Total Immersion

It always seems quite ironic to me when I read how industry analysts are accused of ‘bigging up’ vendor offerings, when I and my peers seem to spend so much of our time resetting the expectations of over-optimistic marketeers. Indeed, without such a position, we would offer a far less useful service - on occasion I have been positively surprised that certain companies have wanted to work with us at all, given the utter trouncing we have given their products or how they are taking them, like Beanstalk Jack and his cow, ‘to market’. I should perhaps apologise (and I frequently do) for being so direct - we want people to get the best out of your technology, we really do, so we’d rather be straight with you.

As such, it can be quite a relief when something comes along that is so clearly, obviously useful to so many organisations. Like Internet Protocol (IP) address management, for example. I can’t confess to know the whole space in technical detail, but here’s the skinny from my perspective. It is a well-known fact that the number of devices that need an IP address to connect to the enterprise network, or indeed the Internet has rapidly outstripped the original numbering standard, of 32-bit addresses enabling a potential four thousand million addressable devices. Such things as Network Address Translation (where a local router/address server allocates IP addresses on an as-needed basis using a local subnet, and then translates between local addresses and a reduced subset of externally-visible addresses) have helped reduce the burden somewhat; as of course has the arrival of IPV6, which extends the number of addressable devices to 2^128 (a very big number).

However, a remaining issue is how to manage said pool of addresses. These days the number of required devices has increased dramatically, notably with the arrival of Voice over IP (VoIP) handsets, which are replacing traditional, analogue telephones. From an address management perspective, the Domain Name Service (DNS) protocol is the standard for allocating specific address ranges to specific subnets, but some organisations are ending up with a large number of DNS servers, which themselves have to be managed. The original protocols were never conceived to manage the address allocation, deallocation and reallocation process on such a scale - and don’t facilitate the cataloguing of what address belongs to which department (Microsoft Excel is a more used, but still inadequate tool). Theoretically, organisations could of course allocate addresses statically, once and for all - but all it takes is an office move (requiring a number of devices to move from one subnet to another) and all hell breaks loose.

So - IP addresses need managing, and existing mechanisms aren’t cutting the mustard. This is the breach into which are stepping organisations like BlueCat Networks (who I have spoken to), and Alcatel-Lucent, BT-DiamondIP and Crypton Computers (who I haven’t - but these chaps have) - essentially delivering management tools and distribution mechanisms that really can cope with such huge numbers of addresses and offer quite some respite to those managing the IP network. It is notable that, when I asked BlueCat whether I could speak to a customer, they jumped at the chance and before long I was speaking with Investor AB, a Swedish organisation.

On the call I learned little that was unexpected: yes, the problem existed and was real; yes, it was for the reasons I understood; and yes, the deployment of BlueCat’s address management solution had been a great help. What’s there not to like, I said as we finished the call. And yet, I was left feeling a little puzzled at the end of the call. Notably, whether by agreeing with the problem and solution, I was in some way implicated in yet another attempt to foist unnecessary technology on an unsuspecting public. Particularly in this case - where the solution itself resolves an indisputably technical problem.

But however we might like things to look, the problem does exist and so does the solution. Just as the invention of carpets required the subsequent creation of carpet cleaners, so can today’s overstretched networks benefit from address management. This won’t be a panacea for all ills - it never is, and it should go without saying that technology can never be more than a crutch to poor operational processes or bad managers. I could add a string of caveats at this point but I won’t - rather, I will acknowledge the fact that most network managers do have their heads screwed on pretty well, and defer to their ability to decide whether this would be an appropriate technology for them.

It is always dangerous to speculate, particularly in this industry, which sometimes seems to be more founded on speculation than practical reality. Consider, for example, Power over Ethernet (PoE) - essentially offering a way of delivering power through an Ethernet cable. Today, there are a multitude of different devices that can be attached to a network that – WiFi repeaters, video cameras and so on — whose location may not be near a power socket. It makes sense, therefore, that the wire used to connect the device to the network is also the wire supplying the power.

Where PoE has really come into its own, is with VoIP phones – telephone handsets that use a network-based infrastructure rather than a traditional PBX. Voice over IP handsets are exactly the kind of devices that can benefit from power over the net, just as old-fashioned analogue handsets are powered by the PBX. The alternative is to have a transformer next to every phone, which occupies a socket and is one more thing to go wrong.

The downside of PoE is, of course, in the “P”. I’ve written before about how hard it is for hardware vendors in general, and networking vendors in particular, to claim any sort of green credentials for their equipment. The fact that PoE is delivering power, makes it a bit of an anathema to green – particularly as the latest iteration of the standard enables more power, not less, to be delivered over the Ethernet ports. According to the marketing, such power increases are required to support the increasing complexity of VoIP handsets. Colour screens, bigger processors, more memory – all of these things will take their toll and become more of a draw on the corporate power supply. That’s all very well, but it’s not very green, is it?

