Tag Archives: hp

HP Networking – an introduction part 1 – the company

Posted on by

Let’s start with HP. This post will be more about the history of HP Networking and where I see its advantages lie. Even though I will touch upon some of the important technologies HP offers with their switching portfolio, I won’t go into much depth until my next post.

First, a brief history (go to http://www.hp.com/networking for the full story). HP is one of the key founders of modern networking along with companies like Cisco, IBM, Intel and 3com. Despite this, for a long time they were more focused on technologies than products.

As we moved into the 2000s, the networking world was somewhat static. 3Com – arguably the biggest player in enterprise Ethernet in the 90s – has pulled out of every market outside of Asia; IBM – the champions of Token Ring – had given up its networking portfolio and Intel had pretty much given up on making switches in favor of just producing NICs (along with CPUs, etc.). This left Cisco as the only big player and they built a dominant position in the business market that rivaled Novell’s position with NOSs in the early 90s.

HP was producing switches in the early 1990s but they weren’t really a core product; they were so low on the totem pole that they actually had to have their own separate R&D location from the rest of the company and reported into the printing group. There were even rumors around 2000 that the division was going to be sold off.

Fortunately, that didn’t happen and HP realised in the early 2000s that the advent of demands for technologies like VoIP, large high-speed wireless deployments and secure, managed networks offered an opportunity to build market share. This was particularly relevant as Cisco has been dominant so long that they weren’t really being pressed to introduce innovations, improvements in technology or reductions in cost to the market. HP jumped in offering all three of these and quickly moved to the #2 position as a network vendor globally; a position they have held since 2003 (and have been #1 in 10GbE networking since 2009).

HP’s earliest successes came in the education space (predominantly secondary schools). Their introduction of what is still the industry’s best switch warranty across their entire range of products coupled with their existing brand recognition (vis-a-vis their File Servers, PCs and Printers) and much lower cost than Cisco made them instantly attractive and they gained market dominance in many countries within this vertical.

Their next three major product leaps greatly related to the demand for VoIP and came out at roughly the same time. The first was the release of their 2600 series switches which provided all the technology necessary to deliver a robust VoIP deployment (i.e. PoE, QoS, VLANs and routing) at less than half the cost of Cisco. The second was the release of the ProVision 4 series ASIC with their 3500 and 5400 series switches which delivered GbE PoE, 10GbE and most of the advanced networking features that larger organisations demanded (such as dynamic routing, advanced security, etc.) with both superior performance and less power consumption than Cisco again at a far lower price. The last was their joint authoring (with Mitel) of the LLDP-MED standard which freed all switch and IP PABX vendors from the need to use Cisco’s proprietary protocols (CDP, CDP2) for the dynamic management of IP handets (a critical need as any engineer in the VoIP industry will attest to).

These technologies, coupled with the the reliability and price of their product led HP to build strong market share in many verticals but they still lacked a product set necessary to push into many large Enterprises. These clients demanded more complex technologies from their switches (even if they rarely if ever employed them); technologies like BGP4 routing; the ability to establish GRE tunnels and policy-based routing amongst others. HP’s focus had been on delivering fundamental networking services that the entire market required rather than on more complex services that tended to only be in use within data centers or large campuses.

In the late 2000s, Cisco introduced their UCS series of File Servers which signified another major industry change. Up to this time, Cisco hadn’t really paid much attention to HP as a networking player as their #2 status still put them a great distance away from #1 (market share charts at the time needed two scales for Networking vendors: 1 for Cisco and one for everyone else as otherwise they would not fit together!). Up until that point, Cisco and HP were strong partners in the DC space with Cisco switches being embedded in HP Modular Server arrays for both Ethernet and Storage networking. Cisco’s release of Enterprise-grade File Servers targeted directly at the DC was seen as a direct attack against what was one of HPs most important markets.

