Pacific fiber and data operator Pacnet announced a new initiative today. They plan to convert their Cable Landing facilities into ‘Data Landing Stations’. Pacnet of operates an extensive submarine cable system in Asia and has such cable landing systems in Hong Kong, Singapore, Japan, Korea, Taiwan and the Philippines amongst others. Leveraging unused space to address the data center market is indeed an attractive concept. In the words of Pacnet’s Bill Barney:
“Over the past few years, we have seen exponential growth in the demand for data center hosting services and customized applications. Yet, the cost of power, real estate, maintenance and connectivity makes it challenging for existing players to upgrade and new players to enter into the data center space. Pacnet’s cable landing facilities across Asia are equipped with massive power, space and capacity, therefore, it is a natural step for us to move forward in building DLS facilities to support immediate demands in the marketplace.”
And yet, I’m still scratching my head a bit here. Just how big are these cable landing stations? And if they’ve got so much extra space and power, why hasn’t this happened before? I think most of these facilities date to the Asia Global Crossing buildout a decade ago, did they overbuild it back then? Or is Pacnet embellishing a tad aggressively? I suppose we’ll see when the space comes online and customers move in.
If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!Categories: Datacenter · Undersea cables
Since most subsea cable landing stations are owned and operated by consortia, they tend to be no larger than required to meet the needs of termination and forwarding.
I suspect the only ones that are now conducive to the type of multipurposing activities that PACNET is pursuig are either (or both) newer and probably proprietary in nature.
Some of the older landing stations are surprisingly small, akin to terrestrial optical repeater stations (in fact some of them are built to the same based building specks), while still a few others along shorelines are retrofitted radiotelephone transmitter and receiver stations that ceased operating during the late seventies to early eighties.
Back in November of ’08 the ITU urged the sharing of these and other types of facilities. See:
I have been in several landing stations, and I agree with the earlier comment by FAC, most are old and undersized. Almost all of these are built with 48v DC power, have no UPS, limited cooling, little to no fire suppression, some with limited terrestrial fiber to get them to a population center.
If Pacnet is serious, I hope they have a boatload of cash to pay for the massive upgrades it would take to turn one of these facilities into a real “data center”.
Stay away from Tsunami Data Centers or coastal Hurricane Data Centers … I prefer inland data centers myself ….
Heh, actually it would be Typhoons in the far east, but the risks of shoreline data centers did occur to me also. I suppose that depends on the particular geography.
A few carriers have toyed with this idea and if the facility is right, in theory it can work. Pacnet’s landing stations are relatively new, large and owned fully by them – their key sub cable assets are the former Asia Global Crossing systems, which were independent not consortia.
Submarine systems do require little physical colocation space but some of the landing stations are built pretty large anyway, to enable unconstrained meet-me-rooms and to allow for other systems to land at the same location. Of course, some are multi-storey and if the local planning laws allow it, PacNet can just add another floor.
If your facility is adequate, you main problem is then customer-related. Customers will have to be prepared to traipse out to some god-forsaken location to access their cabinets and you will have to staff your landing staff with appropriate people 24/7. That becomes an expensive proposition. It also means that each market will be different and so you could only ever implement this strategy on a case-by-case basis. Look for very expensive and capacity-constrained markets where the landing stations are close to the city centre and readily accessible. Singapore, perhaps.
There is one interesting point here though. PacNet is the major partner with Google on the Unity trans-Pacific system, and partnering also with them on another pan-Asia system. I can well imagine a deal whereby Google seeks to lease excess colocation space from its other system partners for its own data centre requirements and PacNet wins the gig to manage this. With PacNet’s own systems being seamlessly connected to the Google-consortia systems, PacNet could then make similar space available to Google through its Asia network.
Result: PacNet gets an anchor tenant for rent and subsequent capacity sales on its pipes, customer access isn’t an issue as with a single customer you can work something out in terms of staff costs. Google gets quality colo space sitting right on top of a fat cable node. I can see that working.
Thanks fluids, that’s good intel and an interesting angle on the Google deal potential – perhaps something like that is what is driving Pacnet’s initiative.
Another consideration occurred to me while reading fluid’s (most interesting) account. Submarine spans are among the most vulnerable, by nature, but also make the worst survivor elements during outages, since they are routinely bypassed entirely either through meshed or ring restoration schemes, the moment a break occurs. Inland nodes survive via alternate paths to other crossings, but the landing stations that are attached directly to failed subsea sections remain stranded until deep water rejointing is completed. Unless, of course, extrarodinary measures are put in place overland to backhaul from other subsea links.
FAC – that is a very good point. In terms of the landing stations I know about, most (though not all) are in fact backhauled to other landing stations so that carriers can enjoy ring protection on their routes. One of the problems Pacific Crossing had when it was trying to sell its PC-1 system was that it lacked connectivity to other landing stations on the Japan side. It then made it a priority to implement this and a little while later managed to do a deal. So your point would definitely be a major consideration and a deal-killer if protection wasn’t in place.
converted telco pop’s aren’t data centers.