On the surface, then, Power over Ethernet can hardly be held up as a poster child for green IT. That’s not necessarily the end of the story, however. Let’s consider some of the plans, and likely developments in the PoE space: not least that it may well become built into switches by default, rather than as an exception. From a systems architecture (and indeed, from a manufacturing) perspective, there is little difference between powered and unpowered Ethernet ports. One of the larger network vendors told me that the chances were most of their switches would build in PoE to all ports, at some point in the future.

In principle, that’s still not very green – but there’s more. There are no concrete examples yet, but vendors are also talking about incorporating power regulation directly into network switches: put simply, enabling the switch to regulate supply according to demand. It is not beyond the realms of possibility to imagine the automatic power-down of devices outside certain hours, or indeed, when no data signals were detected (pretty obvious for IP phones, for example). To take this one step further, it is within the realms of possibility to produce handsets that require only a trickle current when in standby mode – and which could signal their requirements to the switch.

Taking such thoughts to their logical extreme, would it be possible to furnish an entire building with a highly regulated, low-voltage, direct current power circuit based on flood wiring (that is, the networking sockets on the wall)? In principle yes – though indeed, there are a number of hoops to be jumped through first. Not only are there the technological hurdles such as the ones above, but also some basic truths, such as the fact that most network wall sockets are not actually enabled: they may connect to a patch panel somewhere, but this will not necessarily be connected to a switch.

All the same, while it may not yet be possible, there is certainly potential. Such a circuit might, for example, be able to replace the currently obligatory raft of telephone and PDA chargers that litter our offices – indeed, I discussed such a thing with one of the senior guys at network wiring specialist CommScope (who brought up the “not-all-ports-are-wired” issue – thanks Ian). Perhaps it might never happen, but it is often only in hindsight that we understand how technologies are to be used: in this case there has already been a precedent set with the charging potential of USB. Why not the same with the network? Such an infrastructure would be able to support a broader range of devices, far more straightforwardly than relying on the mains: as my colleague Tony Lock has pointed out, consider the efforts of the thin client vendors such as Wyse, who are bringing out devices with power requirements small enough to be powered by PoE alone.

Indeed, it can be dangerous to speculate. But equally, just as many technologies also have a downside, so there may be some upsides of PoE we are yet to experience. Just perhaps, and even taking into account the cost of manufacture, Power over Ethernet might just offer an opportunity for networking to demonstrate its green credentials at last.

As part of the briefing cycle for Aruba’s announced acquisition of Airwave Wireless, I had a very interesting conversation with Roger Hockaday, EMEA marketing director for Aruba. In part it was about the announcement, but it quickly turned (as these things do) to a discussion of the wider picture of wireless, and indeed wired network management. “Discussing the wider picture” can sometimes mean, for analysts, expressing poorly veiled disdain at the fact that a vendor has not taken things far enough - a bit like when the triumphant person comes into the room to demonstrate, after 6 months of hard graft, that he can now juggle with 3 balls, only to be shot down by some smart alec who says, “yes, but to do it properly, you’ll need to jugggle with 4.” Not that I ever would of course, and certainly not with this - because the challenge is not one that can be resolved in one go.

In this particular case, it is more like juggling with a ping pong ball, a meat hook and a chainsaw. Wireless networking protocols remain all over the place as the bubble-headed wonks of 802.11 land continue to squeeze yet more bandwidth, and indeed distance out of some highly unreliable physics; on top of the base protocols are build a number of security and management capabilities, which are supposed to be compatible, but sometimes don’t quite manage to integrate. While all the attention has been (laudably) on driving up bandwidth and resolving compatibility issues, the black hole remains centralised management, particularly for legacy products that were not built for remote configuration and monitoring.

Aruba’s acquisition signifies both the need to centrally manage the variety of wireless switches that are out there, and the resolve to do something about it. Having not researched this specifically I don’t know if this is down to latent demand or direct customer pressure, but it makes sense that organisations which have rolled out wireless access points on a more ad-hoc basis in the past, are now seeking to integrate their operation with the rest of their network management activities. And incidentally, it may be that Cisco have been able to offer a more integrated approach for a while - but it is a rare organisation that is a wall to wall Cisco shop, so the same issue arises.

While WLAN remote management might be able to bring wireless management into the same room as wired networks, this is still a step away from bringing both onto the same console (and I don’t mean through screen scraping, or “sure, we can do SNMP traps” type conversations). It is true that wireless network configuration and monitoring has different drivers to the wired equivalent: one will largely be to support roaming end-point devices running a limited set of functions, while the other needs to consider the entire network architecture; there will be different (e.g. security) policies applied to each, etcetera. However, more integrated management tools lead to more efficient (and therefore, less costly) management.

We should therefore see this acquisition, and the impetus behind it, as a flag along the road - an indication of where we are in this work in progress. From the end-user perspective, where (as we have seen in recent studies) people care little about how they are getting their bits, just that they can get them, networks and their management tools should be able to see not just wired and switched wireless, but also “mobile” protocols such as HSDPA as part of the same, managed network architecture - particulalrly when things like Unified Communications really start to tip, and picocells extend such protocols into the office. Now that, if and when it comes, will be juggling.