Mainframe Computers in Shadowrun

The only mention of pricing or functionality for mainframe computers in Shadowrun is in the VR 2.0 rules, and that is frankly silly. They suggest that a mainframe should cost 5 million Nuyen per Security Value point. This doesn’t account for the obvious fact that a Red-8 system should be vastly more expensive than a Blue-8 system. It also means that mainframes would be amazingly lucrative targets for runners (or perhaps that for some magical reason the kind of power described as a mainframe is actually astronomically larger than any of us in our gaming group (most of whom are engineers) can believe, such that a mainframe would be so large as to be nonportable even to a troll).

So here is the system we worked out. Here are a couple of the other goals and criteria we had:

• Mainframes should have a “rating” that works similarly to other “ratings” for gear. That is, the nominal range of a rating should be about 1 to 10, and the cost per rating point should increase in a chunky fashion.
• A given mainframe should be capable of running at any Security Code from Blue to Red, with a dramatic decrease in number of users and applications supportable as the Security Value rises. This would explain why public-access databases are all Blue and Green despite the graffiti problems.

  A system’s Security Code— its color— represents the level of robustness the operating system itself is maintaining. A Blue system is fine for a library computer or Matrix information site, but is insufficiently robust to handle serious programming tasks— it’s too easy to accidentally crash the system, lose data, and so one during the process of working the bugs out of a program. Green is the bare minimum for a serious programmer, and Orange is generally required to handle large numbers of programmers working simultaneously. A program under development can exhibit runaway bugs that can crash a Blue system, slow a Green system to a crawl, die of resource allocation problems on an Orange, or be frozen for the inspection of the system operator on a Red.

  A system’s Security Value— its numerical rating— represents the level of paranoia the system is operating under with regard to cross-checking user input against privileges. This is often brought up to a high value in order to detect incoming deckers, but is also necessary on systems full of programmers in order to keep programs under development from interfering in the dataspaces of other users and the system itself. (In general, a system can only support development of programs that have a Rating less than or equal to the Security Value.)

• All the other things like subsystem ratings should have reasonable impact on the needed system capacity.
• The system need not cover “ultraviolet” systems. Those are defined as magical and huge, beyond rating in nuyen or direct comparison to a normal mainframe.

It turned out that after we sat down and wrote some more detailed statements about what should come out of the desired system, it was obvious to the mathematicians in the group that not only could we meet them all but that doing so gave an obvious simplest result. Here it is:

Basic Rating/Capacity System

Quality	Rating	Cost	Capacity Range	Cost per Capacity Point	Density (per card)
Standard	1-3	100,000¥ × Rating2	1-180	5000¥	5
Alpha	4-6	250,000¥ × Rating2	181-720	12,500¥	10
Beta	7-9	500,000¥ × Rating2	721-1620	25,000¥	20
Delta	10+	1 M¥ × Rating2	1621+	50,000¥	40

The capacity of a mainframe is measured in system capacity (SC), the basic unit for the remainder of this system. For any mainframe, it is 20 times the rating², though a mainframe can be configured to any arbitrary number of system capacity points.

System components have different quality levels; we’ve used ratings similar to cyberware to give an appropriate feel for the levels of quality. The given tiers of mainframe are the ones that require a given quality level of components in order to keep the system from bogging down. In theory, you could use standard components to put enough capacity in one place to make a Rating 10 mainframe, but the components would be unable to correlate with each other fast enough, and the system would crash frequently from routine operations. Since higher-quality components are smaller and faster, you could always build a low-rating mainframe out of higher quality components, which would generally result in a smaller machine.

Mainframes generally take the form of a chassis with a power supply and a rack inside that can hold cards (about the size of a modern PC motherboard) that contain a given amount of processing hardware or interfaces to other systems. A Rating 1 mainframe is no larger than a desktop computer of 1996; larger ones are the size of a refrigerator. Higher-quality components are smaller, faster, and fit more on a card, so it’s possible to make a mainframe smaller by building it out of components that are of a higher quality than necessary for a system. You can get 5 Standard points on a single card; 10 Alpha grade components will fit in the same space, or 20 Beta grade, or 40 Delta grade. (Yes, for 2.5M¥ you can fit a 900,000¥ rating 3 mainframe into a rating 2 chassis by making it out of Beta and Delta grade components; it also means that a Rating 7 mainframe occupies about as much space as a rating 5 or a rating 10.) Machines containing the more expensive cards often have elaborate security systems to prevent someone from making off with 2M¥ worth of components in a briefcase.