HP responded by making an aggressive move into the Enterprise networking space: they acquired the market leader of the 90s – 3Com. Throughout the 2000s, 3Com has quietly built up a switching and routing portfolio to nearly rival Cisco’s along with a major Asian market dominance. Further, 3Com’s switches offered all the advanced technologies that HP needed to directly target Cisco’s core customer base and unlike Cisco – who had become complacent in their switching R&D due to their lack of effective competition in their core markets of Europe and North America – 3Com (like HP) had been aggressively developing newer and better technologies to offer their customers.

The 3Com acquisition gave HP not only a network switch portfolio to rival Cisco but arguably switches that are class-for-class better (both in features and performance) than what Cisco has on offer. Cisco’s reliance on their venerable Catalyst class of switches (which were excellent switches but are now very outdated) meant that they had to do the unthinkable and acquire another switching company (Nexus) in order to keep up. Even this acquisition wasn’t a resounding success as the Nexus switches still don’t have the feature set of either HPs or even their own Catalyst switches which is one of the reasons why the uptake of Nexus has been so poor and why Cisco is keeping the Catalyst line alive despite its 15 year old roots. At the same time, HP extended their Lifetime switch warranty to all but the highest-end modular 3Com switches and still maintains a list price-point that can be as much as 50% lower than Cisco.

Some specific advantages that I have encountered in the enterprise space:

The ability to deliver 10GbE campus-grade MPLS/VPLS on sub-$10k switches (immensely important in University or large research center environments. Cisco can only deliver MPLS on their routers (and for 10GbE, BIG routers); C6500 or N7000 chassis at many times the price.

Switches with both deep packet buffers and bi-directional flow-control for iSCSI deployments (Cisco inexplicably removed flow-control when they they went from the 3650 to the 3750).

A Virtual Switch technology (IRF) common to the entire 3Com portfolio which allows up to 9 stackable or 4 modular switches to be fully consolidated into a single array which both allows all networking features to seamlessly span the array (such as link aggregation, routing, ACLs, etc) and can survive an outage of up to (n-(n-1)) switches when a “fat tree” deployment is used for uplink/server connectivity (e.g. in a stack of 8 switches, 7 could fail and connectivity and function would still be maintained even though performance would be degraded).

I’m not writing off Cisco. They still have the best Enterprise routers around, and they are definitely working on improving their switch portfolio but as the market stands right now, anyone weighing up features, reliability and price without paying attention to “brand recognition” would find it very hard to choose Cisco over HP for just about any size switch deployment.

About these ads

Starting a new series

Posted on by

Hi all,

I’ve got my next general post entitled “Where have all the Integrators gone?” nearly ready to go but I’m going to start a new series that is as much training as it is Blog based around the vendors’ kit that I most like to use and endorse. Most of my last 15 years in the IT field have been spent working around Networking and Unified Communications so they are the areas I’m going to focus on with an emphasis on kit that I like and why I like it. If you would like to know why I use and endorse the following vendors:

  • HP
  • Mitel
  • SonicWall
  • Fortinet
  • Riverbed
  • F5
  • MS Exchange Unified Messaging
  • Procera

Keep watching. Where practicable, I will also post some training information and videos as well; at this stage I only plan to do this for HP (since I’ve already written a course syllabus and I have most of the kit necessary to show demonstrations) but we’ll see how it goes…

VoIP over Wireless Networking – some things you need to know

Posted on by

Hi all,

I’ll get back to The Cloud soon, I have my next post mostly written but it’s a very complicated topic; and what’s not too hard to deliver as a presentation is proving far more difficult to put into a blog post. I’m going to start seeding these with diagrams soon, which should make things easier for me. For now, I’m going to write another post based again on a request – this time from the IT Manager of a major Sydney school (I’ll ask his permission before I publish his name) which is to explain how to get his VoIP working better over his Wireless network.

The short answer is that VoIP will never be ideal over WiFi (for the techies and purist out there, I know that it’s 802.11 networking and not WiFi but that’s what most people think of, and what most labels say). There are several reasons for this that I am going to cover in this post.

