Data Centers (from Downtown/Financial district infill)

[Stlin]

Thanks so much for all that info! Please, seriously, try and shoehorn your datacenter knowledge into every thread in this forum. If people say it's off topic, I'll defend you.

That's wild that this is the only carrier hotel in New England. Out of curiosity, do you know where else in the region would even work? I would have thought places like the data centers on Inner Belt Road would have the kind of power/fiber redundancy for something like this?

Does a data center need to be centralized in a major downtown? Why couldn't they build a dedicated one in the South Boston wasteland somewhere, like maybe the upcoming Gillette site, or on the Dot parcels or something? There's still plenty of space to build a monstrous land-scraper and then get taller/denser housing in DTX. Obviously I don't know the logistics of moving a data center, but don't understand why it HAS to be where it is now.

Admittedly I work less on the telco and infrastructure side and more on the data side of the data center space, but New England generally isn't prime territory for data centers, so the lift in getting a new one approved is... Non-trivial. The 4 major factors people are looking for are land, power, network and taxes. None of those are particularly favorable anywhere in Metro Boston, so most of my work is for the Mid-Atlantic. That said, my understanding is the inner belt data centers combined are a fraction of the size what's above the Macy's, but the biggest factor is probably southbound network connections.

Keep in mind a lot of the modern Internet backbone in the US is built upon the bones of the Bell System of telephone lines, which itself is built on the bones of the telegraph, which did two things - follow the railroads, and have major switching nodes in big cities. And by following the railroads, like the modern CR system, consequently the telephone (and now the Internet system) is also bifurcated into North and South sides.

Look at the telecommunications act of 1996 and the MBTA's real estate revenues - fully half is telecom leases on its rail RoW, and MassDOT highways also earn a significant amount. On the Northside, they converge in the inner belt. The problem then is the Southside connection to Providence and NYC and beyond, which also follows existing RoW - either MassDOT highways or MBTA rail via Worcester or Providence, both of which historically have converged at S. Station. But that historical bifurcation also means that, to this day, ISP providers often own one side or the other of the connection in Boston - not both. (See attached sample provider network.)

Remember that Alexander Graham Bell lived and worked in Boston and invented the phone here; to my understanding the very first switchboard was in what is now downtown crossing on Washington. From the get go, there had to be a connection between the North and South, and pre telephone? The Western Union telegraph office on Devonshire, and later in PO square next to Bell. While the first commercial telephone switchboard was built in New Haven, legacy telephone trunks and that RoW runs underground here - there's also no accident that the ca 1919 central switching building is also downtown just down the street, with the later 1947 building just down the street at PO square. it's some seriously legacy infrastructure. (Though probably in man accessible utilty tunnels given the vintage.) That's why this is where it is. It sits midway between the southern connections and the northern ones, smack dang on top of a century old Bell trunk, so that north and Southsides have "neutral ground" upon which to exchange packets.

The new fiber being installed on the Fitchburg is 96F + 288F and empty blowable duct - the T itself is only going to use a fraction of that, and will most likely lease the rest out as dark fiber. Here at least You don't need NSRL - you could use the Big Dig highway tunnel plenums - but that's still a major ask to connect the two sides, and doing so would in all likelihood still require a building like this, physically in the middle of the network to make that work. Beacon Park Yard could possibly be an option with enough investment in connecting fiber along MBTA and MassDOT RoW, but it's a lot of infrastructure that already exists downtown. And a new build would be similarly proportioned, with worse ground level activation.

 

[Stlin]

Major co-location data centers need to be locate at NICs -- the major interconnect points of the internet. The fiber core of Boston's internet connectivity is in downtown Boston, basically because that is where the legacy telecom provider was located.

The clients of the co-location center are paying to be directly connected to the fiber backbone.

This is less important for something like a traditional colo or the hyperscalers these days - those do just fine in places like Lowell or the inner belt as edge centers. It was admittedly much more important to be close when bandwidth was at a premium, but for this particular site it's a carrier hotel. Which means it provides an interconnection between multiple network providers, where ISPs transfer data to and from each other. In this case, more than 100 of them. Many are commerical non-consumer providers, but the basic idea is that this building is like an airport hub, where like a connecting flight your email sent via your Verizon DSL connection gets shuffled to your friend's Xfinity served home. That makes it much more of a piece of crucial infrastructure (being the only one in New England) in the system than your average data center which may just be a CDN hosting the hot new Netflix show closer to its consumers to avoid the network traffic of sending it from Iowa.

Fastest connections are still on top of the backbone. Co-location on top of the backbone (at an IXP) is prized by certain sectors who demand low latency, fastest data transfer, and access to the maximum number of ISPs and carriers.

Precisely - granted fewer people need that, but for the 100+ ISPs that 1 Summer serves? Absolutely crucial. This building goes down, so does basically the entire Internet in New England. (Or at least northern NE) I suspect the fail over is Albany, but that's a much lower capacity node.

No, the largest colocation centers are in the boonies, often near airports where the power isn't a worry. Fiber runs cheap these days.

Fiber itself may be cheap, but putting in where you want it isn't. The cost for residential FTTH (fiber to the home), which is a single fiber, is estimated at $10000 per home passed, and that's for a aerial installation on a telephone pole in your average suburb. It goes up exponentially with distance and complexity - the MBTA's Fiber resiliency project ran something like $90+M to plow it just under the Fitchburg & Lowell lines.

lol in the very example she gave the colocation is underneath 20 stories of offices

That particular example also is a conversion to a data center usage - the Verizon building was and is also a major telephone system node. It's construction for 70s era telephone equipment made it suitable for conversion to a modern data center once that shrank, but the office portion was retrofitted in, not the other way around, because they couldn't retrofit the existing building to have enough power and cooling to run it all as data center - switches used a lot less power than servers, so it has literally a third of the capacity of 1 Summer, with inefficient floor plates. And, admittedly I wasn't very clear - unlike here, its not critical infrastructure, not being a carrier node with T4 requirements - it's a T3 which is much more in line with average Colos, and is network wise also similar, which means much fewer security concerns, and also means N+1 redundancies, not 2N, which take up less space and power. The carrier hotel honor(s) in NYC belongs to places like the Long Lines building and 60 Hudson, formerly the Western Union building. Again, it's that telegraph legacy at work.

I was trying to acknowledge that it exists as a mixed use data center, and there are other (pre 9/11) examples, but it's also not particularly comparable to what we have at 1 Sumner.
(Also my pronouns are he/him please)