![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
tugrikWell, to start off, what I will not do: I will not load Planetside on my work laptop (repeat to self 10 times). I've got a small block of un-busy time at the moment (a rare thing to find at my office these days) and the last thing I want to do is fire up videogames here. So, instead, I figure I'll write out this entry.
I've had a few people ask me what I do for a living. I've avoided going into length about it on LJ for worry that things I do/say here will get back to the bossman. I think I can edit things down to keep a medium level of safety, however, so why the heck not. I'm not going to break any NDA's or trash anybody I work with, so it shouldn't be any big deal. In any case, read on if you want to know what keeps me busy during the workweek.
The business card says, "Network Engineer". In the 2.5 years I've been at this place it's also read "Network Manager" and "Network Architect". As with most of my last decade's jobs it's probably better defined as "networkus many-hattus'. I have this habit of coming in for one role and filling multiple ones instead, either due to my company's small size needing multifaceted people or my own odd spread of skills. Of the three I like 'architect' the best. Not only does it sound nice, but it's more accurate: I'm not a people-manager (thank goodness and I honestly don't feel comfortable calling myself an Engineer. I definitely don't have a completed engineering degree. Many of my friends do and they take pride in that fact. Claiming I'm an Engineer myself makes me feel like a wannabe and doesn't show respect to my friends who've done the time to do it right. Still, my company wants to call me an Engineer to help present things to the customer. What's a blue to do?
Architect is the title of my choice as it describes the part of the job I like best. The companies I work for come to me with management-level goals and general ideas. I have to take these and translate them into as series of needs/problems and their corresponding solutions. Those in turn get woven into a project and wedged into a budget, usually requiring much fussing and squawking on the parts of those who sign the checks. Then it's on to purchasing, implementing, documenting and finally administrating these ideas until they can be passed off to the rank-and-file sysadmins and network techs for the long haul. If I'm lucky once one project ends another begins... but usually there's multiples of these going on at the same time instead, making for a heck of a schedule-juggling act. I've had to learn a lot about managing my time at work over the last few years. Now if only I could bring that control of time to my personal life... :)
What are these projects? Usually they're enterprise-class networks. Anywhere between 50-1000 user corporate wide area networks, including customer networks (that nasty buzzword 'extranet') and all the security and admin pains that go with them. I guess the best way to explain it is to show you what I do at my current company.
We've got three major things I do here: corporate services, customer services and a fun little thing called the 'reference network'.
Taking care of these three things is my job. I have one project manager/boss, one veep above me, two techs beside me to help me with things and a handful of support folks underneath us to help with on-call support. Most of the tasks are proper mental challenges of how to keep things online or how to expand services for the next 50 million handsets wherever in the world. This makes it satisfying to come to work each morning, even though there's times the overnight support or 'my computer is busted!' user stuff can get to me. They're letting me hire one more person "just like me" but with more bias towards Unix administration, and I'm actively seeking that person out. Interested?
Okay, before I finish and get back to the job: AGPS. This is for those of you interested in funky technology. The whitepapers are out there and for those really wanting in-depth stuff I can point you to them in email. This will be the quick glossy-photos version (without the pictures).
It all centers around E911. The Federal Government, through the FCC, has mandated that by 2004 all cellular telephone companies must have E911 implemented and be able to locate a cellular caller when they dial 911 for help. The telcos have to locate them within a certain amount of feet of accuracy within a certain amount of time, 90% of the time; I don't remember the exact numbers offhand. There's a few ways to do this trick. The two most popular are "tower location" and AGPS.
Tower Location is when the phone company knows the location of all their celphone towers and the relative signal strength between the caller's handset and 2 or more of these towers. They can use that, plus the changes over time (if you're driving, for example) to determine generally where the handset is. This works great in areas with high tower densities but is terrible in rural, big-distant-tower areas.
