Content Delivery Networks
in the Cloud World
With the explosion of content in our hyper-con-
present some fixes needed to get the network
nected world, as one writer in this eGuide puts it,
ready for the IoT revolution, what to consider as
“CDNs are blossoming like tulips in Holland.” But
organizations shift more business-critical applica-
as more and more of the world moves to the cloud,
tions to the cloud, how personalized online content
and user expectations of content access and quality
is changing internet performance requirements, the
continue to grow, infrastructure changes are ahead.
importance of your website or online content being
In this eGuide, InfoWorld and Network World
reachable and more.
feature
cloud chronicles
The 3 fixes needed
to get the network
ready for the
IoT revolution
Netflix is (not
really) all in on
Amazon’s cloud
5 questions to consider
as you shift more
business-critical applications to the cloud
Rain clouds ahead:
10 bold predictions
for the cloud
industry in 2016
Many traditional
networks are still
manual, static and
complex, which isn’t
ideal for IoT.
Netflix has completed
its migration to AWS,
with one big caveat:
Netflix operates its
own Content Delivery
Network outside of the
Amazon cloud.
Consider these five key
questions before making
any decisions about
the cloud.
Warning signs for the
cloud heading into 2016.
(#7 on the list: Content
distribution networks
meet NFV.)
2
4
5
tech primer
opinion
8
IDG contributor network
column
How personalized online
content is changing
internet performance
requirements
My website is up,
but is my content
reachable?
The internet
isn’t ready for
really big news
As more websites personalize content for users,
optimization efforts need
to adapt.
Tracking metrics within the
cloud will give you valuable
insight to help keep your
applications up and running, but what’s important
is whether your website or
content is reachable.
A worldwide, globally
gripping news event
would likely crash the
major media providers, sending us back
to our TVs and radios.
12
14
16
feature
eGuide
IDG Enterprise is
COMPUTERWORLD
The 3 fixes needed to get the network ready
for the IoT revolution
Many traditional networks are still manual, static and complex, which isn’t ideal for IoT.
BY JACK WATERS, CTO OF LEVEL 3 COMMUNICATIONS | Cisco
has to do with fiber-optic networks. Getting cities—and
estimates 50 billion devices and objects will be connected to the
consumers—hardwired with fiber will be a necessity of the
internet by 2020. And that estimate may be low. If consumers
future. The term “Smart City” has been used to character-
count every device that draws power in their home—lamps, light
ize these communities that are investing in infrastructure
bulbs, kitchen gadgets—and then factor in objects at work, there
and advancing science and technology efforts to securely
may be many more billions of connected devices by then.
collect and use data to do everything from decrease energy
But the problem is, many traditional networks are still manual,
consumption to cut overhead costs and improve the life of
static and complex, which isn’t ideal for IoT. To realize the prom-
residents. The White House stepped into this arena last Sep-
ise of a hyper-connected future, three shifts must take place.
tember when it announced a Smart Cities Initiative to invest
more than $160 million in the concept.
1. Fiber
The bandwidth needed for the onslaught of IoT connected
devices should be enough to make anyone think about the
Right now, we aren’t actually seeing an overwhelming adoption
fiber. At the Consumer Electronics Show, a lot of coverage
of IoT devices in personal homes or offices. Cisco stated in
focused on this because of the emergence of 4K ultra-high-
its same report that more than 99% of things in the physical
definition television. But there were poignant examples
world remain unconnected. However, it’s only a matter of time
as well, one involving a medical use case in Cleveland. An
before every single aspect of our life is internet dependent.
organization helped deploy a fiber-optic network capable
In today’s environment, if consumers want their devices
of 100Gbps data transfer speeds to support high-definition
to be accessed outside of their homes or private networks, a
video so remote neurosurgeons could assist in operations,
user has to go into their Wi-Fi router and portmap it to an out-
because obviously buffering and delays are not acceptable.
side network. This is complicated and not very user friendly.
The underlying conversation around 4K and 100Gbps
2
2. IPv6
of 17
The fix: Give public IP space to all of the “things.” Great
Getting cities—and
consumers—hardwired
with fiber will be a
necessity of the future.
eGuide
IDG Enterprise is
COMPUTERWORLD
in theory; however, the rate of new “things” is growing at
up today, we’re going to have a serious security situation.
such a rapid pace, the current method of assigning ad-
Right now, many companies are making internet-connected
dresses won’t be able to accommodate the volume. In order
devices that aren’t able to be patched or easily updated with
to provide addresses for every device, the internet will need
new security rollouts. Considering many IoT devices collect
to transition from IPv4 to IPv6.
personal data, security should be a concern for all.
