Beruflich Dokumente
Kultur Dokumente
Welcome to The State of the State a special report from Light Reading and
Heavy Reading providing original insights into the changes taking place as
CSPs plan their migration to next-generation communications infrastructure.
We know that the entire communications industry is entering a period of
incredible disruption one driven by virtualization, open-source standards,
white-box deployments and OTT service delivery.
The information in this package provides both qualitative and quantitative
analysis of how these technology change agents are impacting CSPs,
including changes to the way they are planning their network configuration,
the impact to their investment and supplier choices and how service
providers specific procurement criteria is changing in the face of new
service requirements.
As Patrick Donegan, Heavy Readings chief analyst, points out in his report,
New Requirements for Network Equipment Vendors, many of the attributes
of next-gen CSP networks already exist in high-end cloudified enterprise
networks. This is prompting CSPs to look for new attributes in their
equipment suppliers, including both IT and systems integration experience.
Our survey of more than 130 service providers contains some genuinely
fascinating insights into their plans. For example, we now know from the
service provider survey that more than half of all CSPs are still at the stage of
researching virtualization, but 100% of them know they have no choice other
than to make the move to NFV eventually.
The payoff for doing so will be agile, virtualized networks that allow them to
reinvent themselves as services innovators, building and launching profitable
new applications and services at will. And the pressure is on them to make
this move before their last-generation networks and business models are
rendered obsolete by new competitors. At the same time, our report shows
that CSPs clearly recognize the migration must be made in an orderly fashion,
and in a way that doesnt disrupt or destroy existing revenue streams.
Disruptors come in various sizes and forms and some from perhaps
unlikely places. Im hoping these findings start some very interesting
conversations this week.
Steve Saunders
Founder & CEO, Light Reading
17%
Proof -of-concept
(PoC) stage
54%
Early stages of
learning about
NFV (pre-PoC)
19%
In trials
10%
Already
deploying NFV
Source: n=79 service providers. Heavy Reading Survey on Carrier SDN, December 2014
0%
Nevertheless, the outlook of some of these leaders shows the level of commitment they have to this transformation
and just how driven they are to execute on it as rapidly as possible. As shown in Figure 2 (below), for example,
AT&Ts CTO, John Donovan, told the Mobile World Congress event in Barcelona in March 2015 about AT&Ts plans for
NFV. He stated that AT&T plans to migrate 5 percent of its 150 relevant network functions from a classic hardwarebased architecture to VNFs running on virtualized hardware by the end of 2015. He foresees this rising to no less
than 75 percent of these network functions by 2020. Telefonica has comparable objectives, targeting being able to
run 30 percent of its new deployments on its virtualized UNICA execution environment by 2016.
Figure 2: AT&Ts Plans for Virtualizing 150 Network Functions
150
130
110
90
70
50
30
10
-10
2014
2015
2016
2017
2018
2019
2020
In the sections that follow we identify some of the critical capabilities and competencies that vendors will need
to demonstrate if they are to succeed in capturing a share of this emerging opportunity. At first glance, these
requirements have a very familiar, almost motherhood-and-apple-pie, ring to them. But this paper will soon dispel
that impression. What each of the following terms means from a CSP perspective in the modern era, and how the
requirements associated with each one differs fundamentally from what they have meant to CSPs up until now, is
then discussed in detail throughout the remainder of this paper.
Modularity
Openness
Innovation
Availability
Analytics
Network Integration
Evaluating, buying and deploying modular solutions from vendors is one of the core competen-cies of planners,
engineers and operations teams in the CSPs not to mention their procurement teams. Hence, designing products
that fulfil changing requirements around modularity is one of the core competences of the vendors serving them.
As well as hardware-based modularity, CSPs are increasingly looking for software-driven modu-larity because it is
much faster to add functions or capacity as software modules rather than as hardware modules. Software modules
can be instantly and programmatically on-boarded rather than needing a long procurement and (physical) installation
cycle. In itself, however, even software-based modularity doesnt protect a CSP from being locked-in to a specific
vendors product line. After all, its possible to have a modular product and still be locked into a particular vendor.