The main issue with VoIP over WiFi is that Wireless Networking (the 802.1<abgn> standard for the techies out there) is a half-duplex standard which means that a device can either transmit or receive at any given time (see http://en.wikipedia.org/wiki/Half-duplex#Half-duplex for a fuller explanation/definition). This provides a direct conflict with Voice conversations which tend to be bi-directional. The technology is kind of like trying to use a phone as a walky-talky where you have to hold the button down to talk and let go of it to listen. Wifi has sufficient bandwidth to cope with this somewhat, but it is far from an ideal medium (whereas all wired Ethernet networks other than the original 10Mbps are Full Duplex and hence better suited to Voice (even Wired networking has some issues with VoIP but that’s an easier topic I will cover in a future post).

Another issue is that the majority of Wireless telephony devices operate using legacy protocols that operate at very low data rates (as of 2009, commercial-grade 802.1g (54Mpbs) handsets were only starting to be introduced just as the 802.1n (up to 600Mbps) standard was approaching ratification). There are a few reasons for this. Firstly, VoIP needs very little bandwidth to operate. Most VoIP systems utilise SIP (Session Initiated Protocol) with G.711 as a streaming CODEC which consumes between 97-105Kbps depending on which VoIP vendor you are talking to. This is far less than the 1Mbps that the lowest level of 802.11b Wireless allows for an the rule for Wireless is that the lower the speed, the greater the range so WiFi phones tend to request/demand very low bandwidth from your Wireless Network. Far from being a good thing, this is a major problem as Wireless is a shared medium which means that bandwidth available (22MHz per channel) is all you get amongst everyone connected to that Access Point. That means that a WiFi phone connecting a 1Mbps consumes just as much of your Wireless Network as a laptop connected at 150Mbps with 802.11n. A Wireless Radio (NB: a Wireless Access Point can contain one or more radios; most APs contain only one radio although 2 or more radios in an AP is becoming more common) can only operate at a single channel or frequency at any given time. This means that most WiFi IP phone will not only starve your network of performance whilst they are connected, but they will also cause constant contention (since they are likely the only low-speed devices on the network).

Our next issue is frequency contention. Wireless Networks run at divisions of either the 2.4GHz  or 5GHz spectra. Virtually every VoIP phone (and most other WiFi devices) use 2.4GhZ. There are three main reasons for this: Range, Price and Power. Lower frequency signals carry further than higher frequencies (which is why mobile phones tend to operate at as low as 850MHz since Cellular towers are kind of expensive) which makes 2.4GHz far less expensive to deploy (less radios/access points needed) than 5GHz. In order to achieve any decent range, 5GHz devices need to consume a lot more power which can take a rough toll on a WiFi Phone (or Laptop)’s battery. There are two big problems with the 2.4GHz spectrum. Firstly, the allowed spectrum goes only from 2.401 to 2.483MHz (Japan allows a bit more). Since Wifi consumes 22MHz per channel, that means that there are only 3 non-overlapping channels in the 2.4GHz band (or only 1 non-overlapping channel if you use 802.11n which consumes twice the bandwidth)! In other words, if you are running 802.11b Wireless phones, the maximum data rate you can provide to any given area is only 33Mbps (11Mbps x 3 channels) or actually half that once Wireless overheads are taken into account. That’s the bad news, here’s the worse news: since 2.4GHz is an unregulated band, anyone can use it and everyone does. Microwave ovens, baby monitors, cordless phones, cordless computer mice, wireless headphones and dozens of other devices all gleefully use this area of the spectrum and there’s nothing you can do about it. I once did a Wireless Survey at a University where the PIR sensors that turned the lights on in the lecture theaters were operating (and absolutely blasting) at 2.450GHz, completely locking out a big chunk of their Wireless just so someone doesn’t have to flick a light switch when they walk into the room. It doesn’t matter where you go, you will always get frequency contention with 2.4GHz. The 5GHz band is far less utilised but for reasons mentioned above is less suitable