So what’s the hold up? Right now there isn’t imminent
A few months ago we saw a nonmalicious hack of a chil-
financial motivation or competitive pressures for broadband
dren’s toy company. The hacktivist, a Grey Hat, was able to
providers to transition. As long as the market can efficiently
exfiltrate photos and personal information of young consum-
broker the remaining, sellable IPv4 address space to those
ers. The Grey Hat took these actions to show the company
that require it, the pressure to migrate to IPv6 will not fully
how unsecure they were and teach the industry a lesson.
materialize.
This wasn’t the first time a child’s toy was in the spotlight
Over time, as the number of IoT devices increases and
IPv4 addresses grow more scarce, financial and competitive
pressures will rise accordingly, eventually leading to eco-
for having security vulnerabilities, and probably only illustrated the tip of the iceberg.
It goes without saying that IoT developers should do
nomic incentive for IPv6 transition. Consumers will demand
more to secure their products. For organizations using
the ability to interwork with every device seamlessly with
IoT, it is essential to do a rigorous analysis of the security
speed and ease, and expect that the “internet” will continue
controls built into IoT devices and services they wish to use.
to be the ubiquitous, any-to-any network they grew ac-
At a minimum, an audit of an IoT device’s communications
customed to in its IPv4 origins. An IPv4 and IPv6 internet,
channel; use of encryption; an analysis of the type of data it
patched together with transitional technologies such as
collects, stores and transmits; and the security of the end-
Network Address Translation, won’t be able to scale to the
point(s) with which it communicates, is paramount.
levels forecasted for IoT devices—not without cumbersome
Given IoT’s growth rate, and the resulting broadening of
constraints. This perfect storm of consumer expectations
the cyber-attack surface, organizations must be ever more
and financial incentive is what is required for IPv6 to be-
vigilant in conducting comprehensive risk analyses and in
come a reality after all of these years.
the implementation of proper governance structures. A riskbased approach is the best way to balance the risk of using
3. Security
IoT with its unlimited productivity benefits.
There is a lot of concern about what it will mean for the
threat ecosystem to have millions of connected devices—es-
3
The promise and lure of IoT is exciting for both consumers
pecially those managed by consumers—available for mali-
and the industry. If these three areas become a focus, we can
cious activity. If we don’t address the security issues rearing
move toward an “Internet of Everything” connected future.
of 17
eGuide
IDG Enterprise is
COMPUTERWORLD
Netflix is (not really) all in on Amazon’s cloud
Netflix has completed it migration to AWS, with one big caveat: Netflix operates its
own Content Delivery Network outside of the Amazon cloud.
BY BRANDON BUTLER, NETWORK WORLD | Publications
of all of its users, their preferences, what they watch, what they
across the Internet have praised Netflix on the triumphant
click on; and the company organizes its massive catalog of
migration of its operations into Amazon Web Services’
content in the cloud. Netflix remembers exactly where you last
cloud. But there’s one caveat: Netflix isn’t quite all in on
stopped watching something and picks it right back up from
Amazon’s cloud.
there hours, days or even weeks later. It seems like magic. And a
Perhaps the thing Netflix is most known for—streaming videos
to users around the globe—is not done in Amazon’s cloud.
Netflix is a big user of Amazon’s cloud—no one can argue
lot of it is run out of Amazon’s cloud.
But there’s a whole lot of other magic that happens that’s not
run out of Amazon’s cloud. Netflix operates its own Content De-
that. It’s been a poster-child example of how to use AWS, and
livery Network, named Open Connect, in which it stores the video
even open sources tools it has built for Amazon’s cloud. Com-
content it streams to users in data centers around the world to
pany executives—including co-founder and CEO Reed Hast-
be as close to its end users as possible.
ings—have been on stage at AWS’ re:Invent conference since
“The best way to express it is that everything you see on
“The best way to
express it is that
everything you see on
Netflix up until the
play button is on AWS;
the actual video is
delivered through
our CDN.”
2013 chronicling their impressive journey of migrating away from
Netflix up until the play button is on AWS; the actual video is deliv-
Joris Evers
managing data centers and relying more and more each year
ered through our CDN,” Netflix spokesperson Joris Evers said.
Netflix Spokesperson
on AWS’ cloud. It published a blog post earlier this month titled
“Completing the Netflix cloud migration.”
Netflix uses AWS for a wide variety of features: It keeps track
4
of 17
So while Cloud Chronicles congratulates Netflix on its use
of AWS, just keep in mind, the company’s isn’t quite ALL in on
Amazon’s cloud.
eGuide
IDG Enterprise is
COMPUTERWORLD
5 questions to consider as you shift more
business-critical applications to the cloud
Consider these five key questions before making any decisions about the cloud.