As discussed further below, software-driven modularity also needs to be supported by openness i.e., the
functionality must be clearly scoped and exposed through an open (i.e., published) application programming
interface (API). This allows the functionality to be integrated with other modules, not necessarily from the same
vendor. Only if these APIs are standard will the modules that are written to them be truly plug and play. And this
will help drive the cost reductions and service agility that the CSPs are targeting.
While this is undoubtedly the direction that CSPs need their vendors to go in to drive service agility, there are
also aspects of opening up APIs to third parties that are contentious both for vendors and for the CSPs. An
important corollary of investing in open APIs is therefore that the right security logic has to be built into the API
from the outset so as to protect against abuse. Depending on the use case, additional layers of security can also be
considered, such as a monitoring solution on the interface or a means of enforcing specific access rights for specific
third parties.
A Commitment to Open Source Software
Vendors need to demonstrate a commitment to open source software as they position them-selves for CSP business.
OpenStack is perhaps the most obvious example of an open source software community that is featuring at the
heart of the transformation toward a more software-centric network architecture and operating model for CSPs. It
is now the fastest-growing open source project in history, involving thousands of developers keen to create an open
source alternative to commercial programmable infrastructure managers, such as AWS and VMware. With so much
support within the CSP community for leveraging OpenStack for NFV, it is now rapidly becoming the clear favorite of
all possible candidates for the virtualized infrastructure manager (VIM) in the NFV management and orchestration
(MANO) stack.
What CSPs like about open source software is that it transforms their relationship with their vendors, making the
relationship a lot more transparent and further loosening the vendors grip on the CSPs network infrastructure.
Traditionally, telecom equipment vendors have built product using proprietary software. Those same products
may have been deployed across multiple other CSP networks, but many of their individual CSP customers have also
invested in proprietary extensions to that code either to accommodate unique aspects of the network or local
market requirements or for competitive differentiation.
This model has served CSPs well enough for the last 20 years, but its no longer fit for purpose. This is because it
results in the vendor developing a monopoly-like position in the network, rendering it ever more difficult to displace
and ever more difficult to negotiate pricing with, with every additional proprietary extension to the code that is added.
In this traditional model, of course, the vendors software starts out as proprietary and remains proprietary until it
reaches end of life. Throughout that time, the CSP is entirely reliant on that vendor extending its software roadmap in
a direction that complies with the companys strategy. They are also reliant on that vendor pricing those proprietary
extensions reasonably in the absence of any direct competition from other vendors, which just isnt practicable.
The only alternative to this unhealthy dependency is usually the threat of a large scale rip and replace exercise. The
problem with that is that it can sometimes be very expensive and is always very disruptive to the CSPs operations.
The open source model is wholly different in that software extensions developed by vendors get integrated into the
trunk code by the community and become open standard over time. So what Vendor A brings to the open source
code can be freely replicated by Vendor B down the line. This is how open source development ensures openness
with respect to the long-term availability at reasonable cost of software developed by vendors that use it. Because
community edition code can be freely downloaded, its worth noting that open source software isnt truly free
to the CSP. There will always be a cost involved in implementing and maintaining it. Users pay either for in-house
expertise to do this or for a vendor implementation that guarantees a reliable, supported code base.
Because of the transparency of the open source community, CSPs are also assured of unparal-leled, real-time insight
and proof-points of a vendors long term strategic intent as regards their software development roadmap. Open
source software also gives them the opportunity to see and experiment with code early in the process. This allows
them not only to see the quality of the code, but also to anticipate potential integration challenges early on, which
helps drive service agility.
Due to the transparent way in which the contributions of participants into an open source devel-oper community
can be viewed, CSPs can monitor the extent to which specific vendors are truly committing themselves to the open
source model rather than just offering token participation to give the appearance of a commitment.