Landing stations nowadays are often built into existing data centres now that the power and space requirements are more modest (about 35m2 of space, for example). Older landing stations on the coast are typically built with solid floors and insufficient ceiling height to add a raised floor. This dramatically affects what can actually be put in there (particularly since the cooling capability is thereby reduced) and limits the attractiveness as a datacentre. They are also very power constrained and as stated earlier often only provide DC power.
I’m not aware of many (if any) landing stations “backhauled to other landing stations.” In locations with many landing stations, such as Singapore, HK and Japan around Tokyo, many providers run their own metro fibre rings to the various stations, but the network is designed to aggregate and trunk back to some core switch elsewhere, not make routing decisions between landing stations, particularly as space in the LSes is scarce and expensive (eg, FT is trying to charge over $1M for 3 racks in the IMEWE landing station).
The metro ring between landing stations scenario is unlikely to exist for Pacnet because they have very limited holdings of capacity offnet to their own cables.
Finally, Pacnet is a very minor supplier to Google in the overall scheme of things, and Pacnet would need much higher grade facilities than the EAC/C2C landing stations to satisfy Google’s requirements.
data centers are not located near oceans.
landing stations are located near oceans.
During the sixties I was involved with the day to day workings of TATs 1 through 4 and became the NA implementation coordinator for TAT 5. During those earlier days up through about the time of Divestiture 1, the topological hierarchy and naming convention for buildings was fairly well defined by the ITU (nee CCITT), consisting of landing stations, international transmission centers (ITMCs) and international gateway offices belonging to International Record Carriers (IRCs). This model, give or take some variations per locale, still holds true for many parts of the globe today. Dedicated backhaul facilities between landing stations were non-existent, but restoration plans between ITMCs and Gateways were routine.
Purveyors of that era consisted of telephone companies and International Record Carriers, primarily, where the former (AT&T, essentially) was franchised (under consent decree, basically) to sell voice message services and the latter IRCs held reign over data services and contract overseas IRUs.
(Yes, IRUs existed way back then, too, at about the time Dr. Schultz was co-inventing fiber optics, but they came in denominations as “high” as 3.2KHz voice frequency channels and telegraph channels as low as ~45.45 baud, or 60wpm.)
In the NY Area, for example, two of the transatlantic cables (TAT 2 and TAT 4) beached at Tuckerton, NJ where the landing station was, and were homed via microwave and coaxial cable (L1/4) to the NY ITMC in Manhattan. The ITMC, in turn, sent voice capacity to the Long Lines message services group and fed the data capacity to the major IRCs (ITT, RCA, French Cable, WUI).
TATs 1 and 3, on the other hand, beached at Sidney Mines, Nova Scotia and homed to the White Plains ITMC via microwave primarily, where again AT&T disseminated voice trunks for message services and sent the data bundles down to the major IRCs in NY.
Failovers were not facilitated by self-healing rings, meshed routing arrangements or anything of the kind, but mainly provisioned through manual patching arrangements coordinated over hoot’n’holler orderwire circuits instead. Satellites played prominently during restorals, more so than cable-to-cable, when it come to making facilities “good” during outages.
Of course, bandwidth denominations were much lower then, too, where the highest capacity line that was available on the undersea side was a 3200 Hz voice frequency channel, and total circuit counts ranged into the hundreds on a busy day, not tens of millions of “equivalents” as is the case today.
Terremark recently acquired 1600m2 in the Nautilus Mediterranean cable head near Istanbul, Turkey, where it will operate a data center in a bet on future liberalization of that market.
Istanbul, eh? That’s an interesting bet…
Having seen some of the PacNet CLS it is hard to see how these sites will ever be viable as world class DCs as they are either remote to major cities (eg. Philippines and Taiwan) or simply have inadequate power (Hong Kong) or have already been converted to other uses (Singapore).
In all cases there is simply a patent lack of diversity needed in a proper DC and being “carrier neutral” or at least “carrier friendly” has been a hallmark of world class DC players.
Also Mr. Barney is (again) not shy of some imaginative press. He states ownership of sites in Japan but I recall they are in fact owned and managed by Pacific Crossing (now NTT) as part of the bankruptcy settlement from the old Asia Global days and PacNet has no rights to these stations. Even more embarassing I noticed in an interview on this subject in Telecom Asia the number of CLS awaiting conversion was actually more than PacNet owns and operates.
I think you will see more and more of these sorts of releases from PacNet as they run up to their long delayed IPO. These bright ideas will all form part of the IPO roadshow to add some substance to what is essentially another subsea operator. The plethora of new cables in Asia have taken the shine out of the PacNet story and they need something new.
I hear the IPO has been postponed indefinitely. Sales are down and the India cable is not going ahead. Insider says pressure on costs and consolidation – back to a carrier centric role in 2010.
I worked with the legal team on the original cable landing station leases in Hong kong and the lease prohibits use of the land other than as a landing station … this is why the HKG government granted the land and low cost rental.
With land costs one of the highest in the world why would the HKG government allow a variation to this lease ?
I am sorry … but is there anyone who even cares about the truth in these statements and calls these people out ?
Hey anyone see any new data centres opened by PacNet at their landing stations … NOPE … thought so !!!!!!!!!!!!!