BY SCOTT HILTON, NETWORK WORLD | The migration to the
quality degradation, your IT team must know how to monitor for
cloud is well underway. More than half of the CIOs polled at
these issues and how to optimize performance.
last year’s Gartner Symposium IT Expo said they would be em-
Another thing to consider is knowing your “mean time to in-
ploying a cloud-first strategy moving forward. Verizon’s State
nocence”—how to quickly and with precision determine whether a
of the Market: Enterprise Cloud 2016 report found that 84%
customer-impacting issue is the result of your application, data or
of enterprises say their cloud usage has increased in the past
compute environment, or an issue with internet routing, outages
year, and half of enterprises said they will use cloud for at least
or a cloud/SaaS provider. While cloud providers, CDNs and accel-
75% of their workloads by 2018. Even more telling, half of the
eration services may claim to be “always up,” that does not mean
respondents use two to four cloud providers.
they’re always reachable by your customers.
Whether you’re looking to move to the cloud, considering the
pare performance and value across clouds, consider these five
2. If you are a global company, how do you
ensure global availability on the cloud?
key questions before making any decisions:
If you have end users accessing your internet assets at all hours
diversification of your cloud portfolio, or simply looking to com-
of the day from points around the globe, you need to ensure your
1. How do you monitor your internet infrastructure?
end-user experience is equal no matter where or when a custom-
While you may be monitoring your internal infrastructure and ap-
er accesses your website, app or content. Determining availability
plications, what information do you have about external internet
means understanding if your service is available to customers and
connections your customers rely on to connect to your business?
partners. This is especially important to understanding the con-
If connections to a specific data center go down or a cloud pro-
nection performance to your selected cloud providers and CDNs.
vider can’t reach important markets, your system admin should
A 2015 Google cloud outage that lasted for almost two hours
have the tools necessary to route traffic to another center to keep
was related to a software issue inside its virtual network traffic
your services operational. And if you’re experiencing latency or
routing. If your business was using a single cloud instance and
5
of 17
eGuide
IDG Enterprise is
COMPUTERWORLD
a Google cloud client and you were not monitoring for network-
traffic as quickly as possible to ensure minimal interruption. As an
wide availability, you would have experienced this outage and
added bonus, having multiple CSPs can help protect against prob-
your availability would have been impacted.
lems before they even occur by allowing you to deploy apps and
So how does your business ensure that you have strong
availability no matter the condition of the internet locally or in a
services to two (or more) cloud instances so they can back each
other up using a global load balancing or failover architecture.
region halfway around the world? Distributing hosting on mulsingle regional outages. It is also recommended that you monitor
4. What are the biggest risks to optimal
internet performance?
your global network to determine load levels. Should you notice
The biggest risk is simply in not knowing the real-time state of
performance issues, there are tools available to balance these
your performance—where are the internet latency issues, what is
loads and ensure that availability remains strong regardless of
global availability and reachability like, and how is the real-world
the time, day or situation.
performance of end-users. Between the time a potential cus-
tiple cloud vendor sites protects customers from single site or
tomer enters a domain and a page opens, a tremendous amount
3. Is using just one cloud provider sufficient or
should you consider using multiple providers?
happens. Errors can occur within the network, at the CDN or ISP
While not all cloud providers are created equal, using multiple ser-
website can be slow for many reasons as well, and most consum-
vices is the only foolproof business continuity and risk mitigation
ers—and companies, for that matter—have little insight into the
strategy. It also can be an effective strategy to manage cloud costs
“how” and “why” of their performance issues.
based on peak usage, cost spikes and global load balancing.
Using a variety of CSPs gives your company access to multiple
level, with your SaaS providers or with a customer connection. A
Pipeline failures, routing anomalies, latencies, packet loss and
security threats (like man-in-the-middle attacks and DDoS) are
cloud instances (locations), allowing you to meet customers in
just a few of the common risks to a smoothly operating network.
local markets and leverage local connections. Using an advanced
Understanding your connections—and monitoring, controlling and
DNS-based Traffic Management solution with geo-location, you
optimizing them—is the true test of internet performance and in
can control which cloud instances serve which customers, and by
understanding and mitigating against weak spots in your network.
working with the right tools you can manage this from one place.
ibility and value, allowing you to scale and deliver an always-on
5. How do I keep my site up in the event
of an outage?
impression. The ability to access different pathways also comes in
Outages are game changing for your business and they happen
handy when there are outages or slow load times—whether due to
with surprising regularity—upwards of 3,000 times every day
a traffic routing problem or a malicious attack. Working with mul-
on the global internet. And it is difficult for most companies to
tiple providers will help you circumvent these issues by rerouting
detect an internet connection failure without a way to measure
Working with multiple clouds gives your business more flex-
6
of 17
Distributing hosting
on multiple cloud
vendor sites protects
customers from single
site or single regional
outages. It is also recommended that you
monitor your global
network to determine
load levels.
eGuide
IDG Enterprise is
between your customers and your assets.