There is a world of difference between a vendor that merely downloads and scripts in open source software, while
making a few contributions, and another vendor that contributes substan-tially to the common source pool. One
merely demonstrates a willingness to use whats there, whereas the other demonstrates greater willingness to
invest a substantial part of its own code therefore its whole strategic direction to the open source model with the
result that whatever the vendor contributes will be re-usable by others over time.
single application or every single service running across the CSPs infrastructure. CSPs need to get comfortable
with making responsible compromises on performance or availability metrics in the case of those applications and
services that dont need them.
Consistent with that, they need to learn how to be able to isolate those applications and services from others
that do require conventional standards of performance and availability. Unless CSPs make that leap from the
well-entrenched telco tradition of universal availability and performance requirements to requirements that are
tailored per service or per application, they will fail to capture all of the opportunities of the shift to softwarecentric networking. The onus is also on vendors to help CSPs differentiate those use cases that do require five-nines
availability from those that dont; for example, where a VNF has been architected in a highly available way or where
a lag in performance wont affect customer experience significantly, such as in the case of a free voice service.
As depicted in Figure 3, in the coming years, CSP networks will become less reliant on platform hardware and more
reliant on both platform software and network application software to meet their requirements for network or
service availability. The traditional approach with network equipment used in the CSP infrastructure has been to rely
on availability being guaranteed in terms of approaches such as 1+1 hot standby of the chassis; 1+N failover so that
a pool of hardware platforms picks up the load from any platform that fails; and card-based resiliency so that if a
card fails another card picks up its traffic instantly.
Figure 3: The Road From Hardware-Defined to Software-Defined Availability
Classic/Legacy
Proprietary network
application SW
(All Vendor A)
Vendor Bs VNFs
Ecosystem partner
of Vendor C
SL A provided by Vendor
A or a n int eg ra t or
S L A p rov i d ed by an d
fo r - ecosy st em par t n er s
Proprietary hardware
platform
(Vendor A)
Reliability is HW driven
S L As are s u p po r t ed
fo r al l V N Fs
As shown in Figure 3, the conventional model in the CSP environment assumes a hardware-driven SLA guaranteed
either by the vendor, by the vendor acting as a broader network integrator or by an independent integrator. Theres
no question that standard servers have nowhere near the internal redundancy and failover capabilities built into
them that purpose-built telecom equipment have. So to capture the cost saving and revenue generating potential of
SDN and NFV, CSPs have to embrace new ways of achieving traditional objectives, albeit in a way that allows more
granular variability in the unique performance attributes required to support different applications and services.
10
What that requires of vendors is to design network platforms and applications that evolve so that they no longer
rely on the hardware providing the reliability but instead are designed to assume the probability of hardware
failures and perform failover in software instead. Hence when a server fails, the cloud management software is
intelligent enough to determine that the network function that was running on it just gets moved onto another
platform elsewhere in the cloud quickly enough to maintain network or service availability targets. Consistent
with this, section 4.2, Network Function Virtualisation Resiliency Objectives, of the ETSI Group Specification NFV
Release 001 V1.1.1 published in January 2015, states: The key objective is to ensure service continuity, rather than
focusing on platform availability.
Software-Based Failover Isnt New
Software-based failover isnt new and is widely used in the IT environment. The unique challenge in the CSP
environment, however, is that most CSP applications are stateful so if there is a failure in failover, connections
are lost together with other state information, such as whether the user is logged in and where they are. There
is certainly still a lot of work to be done here, but there is little doubt that this is the direction CSP networks are
wanting to head in where assuring availabil-ity is concerned. Hence there is also little doubt also that vendors that
demonstrate strong commitment to delivering failover in software either through their own development activity
or by bringing in high availability software modules from third parties will find favor with CSPs.
