial-networks.htm


RSS Feed Subscribe to EtherNEWS Bookmark and Share

Why LTE Requires Low Latency

Sunday, May 9th, 2010
RSS Feed Subscribe to EtherNEWS Bookmark and Share

LTE bandwidth and latencyThere are two obvious reasons why LTE networks will require both lower latency and higher bandwidth backhaul: LTE simply is the lowest latency air interface and features bandwidth that cannot be supported by legacy backhaul protocols.

Edge and EVDO networks can handle average peak data rates using two or three T1 links, but need more than that to handle peak rates. But HSPA networks cannot do so, either efficiently or conveniently.  To support peak rates on an HSPA network, about 45 Mbps is required.

An LTE network using a 10-MHz channel requires nearly a DS-3 (45 Mbps) just to handle average load, and needs an Ethernet connection to handle peak loads.

Also, where older GPRS or EDGE data networks featured round-trip latencies in the 600 millisecond to 700 msec. range, LTE networks feature round-trip latencies in the 50 msec. range.

That means Ethernet speed backhaul and lower-latency performance is required.

By Gary Kim


RSS Feed Subscribe to EtherNEWS Bookmark and Share

Scaling 3G & LTE: SOAM Issues

Wednesday, April 14th, 2010
RSS Feed Subscribe to EtherNEWS Bookmark and Share

We recently caught up with Craig Easley, president of the Carrier Ethernet Academy and board member of the MEF, at CTIA 2010 in Las Vegas.  Craig was at the conference to provide training for Ethernet mobile backhaul operators’ engineering staff as they prepared to roll-out large-scale 3G & LTE services (watch a video replay of this 30 minute class).  One focus of the session was the challenges providers face when implementing Y.1731 Service Operations, Administration & Maintenance (SOAM) to monitor and manage Quality of Service (QoS) in these performance-critical, all-packet backhaul networks.  Following is a short dialog between Craig and Patrick Ostiguy, President & CEO of Accedian Networks, who provide service assurance equipment to leading backhaul deployments to both mobile operators and wholesale backhaul providers.

Speakers:

speakercea

[CE] Craig Easley, President, Carrier Ethernet Academy

[PO] Patrick Ostiguy, President & CEO, Accedian Networks

[CE] If you received a new software release from your favorite switch vendor that supports this connectivity check messages (CCM), they can be configured just to run end-to-end and so you have only two points in the network actually sending and receiving the CCMs, or you can configure them to be processed by each inner management entity as well to provide complete path-performance information.

If you do that at a high enough level of granularity, it’s possible to actually flood the processing of the equipment that’s in the middle in such a way that you get a false positive that you have a problem.  Your data traffic isn’t being processed and handled by every one of those interim points, but the connectivity check messages are.  They are being read and time-stamped and then forwarded along.  And if an intermediate switch is overrun with time-stamping of connectivity check messages, the accuracy will be off and the operator might think there’s a latency problem – when in fact the end-to-end latency of the actual data itself is within spec.

[PO] The sheer amount of OEM sessions that Mobile backhaul is faced with when implementing these standards is creating this very interesting challenge right now. These providers deliver 3 to 4 classes of service per tower and also want to have an OAM Performance Monitoring (PM) session for each of those classes in addition to a CCM continuity check message session going to each of those towers every second.

So considering that you can have 200+ of thoe towers being served by a single Mobile Switching Center (MSC), it rapidly increases the amount of OAM sessions you are dealing with. So, converging at the MSC you can easily have 1000+ sessions that have to be terminated at a critical aggregation and hand-off location – typically served by a 10 GigE link.  That is extremely dense, even for the “big iron” switch-routers that are out there from the big vendors.

[CE] This is something that people are just starting to wrestle with: essentially, all of the big equipment manufacturers are releasing support for OAM – some are releasing new hardware to go along with it, but most are just doing it in software.  And if you just have OAM capability in software there is only certain amount of compute power in those switches that are already deployed.  So the good news is: it’s a software upgrade, you don’t have to deploy anything new.  The bad news is: you may be pouring a little bit “too much sand in the bucket” and exhausting the capability of the switch.

[PO] Like Craig suggests, this level of processing cannot be done by software running on the routers’ existing cards. In that context our customers have asked us to develop a product to alleviate this problem by providing a pure hardware based design that is independent of traffic load and can therefore handle thousands of OAM sessions, while offering microsecond precision one-way measurements. The beauty of this new product, the MetroNODE 10GE, is that because it is Y.1731 standards-based, it allows the operator to test performance to each and every tower, whether the cell site employs dedicated hardware such as NIDs, or uses cell-site routers or base stations supporting this OAM standard.

[CE] Agreed.  More and more people, I believe, will deploy special purpose network interface devices like the Accedian units to make sure that they get accurate data coming back from the network in terms of the SLA, especially in mission-critical and ‘zero tolerance for error’ latency environments like mobile backhaul.

You can watch a more detailed overview at Accedian.com/10, or watch the CTIA training session by Craig Easley at Accedian.com/cea-ctia.


RSS Feed Subscribe to EtherNEWS Bookmark and Share