Constant monitoring of your internet-facing architecture is the
COMPUTERWORLD
the risk in a loss of sales, revenue and brand confidence and loyalty. Working with the right partner with true visibility into your
only true way to ensure you will not be affected by outages for
infrastructure will allow you to mitigate the risk of an outage,
a prolonged period. If you cannot properly detect outages your
by identifying it before or during a disruption and making near-
business will not be in position to manage a response, mitigating
instantaneous rerouting decisions.
7
of 17
eGuide
IDG Enterprise is
COMPUTERWORLD
Rain clouds ahead: 10 bold predictions
for the cloud industry in 2016
Warning signs for the cloud heading into 2016. (#7 on the list:
Content distribution networks meet NFV.)
interrupting anyone’s business plans in particular, I have a
2. Digital encryption divides will cause
massive cloud exception handling
jaundiced eye toward how clouds are bringing us rain, and how
As the SHA-1 encryption protocol is retired, websites will be
in 2016, the rain will drench us and try to drown us in data.
pushing users from the cliff of dead encryption. SHA-1 doesn’t
BY TOM HENDERSON, NETWORK WORLD | While I don’t like
take much computing power to use, and its users are gener-
1. More cloud services shakeout/shakeup
ally those with old phones, ancient operating systems on slow
Already we’ve seen the shakeout begin. HP Helion is now ad-
desktop and notebook hardware, and even early (and cheap)
junct to HP’s old ally Microsoft and its Azure Cloud. Microsoft,
tablets. Will users of low-value equipment be able to success-
which is in turn now scared to death of unlimited storage and
fully transition back to the web?
the resources it’s chewing up, imposed strictures to OneDrive.
No more happy terabytes with a license of Windows 10.
Portals as a business investment are down. Yahoo tried to
Once the lights are turned off on SHA-1, easy-to-do encryption
by millions of older devices will be eliminated. So will those users’ ability to run simple https web pages.
do things with Flickr, but now seems poised to sell much of its
in the game, but Facebook, the world’s greatest time suck, chal-
3. Chromebooks become the cloud
access device of choice
lenges all.
As cloud app savvy increases, the need for huge notebook stor-
non-Alibaba holdings. There are many minor portal players still
Here’s where I believe that something will happen: Cloud stor-
age and processing power is in decline. Instead, we have cloud
age will grow slowly, but steadily, (see No. 3) because as we all
apps, simplicity and generally less stuff to worry about being sto-
know, nobody erases anything, and data will grow to fill the space
len or replaced. It becomes much more troubling for companies
allocated to it. How much junk is snoozing in your SAN? Been on
that make traditional stuff, as production lines shift rapidly toward
Pinterest or Instagram recently?
Chromebook sleekness as the alternate device to a smartphone.
8
of 17
Here’s where I believe
that something will
happen: Cloud storage
will grow slowly, but
steadily, (see No. 3)
because as we all know,
nobody erases anything,
and data will grow to
fill the space allocated
to it.
eGuide
IDG Enterprise is
COMPUTERWORLD
4. Your television now rats you out to the
cloud (and maybe IoT, too!)
Stir Fry, Seasoning Salts, and the infrastructure bits.
Huge unsorted piles of data are being generated by your
zation harmony will be given more than superficial lip service, as
holiday gift of digital ears in the living room. Yes, that cute
each release of OpenStack requires even more studied intimacy
little Samsung television is trying to take as much data as is
at the API levels. Glue apps and components will do their best to
rationally possible and transmit it to motherships for purposes
become stickier and stickier as the elements of OpenStack look
of analysis.
less like a stack and more like a homogeneous methodology. My
At Best Buys or the Big Box, you’ll need to carefully read the
This also means that the dreaded specter of interstack organi-
prediction: OpenStack Releases actually slow down as adoption
terms of service (oh, right, it’s in the box) prior to connecting
climbs, and more cooks get into the OpenStack kitchens, poten-
that TV to your Wi-Fi. And wear your bathrobe, please.
tially causing drain sludge.
Great piles of data grub might be something you’ve actumyriad new verbal assistants that not only are incredibly handy,
6. New for 2016: Huge Organization-as-aService (or, HOaaS)
but which know your location and are dutifully listening for your
Huge-Organization-as-a-Service (HOaaS) like the Dell-EMC-
next command.