Now we move on to the biggest bugbear of all VoIP (not just Wireless) which is Quality of Service. Nothing you run on your network demands more “real time” bandwidth than voice. Voice quality can be horribly distorted if any packet loss, delay or jitter is present on your network and that’s particularly bad news for WiFi (which is a slow, shared, contended media as described above). To deal with these issues, VoIP traffic needs to be granted absolute precedence on your Wireless network over any other type of traffic or it will be very poor. This is one of the reasons that VoIP over Wireless tends to starve your network’s performance (as described above). Until around 2007 (when the WME extensions to 802.11e were released) there was no standards-based method of providing QoS for Wifi which made deployments very problematic. One company – Spectralink – released a method of providing QoS for Voice over WiFi called SVP (Spectralink Voice Protocol) which depended on a dedicated appliance to govern the traffic in conjunction with compatible access points which many vendors (including Cisco, HP, Motorola, Proxim, etc.) subscribed to. Whilst Spectralink worked reasonably well, it was expensive, difficult to configure and allowed only the most basic (and easily hacked) encryption to be used on the network. With the release of standards for Wireless Media Extensions (or WMM for Wireless MultiMedia as its more commonly referred to), SVP became obsolete but many AP vendors still support it as there are still many older SVP-based IP handsets in use (mainly due to their high cost of replacement).

Those are the main issues with VoIP over WiFi, there are others but those are enough for you to see that it can be a challenging solution to deploy.

I’m sure that by now the gist of this post must seem to be don’t deploy VoIP over WiFi but nothing can be further from the truth. This technology is very useful and can be successfully deployed if the right strategies are used.

Before I get into how to do it right, I’m going to digress into the single biggest issue I encounter with not just VoIP over WiFi but Wireless Networking in general: More is better. I’m going use use bold, underline, italics and caps for the following: NEVER EVER BELIEVE SOMEONE THAT TELLS YOU THAT YOU CAN OVERCOME YOUR WIRELESS ISSUES BY FLOODING YOUR SITE WITH ACCESS POINTS OR RADIOS!!!

I use so much emphasis because I encounter this so often and it is upsetting to see organisations being ripped off by forcing them to buy needless amounts of equipment and spend huge amounts of money (it’s not just the APs, it’s the cabling and the data ports they consume along with their licensing and management that really costs money) with the result being an environment that’s worse than they started with. Yes, worse! As detailed above, there is very limited spectra available for use and the more APs you have in any given area, the more that spectrum gets carved up and the less efficient the overall network becomes. This is a particularly big problem for VoIP over WiFi since people tend to be moving when they are on the phone; not only does a site “flooded” with APs tend to cause the phone to “radio-hop” (i.e. jump from radio to radio or AP to AP) excessively, the results of that hopping can cause havoc on the rest of the network (also as detailed above). Each WiFi phone will be like a little ghost roaming around the premises causing all sorts of issues wherever it goes. It doesn’t even need to have a call active since the phone (like any other Wireless device) is always “on” the network (how else could it receive a call?) unless it is switched off. If you aren’t sure if you have too many radios or APs, download a program like Netstumbler (http://www.netstumbler.com/) and run it at several locations around your site. If you see full coverage of the three non-overlapping channels (i.e. 1,6 and 11) from 3 distinct radios and maybe a small amount of signal in those ranges from other radios then you probably have the right number of devices. If you see the same channels being strongly presented by multiple devices then you have a problem.

What can you do about it?

OK, enough with the doom and gloom. Let’s move on to steps you can take to ensure decent VoIP over WiFi coverage. These are only rough guidelines to help you; doing this right really needs a trained Wireless Networking professional with the right tools and knowledge but a lot of people claiming to be that really don’t know what they are doing so see if they are following these steps and if not, you might want to consider how qualified they really are.