Uncertainty does remain with respect to the optimal balance between extending the capability of network
application software to support failover on the one hand and extending the capability of network platform software
on the other. Work in both areas is clearly required, but the jury is out as to exactly where that balance will fall.
There is also uncertainty with respect to how quickly, and in what stages, different CSPs are going to be able to
migrate their approach to availability from a highly hardware-centric to a highly software-centric model.
The procurement teams in the CSPs are used to performing quality assurance on network equipment according to
their own conventional model. They must undergo a profound overhaul in the way that they evaluate and approve
vendor solutions from an availability or reliability stand-point in order to become comfortable with a highly
software-centric model.
Figure 3 points to an intermediary phase in the evolution of the NFV roadmap, whereby unique ecosystem partners
join forces to guarantee performance of one anothers products, such as one vendors VNF operating well on another
vendors NFVI. This is largely in accordance with much the same principles as conventional integrator models in the
CSP space, albeit in a far more modular, multi-vendor environment.
We define this phase as state of the art because we see evidence of this approach taking hold in the market today.
We view it as a very important interim phase of the markets development, pending a future, fully mature state of
the market, when carrier-grade requirements, including availability, have been fully developed across the vendor
ecosystem and any and all applications can deliver good performance on any infrastructure.
11
Customer analytics are equally important. They enable better understanding of what applications and services
customers are using as a means of automating the generation of customer profiles as a platform for offering them
new capabilities that are customized to their unique behaviors and requirements. Vendors that demonstrate an
ability to appropriately compose the right kind of advanced analytics either in house or from third parties into
their products will also be well positioned for emerging CSP requirements.
43%
34%
19%
2%
2%
Source: Heavy Reading survey of 57 mobile operators on virtual EPC, October 2014
As shown in Figure 4 (above), there is strong Heavy Reading survey evidence that CSPs are also very interested in
having an independent, specialist integrator bring together the mix of hardware and software components in the
emerging software-oriented network. In Heavy Readings October 2014 survey on virtual EPC, 43 percent of CSP
respondents stated that they will likely prefer a mix of vendor solutions using a specialist integrator to bring them
all together, while 53 percent had in mind carrying out the integration role themselves. Tellingly, only 4 percent of
respondents have in mind a single-vendor solution.
12
Heavy Reading believes that the role of network integrator will be highly sought after in the vendor community. We
expect many of the Leading Lights in telecom and IT worlds to prioritize the network integrator as a key strategic
platform for their own future growth in the networking hardware and software businesses. There arent many
vendors that can demonstrate core competencies in building cloudified enterprise IT network, as well as capabilities
in the telecom networking space. Those that can will be well positioned for this type of integrator business.
Summary
Heavy Reading believes that the criteria against which vendor selections are made in the CSP environment are
undergoing a significant transformation. While many of the same terms will continue to be used, what these terms
actually mean and the requirements for how vendors will deliver on them is becoming markedly different.
Vendors that support modularity, openness, innovation and availability according to the increas-ingly softwaredefined requirements of CSPs as already driven in the IT market will succeed. Those that dont, or those that are
behind the curve, will lose out. Analytics will play an increas-ingly important role, too. And we believe that the role
of network integrator will be highly sought after among vendors because of the strategic relationship it will bring
with CSP customers in navigating both the challenges and the opportunities in front of them.
About HP
HPs objective is to help CSPs thrive in this disruptive environment by accelerating their journey to NFV. Its OpenNFV
Program which includes an open standardsbased NFV reference archi-tecture, HP OpenNFV Labs, and the HP
OpenNFV partner ecosystem, helps CSPs speed interoperability and time-to-market of new services.
One of the key tenets of the OpenNFV architecture is that its based on open standards and leverages open source
technology projects such as OpenStack and OpenDaylight (ODL). This approach gives other ecosystem players the
ability to bring in new innovations. We do not believe that NFV is an environment where one vendor will do it all.
For more information follow @hpnfv and visit www.hp.com/go/nfv.