VMware merger will become commonplace, as Oracle, IBM,
ally enabled yourself via Amazon’s Echo, Siri, Cortana and the
When will the temptation for organizations to just, well,
Dell, Microsoft and even Rackspace attempt to eat Amazon’s
analyze everything you do commence? How are you sure it’s not
lunch. Dazzling, dizzying and absolutely opaque-to-compare
happening right now? Plenty of people put black tape over the
offerings will emerge, with price lists published biweekly (daily
cameras on their tablets (perhaps for good reason, too) knowing
in Washington, D.C.) in an attempt to spark a cloud services
that the cloud is like a sieve for your personal information. And
market share war, if only for stockholder visibility as everyone
IoT analytics will strengthen as an industry. My stepson once
tries to remain relevant.
wore tinfoil for a short period of time, carefully on his head. He
was a pioneer.
The HOaaS idea, time-honored, is that organizations that are
busy, resource-challenged or in need of that certain propellant
that takes them either wholesale or a little bit into the cloud
5. OpenStack recipes for cloud containers
become dominant
sometimes need to obtain all of their kit from just one contact
So you wanted to cook in the cloud, eh? You’re not alone. Now
expand it all. The One-Stop-Shop approach has merit for many
that there’s an Open Container Initiative with seeming teeth,
reasons, and in the cloud, captivity may not be as difficult, due
battles over deployed systems security provenance may be
to an increasing number of cloud transport/mirroring apps.
quieted, at least for a while. Value-added pieces evolve, with
Food Channel-like organizations booming to provide Container
9
of 17
that they can scream at when things go awry—or they want to
This is a space where some organizations will put on bright,
shiny new clothing, dubbing themselves newly evangelized (and
eGuide
IDG Enterprise is
Suddenly Agile!) service providers, while behind the curtains,
COMPUTERWORLD
Why? Chromebooks, tablets and other generally cloud-serviced
the same old smoke and dance will fail to enchant all but their
devices will continue to dominate sales. People hate to back up.
leaden client base. HOaaS will need to be fleet-of-foot to survive
Some have never done this. Worse: Some businesses never do
the tracks laid down by the leaders.
this. Still worse: Some government agencies don’t either.
As HTML5 grows up, web services—oops, “cloud services”—
7. Content distribution networks meet NFV
will continue to grow. Consumers, finding that a 16GB iPhone
CDNs are blossoming like tulips in Holland. They’re the only
or 32GB NextBook are simply too small for their storage needs,
way that many cloud providers are surviving, as NetFlix,
will be happy to avail themselves of network storage, backup and
iTunes and even captive CDNs from Microsoft and Amazon
archiving, and not having to take out the trash.
attempt to service their clientele. But the ability to rapidly
reconfigure and take advantage of regionalized content distri-
In some ways, the cloud is like computational crack: Once addicted, walking away might be murderous.
bution for cloud-to-branch and Safe Harbor quarantine means
rapid configurations as situations warrant.
Open a pack of NVF—network virtualized functionality—sprin-
9. The international data blockade year begins
Without new treaties, cloud services and providers will now
kle across your network bottlenecks, and suddenly intercarrier
need to sequester data into various new international geo-
operations and network reconfigurations become a cinch!
graphic regions. The European Union now bars many kinds of
OK, perhaps it’s not going to be prevalent in 2016 or even 2017,
international data transfers, allowing Safe Harbor agreements
but the primitives are starting to emerge that allow large cloud cli-
that once permitted such transfers to push cloud providers
entele to do rapid network reconfigurations that permit major CDN
into dividing data—and data business locales.
functionality changes—especially within the spreading circuits of
major cloud providers done in minutes, not months.
Will NVF and SDNs mean the difference between servicing all the
queues or extreme network cloud constipation? We’ll soon find out!
Data must now have provenance, kind of like fish and wine, so
that we know varying rules and regulations—largely meeting the
needs of privacy—aren’t violated. Many cloud hosting organizations are opening multinational hosting sites for nexus, so as to
comply with privacy regulations. How does a German traveling
8. 2016: The Year of the Consumer Cloud
in Aruba deal with data sequestering? How about my upcoming
Although Microsoft and phone service providers have begun
trip to Barcelona? Can I email home? What about those ads I’ll
to impose constraints on the concept of “unlimited,” the
click on while in Catalunya? Does my click stay in Catalunya, or
number of online streaming and storage services will rise and
is it MY click, and can it go home with me? Must Cortana speak
clog networks like no time in history, I predict. Whether social
Catalunyan and not Castillian Spanish? Can I use sign language?
media, services applications or raw storage, consumers are
adapting to offloading their lives into the cloud.