Do a Wireless Survey. Whether you are planning a new deployment or are having issue with your existing environment, a proper Wireless Survey is a must to work out optimal radio placement and potential blackspots. Top of the line survey tools (like “AirMagnet”) can run to over $10k but they will provide very exacting information regarding what you need to do to get your Wireless Network deployed right. Note that a Wireless Survey of a large campus can take a week or longer so expect this to be a fairly expensive exercise but if you have a lot of problems, or if you can afford it, it is worth it when done properly.

A follow-on to a Wireless Survey is investing in decent Wireless Management tools for your environment. Most major vendors offer pretty decent self-branded tools that will allow you to easily prepare “heat maps” of your Wireless Network as well as to detect “rogue” devices (i.e. APs that don’t belong to you but are transmitting in your space) as well as blackspots (where there is no coverage), hotspots (where there is heavy congestion) or areas where there is too much congestion. These same tools can often also help you “tune” your network (see below). If your vendor doesn’t offer the right tools or you are running a multi-vendor environment (more on this below as well), there are excellent 3rd party solutions (like Aeroscout http://www.aeroscout.com/) you can use.

Tune your Wireless Network properly. I made strong mention of the issues of too many APs above, but how many is too many? The answer is based on the amount of channel overlap you have. A lot of environments (most notably secondary and tertiary schools, where every student has a Wireless laptop or similar) really do need a lot of coverage. It is not uncommon to even see two radios per classroom in some schools and this is fine, as long as they are tuned right. Avoid the temptation to turn the gain right up on the antennas. If anything – in high-density environments – you need to turn them right down in order to make sure their coverage is as localised as possible. Some vendors’ APs are “self-tuning” but my experience is that this is usually less effective than taking the time to set the antennas properly and lock them that way. Use the tools I mention above to do this and review it at least once a year (or whenever there is a problem, or the Wireless Network is being expanded).

Move high-bandwidth devices to the 5GHz band only. These days, most laptop and tablet vendors offer 5GHz WiFi in all but their cheapest models. Not only is the 5GHz band less cluttered, it also has many more channels to use for overlapping coverage. The number of channels differs by country (due to local regulations) but Australia – which is where I live and allows an average number of channels – has 21 non-overlapping 20MHz channels (or 10 for 802.11n). This firstly allows you to deploy far more APs/Radios than 2.4GHz (which allows you to provide far more data) but will also pull most of your contentious data off that spectrum leaving far more available for Voice. The only devices that should be on your 2.4GHz network should be Smartphones (which still mostly use 2.4GHz due to power constraints) and VoIP WiFi phones. This may mean investing in 5GHz dongles for some legacy laptops but these can be obtained for under $50 each. This will also mean getting rid of any 2.4GHz wireless printers (or print servers) you may have deployed. Given how much data is consumed by printing, these are likely causing issues anyway!

Consider reserving a channel just for VoIP. This is a fairly draconian but effective measure for maximum VoIP over WiFi performance. Choose one of the three non-overlapping 2.4GHz channels and use it exclusively for VoIP, allowing data devices to use only the other two channels. This will guarantee that your VoIP usually has decent bandwidth available (subject to interference from non WiFi devices as detailed above; use the cleanest of the three channels for VoIP). If the previous strategy is employed, this will only reduce the amount of data available to Smartphone devices.