A-GPS, or Aided Global Positioning System, is a modification of the more well-known GPS, like the receivers one uses to go geocaching or to navigate on the highway. GPS watches a constellation of satellites known as "Navstar", listening for the Doppler shift in precisely-timed signals each satellite broadcasts. The reason for A-GPS over GPS has to do with that nasty "time to fix" requirement, battery power, and poor reception area coverage.
Now, keep in mind that this is just a quick gloss-over. A true GPS junkie will probably @snit at me over the niggling details, but it goes like this. A GPS receiver needs to 'hear' three things to operate: Almanac, Ephemeris and "chips". The first two are collections of data telling the GPS receiver about the satellite cloud and the orbits of the individual satellites. Think of this as being told who you're going to listen to and where they should be. The "chips" are the rhythmic little time-pulses each satellite continuously puts out. They're at extremely regular intervals, driven by an atomic clock. Think of this as listening to a beeper on something moving past you: the beep Doppler shifts as it passes you, changing pitch from high to low. If you know the tone and interval of the beeps, you can tell how fast something was moving past you and that helps you find out where it is, and where it's going.
Ephemeris and Almanac take a while to download and are quite low power signals. This is why a normal GPS receiver needs a nice clear view of the sky when it's starting up and takes a little while before it gives position. Once that data is stored it starts listening for the chips off of individual satellites. It figures out the Doppler shift of each, does some happy math, and figures out the distance and delta-v to each satellite it can hear. Put them together and voila, you can triangulate your position on the planet's surface. This is a gross simplification, mind you (there's also things like ionospheric compensation, comparison to terrain models, etc that can occur). The end result is that your GPS receiver had to be 'on' the whole time, draining battery - and it had to have a clear view of the sky for most of it. With a good antenna you can wander under trees or even light building cover and still hear the 'chips', as anybody with a modern Garmin receiver (for instance) can tell you.
A-GPS seeks to shortcut the process and get right to the chips. An A-GPS system has some kind of communications channel (say, SMS on a celphone, a pager, the internet, etc) where it can download the Almanac and Ephemeris ahead of time and store it. That data is good for a few hours at least (some of it far longer), so you only have to do that now and then. You then turn on your receiver only long enough to 'sample' a few short seconds of signals, and then hand it off to the processor. It rakes through the signal repeatedly looking for the 'chips', using the previously stored A&E data to figure things out. Once again, a position results. Only this time, you've only left your receiver on for a few seconds or less. You also didn't have to be out in clear sky; you already had the data you needed and just wanted those chips.
Our product is an 'indoor AGPS solution'. This means that we have a very good receiver that's quite adept at picking out those 'chips', even in extremely noisy situations. Using what's called correlation, we run the same math over the recorded sample, oh, 16,000 times in parallel. Combine this with the fact that 'chips' are so predictable (we know almost exactly where to find them) and we can verify the Doppler-shift at insanely low signal conditions. For those who know signals, we can chug down to -158dB or so. This means that we can hear the 'chips' through most types of roofing material and walls and the reflected signals that come in from low-on-the-horizon satellites make it through windows and layers of parking garages.
In the end this means that a phone with our company's solution on it will get 'aiding data' as needed (not very often), and when you need a position it'll have it within a few seconds outside or in under a minute even indoors or in a parking garage. It's also a bit more accurate and can do a good job of telling where someone is within a building at least enough to narrow things down for emergency personnel. Our solution also uses so little battery that its effect on the average celphone's battery life is almost unnoticeable. Compare this to some of the phones out today that have a traditional GPS receiver in them: they suck battery like maddogs and take a good long while and clear sky to get a position.
So, that's what I do. I fling stations around the planet to study the GPS cloud and hand that data to our engineers. I run the in-house networks those engineers use to develop the tools to make AGPS data and do the daily-business of the company. My team and I together run the customer networks that we use to distribute that aiding-data to the cellular companies. This makes it all the way down to you when you own an E911-compliant celphone in the future. Being a part of this is honestly the most fun I've had in a job since moving to California.