10
of 17
The devil of the details of what Safe Harbor and data sequestering mean must become the crux of policy—not easily negoti-
CDNs are blossoming
like tulips in Holland.
eGuide
IDG Enterprise is
ated in a U.S. Presidential election year. In the interim, many in
COMPUTERWORLD
What’s woefully apparent is that car cloud security is as primi-
the cloud services sector are looking very closely at outcomes as
tive as a bad password on a good day. It’s become the crux of
infrastructure becomes replicated into the EU, but also Canada,
lots of laughter and mayhem at BlackHat. So gruesomely bad are
and soon, great walls of data will become the norm … until
the minimums—car door locks—that key replicators pop up in
something else happens.
disguise on eBay frequently. And nothing talks to nothing.
10. Cars meet cloud
of whom traditionally invents its own secret sauces, and with
Although a slow maturation in automobile (and service vehicle)
luck, OEMs them to each other. Once automotive cloud sites are
data has been around for a while, virtually all cars today have a
breached—which they inevitably will be—the hardwired-thinking
There is little interoperability among car manufacturers, each
transponder built in to them. Whether it’s autopilot or GPS sug-
and design of automotive cloud technology will be devilishly dif-
gesting the nearest Starbucks, cars will talk to the cloud, each
ficult to repair, recall or retrofit. Worse, at 80 mph on a Wyoming
other, hoarders and sifters of big data and perhaps emergency
freeway, it could mean death and litigation, the size of which
responders (as some do now). There is money to be made. As
must cause insurance company actuaries to awaken in cold
usual, it’s money first, and not security and safety. This fact will
sweats in the middle of the night.
become amplified in 2016, I predict.
11
of 17
But hey, it’s the cloud. Damn the security; full profits ahead!
eGuide
IDG Enterprise is
COMPUTERWORLD
How personalized online content is changing
internet performance requirements
As more websites personalize content for users, optimization efforts need to adapt.
BY ZEUS KERRAVALA, NETWORK WORLD | Content delivery
years, but an even bigger driver is personalized content. Content
networks (CDNs) have been the de facto standard method of
personalization comes in many different forms, but it revolves
improving internet content for years. CDNs work by prefetch-
around the segmentation of the audience the business is trying to
ing content and putting it as close to the user as possible. The
capture. This includes:
closer a user is to a CDN point of presence (POP), the better
the performance.
However, as the great poet and song writer Bob Dylan once
• Location information: Everything from apparel to cars to
coupons can be personalized based on location. The clothing
said, “the times they are a-changin’,” and the way internet con-
needs of someone living in Winnipeg will vary quite differ-
tent is optimized needs to change as well. CDNs have been effec-
ently from those of a person living in Arizona. This is why so
tive because most of the content flowing across the internet was
many websites now try to capture location information as a
somewhat static, meaning it is cacheable. Today, that’s not the
case, as more and more content is becoming personalized and
dynamic, and therefore noncacheable.
The challenges behind optimizing static content are well
first step in a personalized experience.
• User behavior: Online marketers are starting to study how
visitors to a website behave, and then capture this in real
time. Traditionally, email campaigns relied heavily on past
understood today. This is content that doesn’t change very often,
behavior, but the personalized web depends much more on
and when it does, the changes are predictable. JavaScript and
current data and analytics.
images are great examples of static content that can easily be
• Unique interests: Not everyone has the same interests, so
cached. CDNs came into existence specifically to address static
it’s important to capture the motivating factors behind each
content as it’s cached and served from the edge of the network.
visitor to a site. For example, on a travel site one person may
Over the past few years, the industry has seen an explosion
always be searching tropical destinations, while another may
in content that is uncacheable. Dynamic objects and AJAX calls
be on the prowl for great ski resorts. The travel site’s ability
have been increasing the amount of dynamic content for a few
to understand this and personalize the information gives
12
of 17
eGuide
IDG Enterprise is
COMPUTERWORLD
them a significant advantage over sites that push the same
congestion. Many of the use cases I highlighted were for bidi-
content to everyone.
rectional content, such as media and gaming, and the company
has had some early success with file sharing and backup and
I’ve discussed this topic with a number of marketing professionals, and almost all of those I interviewed agreed that content
recovery services.