Go with a single-vendor, managed WiFi solution. Wireless Networks come in two general flavors: un-managed and managed. Un-managed networks are where each Access Point is configured (and runs) individually. Managed networks use a central controller to administer and propagate policies to all the APs. Managed networks allow technologies like fast-roaming (i.e. quick handover from one AP to another during times of movement or congestion) and Layer 3 roaming (roaming between different network subnets) as well as better overall network management and congestion mitigation. But…

Avoid “Captive-Portal”-based managed WiFi solutions. There are two ways that managed Wireless Networks function. One way (and still the most common) is to tunnel all the traffic from the AP through the controller before sending it to its ultimate destination. The other way is for the AP to receive policy from the controller but to send the traffic is receives directly to its destination. The former solution is a major reason for poor VoIP over WiFi since the controller itself becomes a major bottleneck point. If you already have a Captive-Portal System…

Consider a dedicated Wireless Network just for VoIP. This isn’t as expensive as it may seem. Remember that the lower the data rate, the greater the range. Whereas 11Mbps 802.11b has a nominal transmission range of around 30m (indoors with typical walls), 1Mbps 802.11b has a range of around 90m! This means you can deploy far fewer APs to attain the same coverage for VoIP as you ordinarily would need to use for data. Further, performance could be greatly optimised since not only could you reserve a channel for VoIP, you could also lock the data-rate of the AP to match that of phones. This should be a last resort but if VoIP over WiFi is really important to your organisation it may be the only solution.

 

There you have it, my view of what issues you can expect to encounter with VoIP over WiFi and some of the ways you can deal with them. I must again stress that this is an area where good expert assistance will make all the difference to a successful deployment.

I would like to make one final note, and this is more around security than performance. Remember that many WiFi handsets only support very basic WiFi encryption which is very easy to hack so make sure that you isolate your “Voice” network from everything else using appropriate security policies on your APs, switches and routers/firewalls. If you consider the confidentiality of your VoIP traffic to be important then make sure that you invest in devices that support strong encryption (such as WPA2) if you don’t want people listening in on your conversations.

ASIC or Merchant Silicon – HP Switches

Posted on by

Another re-post I’m afraid. I have a bad cold and can’t concentrate. the original post is at http://h30507.www3.hp.com/t5/HP-Networking/ProVision-Inside/ba-p/94329. This was my weigh-in on the debate over what was preferable to use in network switches following HP’s acquisition of 3COM (since HP has always focused on ASIC whilst 3COM was pretty much all MS). Here we go…

 

Whilst initially it might look like there might be some conflict at HP between merchant Silicon (i.e.. A-series and other 3Com acquisitions) and ASIC (i.e.. E-series or legacy ProCurve) I believe that a valid strategy exists for both.

In Data Centre space, standards are king. MPLS, BGP, OSPF, etc. Everything has to interoperate cleanly and it is unlikely in the extreme to encounter a mono-vendor environment when taking into account: switches, routers, firewalls, load balancers, wan accelerators, traffic shapers, etc. In this sort of environment, merchant silicon makes perfect sense for switches and routers. New technology is rarely introduced into the core of a DC environment; and the fundamentals change glacially slowly. I think the IPv6 (and its associated routing protocols) is the only technology likely to start becoming prevalent as a core component of DCs that is much less than 10 years old. With such tried and true technology, it makes sense to use less expensive and more standardised Merchant Silicon than investing time in developing customing ASICs that have features that will never be used.

Out on the edge, the reverse is true. VoIP, UC, Web 2.0 (3.0, 4.0…), and other emerging technologies along with the vast increase in component performance makes customised technology not only advantageous, but necessary. Most organisations these days are DC-centric but still require a number of technologies to be deployed at the edge/branch level including: PABX, Switching, Routing, Security, Wan Acceleration, Print Server, etc. Historically this has meant a fairly substantial wiring cupboard (or more likely a full server room/rack) to be implemented at each branch along with associated cooling and power redundancy. Using the ProVision-based E8200/E5400 most if not all these functions can be consolidated into a single, highly-available device. This offers both cost and efficiency savings and will certainly appeal to those interested in Green IT.

It’s for this reason that I’ve started to think of the ProVision switches as “Application Switches” with their own niche that other vendors are woefully behind. Unlike so-called Integrated Service devices which offer extremely cut-down versions of the services their are purported to support, the ProVision Application Switch architecture allows the deployment of multiple full-blown systems (often integrating other leading vendor technologies) into a ludicrously small platform.