Spending System Capacity

Color Multiplier

Almost everything you spend system capacity on has its cost multiplied by a number based on the Security Code. This multiplier represents the amount of overhead the host is using to ensure that programs are running safely.

Color Multiplier (CM)

Blue	1
Green	2
Orange	4
Red	8

System Infrastructure

The capacity cost for the actual operating system and other infrastructure is computed as follows:

Normal Things

Notes:

• Yes, when you increase the Security Code, not only do things all cost more, but the Security Code itself costs more.
• One running IC program or construct counts as one user. If a system ends up with more users than it can handle, start applying penalties such as target number increases and I/O speed decreases. At some upper limit, like twice the allowed number, it should just flat deny further users and/or begin an orderly shutdown, or even flat out crash.
• These costs assume that you are putting the function in its normal subsystem. If a connection is not in the Access subsystem, a slave is not in the Slave subsystem, etc., the cost is 3 times normal.
• “Offline” storage, as described in VR 2.0, is not quite “offline” in the modern jargon. This is auxiliary storage which is very slow to search, retrieve or modify, and corresponds to modern hard disk storage. The system cost here is that for a very small, hence very slow to search index. Actions such as Locate File, Read File, etc. take ten times as long as normal on this data. This is the sort of storage that, say, the IRS would use for tax returns over 5 years old. Offline storage costs 5000¥ per terapulse; a one-terapulse unit is roughly the size of a hefty toolbox (about 45cm × 30cm × 15 cm). Modules like this can be added and removed to the system trivially, and often come with handles for easy carrying.
• “Indexed” storage is similar to offline storage except that “online” storage is used to hold an index to the offline data. This allows searching actions (Locate File) to proceed at normal speed, but reading actions take the same ten times as long, and edit/writing actions take even longer, twelve times as long because the index must be updated.
• “Active” storage is fully available at full speed.

Unusual Things

As mentioned above, putting something in the wrong subsystem costs three times as much as normal. There are some other special cases mentioned in VR 2.0 that this doesn’t cover:

• A timed vanishing SAN costs twice normal. A teleporting SAN with a fixed repertoire costs triple, plus you must be paying the telecom companies for all those SAN addresses. A true teleporting SAN costs four times the capacity.
• The “black box” software to locate a teleporting SAN takes up one Gpulse, which is either 1 × CM capacity on a mainframe or (rarely) a Gp of active storage on a PC or deck. Usually if you want to connect to a teleporting SAN from a small machine you have to go through a mainframe that is running the black box.
• A virtual machine takes capacity on its host as if it were really of the Security Code as the host. Nested VMs work the same way. For example: a Green VM that would take 30 capacity points if it were real is running inside an Orange VM which is running inside a Red operating system. It costs 30 * (4 / 2 = 2) = 60 capacity points on the Orange VM. The part of the Orange VM that is running the Green VM is taking 60 * (8 / 4 = 2) = 120 capacity points on the real Red machine.
• A “bouncer” system, along with similar ideas where the configuration of the system changes on the fly, and including systems that have more than two configurations, must obviously have the capacity to run all configurations. It must also dedicate at least one capacity point times CM per available configuration, under all configurations, to hold the configuration information, and an additional capacity point times the highest involved CM under all configurations to hold the switchover controlling software. Example: a Green-8/Orange-10 bouncer system must allocate 8 capacity points (2 (configs) × 2 (Green CM) + 1 (constant) × 4 (Orange CM) = 8) under the Green mode and 9 (2 × 4 + 1 × 4 = 9) under the Orange configuration to the fact that it is a “bouncer” system.