At first glance, these use cases may seem quite different, but
personalization was a critical success factor moving forward.
the one point of commonality is that these are all highly person-
However, almost all of the businesses I interviewed expressed
alized, and therefore uncacheable services. Personalized content
a lack of understanding of how to optimize the performance of
is growing in popularity across almost all verticals and company
personalized content.
sizes. Any organization that has a digital strategy today should
So, if CDNs aren’t the answer here, what can improve the
be thinking about personalized content, and I believe this will be
performance of the highly personalized internet? In October
the low-hanging fruit for companies like Teridion that can solve
2015, I wrote about a startup called Teridion. The company’s
the challenge of optimizing personalized content.
Global Cloud Network was also featured in a Network World new
product roundup.
In my post, I compared Teridion to being Waze on steroids,
The internet is continually evolving, and the digital era will
change the internet experience into something that becomes
increasingly personalized. Businesses should look beyond CDNs
inasmuch as the company continually computes the fastest
to ensure that customers have an experience that keeps them
path between two points and creates a virtual overlay to avoid
coming back.
13
of 17
The internet is continually evolving, and the
digital era will change
the internet experience into something
that becomes increasingly personalized.
Businesses should look
beyond CDNs to ensure
that customers have an
experience that keeps
them coming back.
eGuide
IDG Enterprise is
COMPUTERWORLD
My website is up, but is my content reachable?
Tracking metrics within the cloud will give you valuable insight to help keep your applications
up and running, but what’s important is whether your website or content is reachable.
BY MATT LARSON, NETWORK WORLD | I recently read an article
you can’t control whatever outages or downtime the customer’s
in InfoWorld by David Linthicum entitled, “Good Cloud Ops Need
ISP or other ISPs in their path are experiencing.
Good Cloud Metrics.” It caught my attention for two reasons.
First, Linthicum is correct in his assessment that metrics are
Measuring performance across the network, however, is the
first step to help mitigate issues for your customers. With that in
as important in cloud operations as they are in any other aspect
mind, let’s take a look at Linthicum’s three points about cloud
of your operations. To paraphrase Lord Kelvin, “to measure is to
metrics:
know.” And knowing lets you do something about it.
rics he describes are important, they are only part of what you
1. “You can trend data and spot issues with
recent operations.”
need to measure to fully understand your internet performance.
You need to measure network performance continuously to under-
Linthicum focuses on measuring the performance of the systems
stand the baseline and detect deviations, and there are different
and applications within the cloud provider. But as I’ve discussed
mechanisms for doing so. One is Real User Monitoring, or RUM,
before, system and application performance is only part of the
where you instrument your web pages to send code to run in the
story—network performance across the internet between the
user’s own browser to take performance measurements (there’s
cloud and end users is a huge factor, too. Because, of course,
an entire JavaScript API for this purpose). You can also infer user
those end users are your customers.
performance from “synthetic measurements.” For example, if you
But second, it caught my attention because while the met-
Another way to look at it is that tracking metrics within the
know performance from a given network provider in a certain city
cloud will give you valuable insight to help keep your applications
to a particular cloud provider’s data center is degraded, and you
up and running, but the most important point is not whether your
host content in that cloud provider’s location, you can reasonably
website or online content is up and running. Instead, what’s most
assume performance for users of that provider in that city will
important is whether your website or content is reachable (i.e.,
suffer. You can set up and run your own RUM or synthetic mea-
can a customer connect via her local internet service provider?).
surements or use a vendor who specializes in such measurement.
Optimizing the customer experience can be challenging, since
And don’t limit yourself to cloud providers; you should be mea-
14
of 17
Measuring performance across the
network is the
first step to help
mitigate issues for
your customers.
eGuide
IDG Enterprise is
suring performance to anywhere you have content, such as CDNs
and hosted data centers.
COMPUTERWORLD
3. “You can make your systems in the
clouds self-healing.”
So once you have the network performance data and are finding
2. “You can use the data to provide
predictive analytics.”
patterns in the data, what can you do? There are multiple ways
Once you have enough network performance data to the places
an obvious action is shifting traffic from a poorly performing
where your content and applications live, you can explore to find
path to a better one when you detect a degradation or outage.
patterns and make predictions. Does performance change on a
Having worked for a long time with and around DNS, I need to
regular basis? You might be surprised at how often this is true.
point out that using DNS answers “steered” by performance
Many systems used by humans follow a diurnal pattern (showing
data is a simple and effective way to route your users to the
the same changes every day). For example, the end of the busi-
best-performing path.
ness day on the U.S. East Coast is evening in Europe and early
to take action. Since we’re talking about network performance,
So start measuring—not just within the cloud, but the entire
morning when people are waking up in Asia, and this busy time
network path starting with the users. Once you have the data,
can affect traffic and change performance.
you know. And once you know, you can take action.
15
of 17
eGuide
IDG Enterprise is
COMPUTERWORLD
The internet isn’t ready for really big news
A worldwide, globally gripping news event would likely crash the major
media providers, sending us back to our TVs and radios.