I've had a few people ask me what I do for a living. I've avoided going into length about it on LJ for worry that things I do/say here will get back to the bossman. I think I can edit things down to keep a medium level of safety, however, so why the heck not. I'm not going to break any NDA's or trash anybody I work with, so it shouldn't be any big deal. In any case, read on if you want to know what keeps me busy during the workweek.
The business card says, "Network Engineer". In the 2.5 years I've been at this place it's also read "Network Manager" and "Network Architect". As with most of my last decade's jobs it's probably better defined as "networkus many-hattus'. I have this habit of coming in for one role and filling multiple ones instead, either due to my company's small size needing multifaceted people or my own odd spread of skills. Of the three I like 'architect' the best. Not only does it sound nice, but it's more accurate: I'm not a people-manager (thank goodness and I honestly don't feel comfortable calling myself an Engineer. I definitely don't have a completed engineering degree. Many of my friends do and they take pride in that fact. Claiming I'm an Engineer myself makes me feel like a wannabe and doesn't show respect to my friends who've done the time to do it right. Still, my company wants to call me an Engineer to help present things to the customer. What's a blue to do?
Architect is the title of my choice as it describes the part of the job I like best. The companies I work for come to me with management-level goals and general ideas. I have to take these and translate them into as series of needs/problems and their corresponding solutions. Those in turn get woven into a project and wedged into a budget, usually requiring much fussing and squawking on the parts of those who sign the checks. Then it's on to purchasing, implementing, documenting and finally administrating these ideas until they can be passed off to the rank-and-file sysadmins and network techs for the long haul. If I'm lucky once one project ends another begins... but usually there's multiples of these going on at the same time instead, making for a heck of a schedule-juggling act. I've had to learn a lot about managing my time at work over the last few years. Now if only I could bring that control of time to my personal life... :)
What are these projects? Usually they're enterprise-class networks. Anywhere between 50-1000 user corporate wide area networks, including customer networks (that nasty buzzword 'extranet') and all the security and admin pains that go with them. I guess the best way to explain it is to show you what I do at my current company.
We've got three major things I do here: corporate services, customer services and a fun little thing called the 'reference network'.
- Corporate Network
- Customer Services Network
- Reference Network
This is the part of the network that is the most familiar. It encompasses the parts that the employees themselves use on a daily basis. My current company has about 50 employees spread around the world with offices in California, Colorado, New Jersey and Spain. We also have little one-dude-and-a-link places scattered around the globe for each particular market. At each site you'll have a network core (gigabit at the big sites, 100mb at the smaller), distribution to each desktop and then frame-relay connections keeping them all linked. Each site has its own internet connection and with it a proper firewall. Everything runs on OSPF, a basic routing protocol, so if any one link or internet-service goes down employees get routed around the problem. It's taken me a while to get it set up properly but now it works satisfyingly well. If someone in New Jersey trips over a cord and knocks out their T1 to the world, they're routing out through Colorado without even knowing it. We get a little beep on our management consoles in San Jose and can fix it before they knew there was a problem.
Also included in the corporate network side of the job is where the rubber meets the road of user support: daily work-tickets. We're a small company so we're not quite ready to hire a horde of tech support folks so even though I'm their enterprise network guy I still get my share of "hey, my mouse won't work!" problems. Usually it's a manager who let their kid install games on their laptop over the weekend and show up Monday morning with a pooched system. While this part of the job can be annoying there is one well defined reward to stepping down from the large-scale stuff to fix errant keyboards and jammed printers: it's social. I'm pretty good with user-relations. This way people see me every day and keep me in mind instead of letting me become a line on an org chart or a department to complain to. This makes for happier users who are more willing to tell me their problems in a timely manner, which makes fixes happen faster. A positive feedback cycle, if you will.