Subsystem Ratings

At first these seemed problematic, but then we realized that these are inflated to give the right ranges of target numbers.

• Systems of extremely low Security Values— 4 and under— have base subsystem ratings of the Security Value plus 4. A system may not have a Security Value less than 1.
• The base subsystem ratings for systems of Security Value 4 and higher are 8/8/8/8/8.
• If you buy a Security Value greater than four, you get (level minus 4) extra points to distribute among the basic subsystem ratings. Thus, a rating 9 system can have subsystems 9/9/9/9/9, 13/8/8/8/8, 10/10/9/8/8, etc. These extra rating points do not have any extra associated costs.
• Additional points in subsystem ratings (added after any extra provided by Security Values above 4) cost separately per subsystem, and do not take normal Color Modifiers. The first extra point in a subsystem costs 1 capacity; the next costs 2; the next 3; etc. These costs are doubled on Orange and Red systems. Note that this makes the “bonus” points from the Security Value more valuable the higher you want to make your subsystem ratings.

Using a Mainframe to Write Cyberdeck Programs

(This is the other replacement for text in VR 2.0. It retains the flavor and nature of the VR 2.0 rule but has numbers for this system.)

To use a system in developing cyberdeck programs, the system must be running a Security Code of at least Green. The maximum rating of program that can be developed is equal to the Security Value. Program development requires 1Gp per 10 Mp of the program’s design size. The 300,000¥ mainframe programming suite, if present, takes 8 Gp. (As usual for a mainframe app, the size in Gp times the CM gives the capacity cost for running the app.) See the example systems for an example of this.

What about all those little shops with Orange security?

Most of the shops in Sprawl Sites that have impressive security are just leasing space on someone’s mainframe— possibly even a small virtual machine— but certainly don’t actually have Orange systems in the shop waiting to be stolen.

Example Systems

System One: Rating 1 (100,000K¥)

This is a system for a small business. It has 20 capacity points, runs at Blue-4/8/8/8/8/8, has 1 PLTG connection, 1Gp business data & software, and a total of 14 users + slave systems. (It is connected to a PLTG rather than an LTG because it would be a complete cakewalk for even a novice decker, so the business buys LTG service from another company that runs a firewall with a much tougher system.)

Blue-4/8/8/8/8/8	4
PLTG connection	1
1 Gp business data & software	1
14 users + slave systems	14

System Two: Rating 3 (900,000¥)

This is an Easy system at Green 5/9/9/9/9/9. It has 180 capacity points, and is using 8% of them for its operating system and security.

Green 5/9/8/8/8/8	10
Four subsystems at +1	4
83 Gp/connections/users/slaves	166

System Three: Rating 3 (900,000¥)

A decker wanted a good system for programming and put together a small run to swipe the system from the above example. Note that it costs only three times as much as the mainframe programming suite (assuming the decker bought it, rather than pirating it from a contact who owed them a big favor). The system is configured for Green-10/10/10/10/10/10, with no connections, 1 user, 1 Gp misc. data & software, no slaves, a mainframe programming utility and capacity for working on programs up to rating 10 and 700Mp.

Green-10/10/10/10/10	20
1 user, 1 Gp data	4
Mainframe programming suite	16
700Mp program	140

System Four: Rating 5 (6.25M¥)

This is an Average system at Orange-8/13/13/13/13/13. It has 500 capacity points, and is using 73% of them for its operating system and security.

Orange-8/9/9/9/9/8	256
Four subsystems at +4	80
One subsystem at +5	30
33 Gp/connections/users/slaves	134

System Five: Rating 12 (144 M¥)

This is a powerful mainframe designed for a few high-level users to do extremely secure development, at the top end of VR 2.0’s Hard category at Red 12/18/18/18/18/18; it only has room for 18 Gp of non-security code and data. It has 2880 system capacity points, and is using 95% of them for security! A rating 13 system using this configuration could have 80 Gp for non-security purposes, but at the cost of another 25M¥.

Red-12/10/10/10/9/9	1152
3 subsystems at +8	864
2 subsystems at +9	720
18 Gp/connections/users/slaves	144