Rather than diminish the importance of the legacy ProVision architecture, I believe that the introduction of the A-Series range has created a differentiation that has clarified the role of a product that is years in advance of its competitors.

Michael

What Network do I need for my Small Office

Posted on by

Small businessmen have real challenges when it comes to IT in their business. They all need it and most of them don’t understand it. What makes it worse is the amount of terrible advice they get ranging from recommending they spend far too much money to “saving a packet” and having an unreliable environment that causes losses due to down time. I thought I’d write a few posts about this topic starting with The Network.

The Network is what ties everything together in the office. It’s driven by that pizza-box shaped thing that all the blue cables with the over-sized phone connectors plug into (otherwise known as a switch which has Ethernet cables with RJ-45 connectors plugged into it). The Network is one of the most innocuous and yet vital elements of your environment since if it fails, everything breaks. Because of this, it’s worth investing in a quality device. On the positive side, a quality network switch should last your business 5-7 years. There are a few considerations that need to be taken into account with any switch that you purchase; they are:

  • Speed (Throughput)
  • Power (Power over Ethernet)
  • QoS (Quality of Service)
  • Redundancy
  • Maintenance
  • Let’s cover each of these

Speed

A made a point to a colleague recently that Gigabit Ethernet is a waste of money but I didn’t expand on that point. Most PCs support GbE as do most File Servers. Unfortunately, most IP Phones do not. The Digital PABX is dead; very few vendors still make them and they are starting to cost far more than newer IP PABXs (VoIP) to manufacture and hence buy. One of the benefits of VoIP is being able to “double hop” a cable; i.e. plug the phone into the wall (and back to the switch) and then plug your PC into your phone. The problem here is that the majority of IP Phones only use 100Mbps Ethernet (IP telephony only consumes about 100Kbps) so even if you have Gigabit Ethernet at your switch, you still only get 100 Mbps Ethernet to your desk. The next issue is your File Server since it’s probably plugged into your network at GbE as well-meaning that if you are in an office of – for example – 40 people then you are sharing that GbE at a 1:40 ratio during peak periods of the day (NB: 100Mbps switches usually include a few GbE ports specifically to connect to File Servers (or other switches) specifically so that the Server has more bandwidth than any one user can consume). To fix this issue requires either aggregating multiple GbE links into your Server (generally you can use up to 8 without much difficulty) or investing in 10GbE between your Server and your switch. The problem with the former solution is that is is messy (up to 8 cables) and consumes a lot of network ports. The problem with the latter solution is that it can be fairly expensive. For these reasons, I generally recommend against GbE. The reality is that unless you work with very large files all day (e.g. desktop publishing, architects, etc.) you really won’t notice the difference. This becomes even more relevant if a business starts migrating towards a Cloud Infrastructure – where much (or all) of the data and applications are hosted elsewhere. In this situation, the performance of the network becomes almost completely irrelevant since the limiting factor (i.e. bottleneck) becomes the business’ link to The Internet or Private Network that connects them to the Cloud. Few small businesses today have links of speeds greater than 2-4Mbps (or ADSL but since that only provides high-speed downloads, it is less than ideal for Cloud deployments) although some in heavily urban areas may have up to 10Mbps – still far shy of the 100Mbps network they probably have deployed.

There is one caveat here though which is the relatively new high-speed Wireless network (802.1n) which requires that you run it across Gigabit Ethernet since it allows for connection speeds greater than 100Mbps (up to 600Mbps is allowed for in the design). My general recommendation for most small businesses is thus to deploy 100Mbps to the desktop and Gigabit Ethernet to the File Servers and newer Wireless Access Points.