The company I work for services and sells to the cellular telcos of the world. If you pay for a cellular telephone somewhere on this planet the odds are pretty high that you're giving money to a company that I have to string a network to. We sell a service known as "A-GPS" (more on that later in this entry) to cel providers. The data being sold is small in size and speed but it has to be both very timely and extremely reliable. The FCC breathes down our neck heavily over this issue as the data is related to Enhanced 911 services ("E911" for short). We maintain what's called a "5-nines" level of support, otherwise known as "staying online 99.999% of the time". To maintain this we can only be offline for 5 minutes and 15 seconds per year. Yes, this is a difficult challenge. The secret? Redundant redundancy, with a side of redundant servers on the side (yes, I'm being cheesy, I know). At least this company realized this from day one and has happily let me build in heaping gobs of redundancy into the customer network. We've been actively selling to customers for a year and a half and have made our 5-nines handily so far.
I call this a 'fun little thing' because it's unique in comparison to all the networks I've worked on over the last decade. To create the AGPS data we must monitor the GPS satellite constellation. This means you have to set up observer stations all over the planet, once again with heavy redundancy, to be able to 'see' the sky. All of the sky, for the whole planet. Each observer station is pretty simplistic in itself. I can't go into details (yay for NDA's) but leave it to say they're cheap enough it's not difficult to litter them around the globe. Each one watches the sky with a GPS antenna and phones home on regular intervals. We take this data and re-create a virtual copy of the GPS constellation and use it to generate AGPS data for each customer's location. As customers are cel-phone providers, their 'areas' are anywhere they have users with handsets.
Setting up an international network in the first place is challenging. Doing it to mathematically-determined locations so you don't have the pick-of-the-crop for data centers and such is even more so. Matching equipment to local power, providing local technical services or increasing redundancy for areas where there's none to be found (to cover for longer station outages if we have to fly out a tech) is a pretty fun challenge. The reference station is the definition of a 'hostile conditions' network. It often feels like defending territory in Planetside: the fringes are always being attacked by something (environment, local politics, equipment failures, etc) and you have to constantly shift attention around to shore up frontlines and send in reinforcements. Now and then you just gotta hop in a vehicle and take a trip to the middle of nowhere to fix the problem yourself.
Taking care of these three things is my job. I have one project manager/boss, one veep above me, two techs beside me to help me with things and a handful of support folks underneath us to help with on-call support. Most of the tasks are proper mental challenges of how to keep things online or how to expand services for the next 50 million handsets wherever in the world. This makes it satisfying to come to work each morning, even though there's times the overnight support or 'my computer is busted!' user stuff can get to me. They're letting me hire one more person "just like me" but with more bias towards Unix administration, and I'm actively seeking that person out. Interested?
Okay, before I finish and get back to the job: AGPS. This is for those of you interested in funky technology. The whitepapers are out there and for those really wanting in-depth stuff I can point you to them in email. This will be the quick glossy-photos version (without the pictures).
It all centers around E911. The Federal Government, through the FCC, has mandated that by 2004 all cellular telephone companies must have E911 implemented and be able to locate a cellular caller when they dial 911 for help. The telcos have to locate them within a certain amount of feet of accuracy within a certain amount of time, 90% of the time; I don't remember the exact numbers offhand. There's a few ways to do this trick. The two most popular are "tower location" and AGPS.
Tower Location is when the phone company knows the location of all their celphone towers and the relative signal strength between the caller's handset and 2 or more of these towers. They can use that, plus the changes over time (if you're driving, for example) to determine generally where the handset is. This works great in areas with high tower densities but is terrible in rural, big-distant-tower areas.
A-GPS, or Aided Global Positioning System, is a modification of the more well-known GPS, like the receivers one uses to go geocaching or to navigate on the highway. GPS watches a constellation of satellites known as "Navstar", listening for the Doppler shift in precisely-timed signals each satellite broadcasts. The reason for A-GPS over GPS has to do with that nasty "time to fix" requirement, battery power, and poor reception area coverage.