13
14
1. How much have your companys technology procurement processes changed over the
past three years?
9%
Value
Percent
Count
A lot
46.4%
64
A little
44.2%
61
Not at all
9.4%
13
Total
46%
138
44%
2. How has the importance of these criteria changed in your companys technology
procurement decision-making over the past three years?
Has become more
important
Hardware performance
(capacity, data rates)
84
62.7 %
23
17.2 %
27
20.1 %
134
Reliability
(MBTF, SLAs)
83
62.4 %
18
13.5 %
32
24.1 %
133
Software agility
(virtualization, opensource support)
91
70.0 %
17
13.1 %
22
16.9 %
130
Delivery and
integration timeframe
89
68.5 %
17
13.1 %
24
18.5 %
130
Energy consumption
79
59.8 %
22
16.7 %
31
23.5 %
132
Physical size of
deployed hardware
60
46.2 %
26
20.0 %
44
33.8 %
130
64.3 %
16
12.4 %
30
23.3 %
129
Interoperability with
technology from other
suppliers
83
15
3. How has the level of involvement of the following C-level executives changed in your
companys technology procurement process over the past three years?
Has become more
involved
No change in
involvement
Responses
CEO
31
23.3 %
25
18.8 %
77
57.9 %
133
CFO
43
31.9 %
23
17.0 %
69
51.1 %
135
CIO
61
45.5 %
17
12.7 %
56
41.8 %
134
CMO
36
27.1 %
26
19.5 %
71
53.4 %
133
CTO
73
53.7 %
17
12.5 %
46
33.8 %
136
4. Overall, how do you expect your companys technology spending to change over the next
three years?
Value
Percent
Count
44
27.0%
37
31.4%
43
5.1%
4.4%
Total
5%
4%
32%
31%
137
27%
5. How do you expect your companys spending on hardware to change over the next
three years?
Value
Percent
Count
34.3%
47
38.7%
53
27.0%
37
Total
27%
34%
137
39%
16
6. How do you expect your companys spending on software to change over the next
three years?
Value
Percent
92
11.0%
15
21.9%
30
Total
22%
Count
11%
67%
137
7. How will your companys spending on the following change over the next three years?
Spending will
increase
Spending will
decrease
Responses
Proprietary hardware
41
31.1 %
57
43.2 %
34
25.8 %
132
White-box hardware
50
37.3 %
35
26.1 %
49
36.6 %
134
Proprietary software
39
29.1 %
53
39.6 %
42
31.3 %
134
Open-source software
84
62.2 %
19
14.1 %
32
23.7 %
135
SDN
84
63.6 %
17
12.9 %
31
23.5 %
132
Virtualization
(VNFs, NFV)
99
76.2 %
6.2 %
23
17.7 %
130
Security
102
76.7 %
6.8 %
22
16.5 %
133
Analytics
84
64.6 %
15
11.5 %
31
23.8 %
130
17
8. How will your companys use of systems integrators change over the next three years?
Value
Percent
Count
37.2%
51
19.7%
27
34.3%
47
8.8%
12
Total
9%
37%
34%
20%
137
Percent
Count
71.3%
97
28.7%
39
Total
29%
71%
136
10. How do you expect your companys internal technology expertise to change over the
next three years?
Value
Percent
Count
24.1%
33
48
56
Total
18
40.9%
137
24%
41%
35%
Percent
Count
14.4%
20
25.9%
36
40.3%
56
11.5%
16
2.2%
0.0%
5.8%
Total
14%
26%
6%
2%
12%
40%
137
Percent
Count
U.S.
70.5%
98
Canada
6.5%
Central/South America
(including Mexico & the
Caribbean)
5.0%
Western Europe
10.1%
14
Central/Eastern Europe
0.0%
Middle East
1.4%
Africa
0.0%
Asia/Pacific
(including Australia)
6.5%
Total
10%
6%
5%
6%
71%
139
19