Power over Ethernet

As per the previous section, virtually all Telephony these days is moving towards VoIP which inevitably leads to a need for Power over Ethernet (PoE) networks to power the phones (unless you want a power brick at each desk). PoE can also provide power for other devices such as Wireless Access Points, IP surveillance cameras, etc. The original standard for PoE was for 48v DC @ 14.4w (12.8 delivered) this is enough for virtually any device short of a a Pan Tilt Zoom camera which needs more power to drive its motors. The more recent standard – PoE+ -  can deliver nearly 30w and is frequently all that is on offer these days. Since powering devices via DC over the Network is far more efficient than AC plugs, got with PoE+. NB: Some vendors are promoting even higher power delivery on their switches but these involve proprietary technology and are generally useless since there are few if any devices that  current draw that much current. There is a new standard being developed to deliver higher power with the theory that you will be able to charge laptop batteries or power Thin Clients across the Network but until that standard is ratified, stick with the current technology.

Quality of Service (QoS)

In over 10 years of working with VoIP and Unified Communications, the number one issue I have encountered is organisations’ not deployment appropriate QoS for their IP Phones. The argument is always something along the lines of “VoIP only uses 100Kbps and we have a 100Mbps/GbE network; we have loads of capacity, we don’t need QoS! This is completely wrong as the nature of Ethernet Networking is such that – by default – applications which demand most bandwidth (e.g. print jobs or large file copies) get the most bandwidth (see my upcoming article CSMA/CD is a sledgehammer for a more technical overview of why this is so). This means that a single large print job can starve (and that is the technical term) your IP phones of bandwidth and hence cause poor call quality or even dropouts. This is relevant for other light-bandwidth applications as well (things like Telnet in the past and many current cloud applications). It is for this reason that QoS was developed as it allows for time-sensitive traffic to gain priority on the network over bulkier but less sensitive traffic (e.g. a half-second delay in the delivery of an email is irrelevant but a half-second delay in a phone conversation is unacceptable). There are two broad types of QoS; L2 and L3. The technical details of these are beyond the scope of this post but you should ensure your business invests in a switch that supports both L2 and L3 QoS.

Redundancy

This is a bit of a sticky subject. On the one-hand, a network switch is a mostly solid-state device, with the only moving parts being the cooling fans (if any) which makes it a very robust device with a much lower fail-rate than File Server, Printers or Desktops. On the other hand, as mentioned previously, if the Network breaks you might as well send everyone home. Whilst there are fully redundant Network devices available (where they can survive the failure of any single component), these are generally very expensive and are designed for large deployments or data centers. Most Network vendors offer a good compromise through either partially redundant Switches (most often supplied with multiple Power Supplies and modular data ports) or Virtual Chassis (where multiple network switches are “stacked” to act as a single switch, and File Servers  can have links to each of them so that if a unit crashes, the File Server and at least part of the network will continue to function. Both approaches are quite solid; a redundant device has a far greater impact if it fails than does a Virtual Chassis but a redundant device is also somewhat less likely to fail. Choose one of these approaches if possible; avoid a single stand-alone device as this will leave the business quite vulnerable to a complete outage.

Maintenance

Maintenance in the IT world means the cost to maintain the warranty and software licensing of a device (rather than physical maintenance of something like an Air Conditioner). It is important that you maintain your device both for security and reliability – so that you can obtain a software patch if a problem is discovered with your switch’s software – and for recovery – so that you can rapidly replace a device that fails. The cost of maintenance can vary wildly from vendor to vendor. May offer “lifetime warranty” with a fine-print caveat that replacement could take up to 10 business days whilst the best offer “lifetime warranty” with Next Business Day replacement. The former offering is useless whilst the latter could still mean a day+ of lost business in the event of a failure. The appropriate maintenance for your Network should be the same as for your File Servers which is generally same day 4 hour response. Make sure that the vendor you buy your switch from offers this level of support and check its price; some vendors will charge you nearly as much (or more) for three years maintenance as the initial cost of the switch (so in its 5-7 year life you are effectively buying it three times).

I hope this helps you with your choice of networking device and I welcome any comments.