Now, keep in mind that this is just a quick gloss-over. A true GPS junkie will probably @snit at me over the niggling details, but it goes like this. A GPS receiver needs to 'hear' three things to operate: Almanac, Ephemeris and "chips". The first two are collections of data telling the GPS receiver about the satellite cloud and the orbits of the individual satellites. Think of this as being told who you're going to listen to and where they should be. The "chips" are the rhythmic little time-pulses each satellite continuously puts out. They're at extremely regular intervals, driven by an atomic clock. Think of this as listening to a beeper on something moving past you: the beep Doppler shifts as it passes you, changing pitch from high to low. If you know the tone and interval of the beeps, you can tell how fast something was moving past you and that helps you find out where it is, and where it's going.
Ephemeris and Almanac take a while to download and are quite low power signals. This is why a normal GPS receiver needs a nice clear view of the sky when it's starting up and takes a little while before it gives position. Once that data is stored it starts listening for the chips off of individual satellites. It figures out the Doppler shift of each, does some happy math, and figures out the distance and delta-v to each satellite it can hear. Put them together and voila, you can triangulate your position on the planet's surface. This is a gross simplification, mind you (there's also things like ionospheric compensation, comparison to terrain models, etc that can occur). The end result is that your GPS receiver had to be 'on' the whole time, draining battery - and it had to have a clear view of the sky for most of it. With a good antenna you can wander under trees or even light building cover and still hear the 'chips', as anybody with a modern Garmin receiver (for instance) can tell you.
A-GPS seeks to shortcut the process and get right to the chips. An A-GPS system has some kind of communications channel (say, SMS on a celphone, a pager, the internet, etc) where it can download the Almanac and Ephemeris ahead of time and store it. That data is good for a few hours at least (some of it far longer), so you only have to do that now and then. You then turn on your receiver only long enough to 'sample' a few short seconds of signals, and then hand it off to the processor. It rakes through the signal repeatedly looking for the 'chips', using the previously stored A&E data to figure things out. Once again, a position results. Only this time, you've only left your receiver on for a few seconds or less. You also didn't have to be out in clear sky; you already had the data you needed and just wanted those chips.
Our product is an 'indoor AGPS solution'. This means that we have a very good receiver that's quite adept at picking out those 'chips', even in extremely noisy situations. Using what's called correlation, we run the same math over the recorded sample, oh, 16,000 times in parallel. Combine this with the fact that 'chips' are so predictable (we know almost exactly where to find them) and we can verify the Doppler-shift at insanely low signal conditions. For those who know signals, we can chug down to -158dB or so. This means that we can hear the 'chips' through most types of roofing material and walls and the reflected signals that come in from low-on-the-horizon satellites make it through windows and layers of parking garages.
In the end this means that a phone with our company's solution on it will get 'aiding data' as needed (not very often), and when you need a position it'll have it within a few seconds outside or in under a minute even indoors or in a parking garage. It's also a bit more accurate and can do a good job of telling where someone is within a building at least enough to narrow things down for emergency personnel. Our solution also uses so little battery that its effect on the average celphone's battery life is almost unnoticeable. Compare this to some of the phones out today that have a traditional GPS receiver in them: they suck battery like maddogs and take a good long while and clear sky to get a position.
So, that's what I do. I fling stations around the planet to study the GPS cloud and hand that data to our engineers. I run the in-house networks those engineers use to develop the tools to make AGPS data and do the daily-business of the company. My team and I together run the customer networks that we use to distribute that aiding-data to the cellular companies. This makes it all the way down to you when you own an E911-compliant celphone in the future. Being a part of this is honestly the most fun I've had in a job since moving to California.
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
no subject
Date: 2003-06-25 04:28 pm (UTC)no subject
Date: 2003-06-26 01:42 am (UTC)no subject
Date: 2003-06-29 04:31 am (UTC)Ya know, when I tell people I'm a security guard, I usually see the person's eyebrow raise slightly, and one corner of the mouth lower a bit.
And they mumble "Oh, well..."
I think security guards rank slightly below burger slinger in our culture. :)