BY PAUL VENEZIA, INFOWORLD | One thing you can say for
that content is sent to me and only me as it streams. It may be
traditional broadcast media: They scale really well. If you put an
cached somewhere along the way, but ultimately, that stream is
analog signal on the air or on a wire with enough repeaters and
unicast and not shared. Also, the content provider must accom-
amplifiers, it will serve every client that connects. That’s not the
modate the bandwidth required for that stream, as well as the
case with most of the network world, unfortunately. Sure we have
resources necessary to deliver it.
multicast, but that’s not on an internet scale—and the internet is
where the problems lie.
First, let’s define multicast as used in IP networks. This is a
Clearly, this isn’t usually a problem. With a suitable broadband connection on a normal day, accessing content around the
internet is a relatively stable and consistent experience, depend-
method by which a single source stream can be accessed by
ing on how adept the provider of that content may be in actually
multiple clients simultaneously, without increasing the load on
delivering the content.
the source itself. Thus, this functions much like an analog broad-
However, if it’s not a normal day, things go south quickly—
cast: You have a single source that a client can connect to at any
and how far south they might possibly go, I don’t think we
time. The downside is that the client is a silent subscriber of the
know. We have never truly seen the impact on our modern
content and cannot control the stream; there’s no rewinding or
broadcast infrastructure of an unexpected event of worldwide
restarting on a per-client basis. This is content broadcast over IP,
significance. I would guess that the internet itself would be
and it’s what television networks use to distribute video streams
fine, but the content providers would get crushed, which could
through their networks, financial institutions to receive stock
potentially lead to a cascade of events that effectively amount
quotes, and so forth.
to an internet media blackout.
On the other hand, the world is rapidly moving to a demand
I clearly recall the events of Sept. 11, 2001. I had a huge net-
model—the younger generations are already there—where
work forklift overhaul scheduled for that day, and as we started,
streaming content is controlled by the client, and the client forms
the world turned sideways. People gathered in front of the only
a one-to-one connection to the content source, versus multicast’s
available television and stayed there for hours. Few people, if
one-to-many approach. If I’m streaming video from a news site,
any, were turning to news websites for 911 updates, and certainly
16
of 17
If it’s not a normal day,
things can go south
quickly—and how far
south they might possibly go, I don’t think we
know. We have never
truly seen the impact
on our modern broadcast infrastructure of
an unexpected event of
worldwide significance.
eGuide
IDG Enterprise is
COMPUTERWORLD
none were loading information on their mobile phones. Univer-
television, and certainly broadcast television would be available.
sally, the terrible events of the day were carried by broadcast
However, the communications resources underpinning the pro-
television and radio.
duction of content distributed over those mediums would likely
The world is a vastly different place today, geopolitically and
be sluggish, if not also down hard. We would effectively return
technologically. An event of similar magnitude would gather a
to the days of 2001, and information would spread mostly via
suitably massive audience, but they would not immediately turn
broadcast radio and television.
on their television. Rather, they would open an app or browser
We have built a unicast-centric communications infrastructure
on their phone, tablet or computer. Those events would not be
in order to deliver astounding functions and services tailored to
broadcast. Instead, they’d be streamed on-demand—and those
the infinite needs of internet users. These infrastructures work
streams would eventually fail.
extremely well when the world is normal, but when the world
The internet and internet services work on economies of scale.
Large websites function with the knowledge that they will have
tilts, that model may collapse under the weight, especially when
only a few large companies provide most of these services.
peaks and valleys of usage, and if they have 10 million users,
We need to hope that we have enough time to build out and
no more than perhaps 1 million will be actually engaged with
spread around the underlying resources to a level where this isn’t
the site at any one time. Never would all 10 million attempt to
a threat. We need more competition and broader
connect at once. And with the advent of adaptive cloud services,
dispersion of internet media resources around
large sites can spin resources up and down to handle the peaks
the globe. Until then, keep a set of
and valleys.
rabbit ears and an FM radio
But a singular event that captures the attention of almost
every connected human on Earth is a spike we’ve never seen. It
would produce a resource load on large news sites, aggregators
and discussion forums like never before. And now that we have
these fantastic cloud services, many of those sites and services
will be hosted with the same providers, in the same data centers,
all vying for finite resources at the same time. If the event and attention is big enough, it could take down entire providers, which
would in turn pull down unrelated sites.
The end result would look like a complete internet blackout,
even if the actual damage were the loss of a few huge cloud
CDNs and providers like AWS. At that point, assuming the cable
providers can still manage their systems, we might have cable