Is capitalism RAID 1 or RAID 5 – The Chain and the Starfish

Strategies for breaking up assemblages must be based on an understanding of the different possible relationships between the components of the assemblage. To examine two different types of assemblage we will consider the chain and the starfish.

The chain is an assemblage of links. For the chain assemblage to maintain it’s integrity under stress, every single link component must not fail. The failure of any single link brings about the catastrophic failure of the entire chain assemblage.

Strategies for breaking up assemblages must be based on an understanding of the different possible relationships between the components of the assemblage. To examine two different types of assemblage we will consider the chain and the starfish.

The chain is an assemblage of links. For the chain assemblage to maintain it’s integrity under stress, every single link component must not fail. The failure of any single link brings about the catastrophic failure of the entire chain assemblage.

The appropriate strategy for breaking a chain is to find it’s weakest link and concentrate your force of attack on that. Thus we say that a chain is only as strong as its weakest link.

The starfish is an entirely different assemblage from a chain. It is a living organism whose assemblage has been shaped by the force of evolution rather than human artifice.

Starfish have the unusual capability to regenerate legs that are severed. This goes beyond the ability of a lizard to shed its tail. Depending on the starfish they can either regrow severed limbs from a whole centre, half a centre or even just an intact fifth of a centre. A very few need even less, such as the Linckia starfish or Blue Sea Star. This remarkable creature does not need any section of centre, it can regenerate entirely from one of its constituent legs. A surviving single leg will sprout four tiny legs, in which state they are called comets, from their appearance, and will eventually regrow to its full five-limbed form.

Of course evolution has produced many wonders in the natural world that we are currently unable to reproduce with our current levels of technology – the use of turbulence by the bumble bee for flight for example. However, before we shift to the sphere of human-artificed objects, let’s take a quick detour via the ancient mystical speculations of the east.

The great Japanese scholar and translator of Zen for a western audience, D.T. Suzuki, often referred to an image from his favourite sutra, the Lankavatara Sutra. In this sutra repeated references are made to a net of jewels which Suzuki interprets as a metaphor where within each jewel are tiny reflections of all the other jewels in the net.

This being the realm of the mystical, the jeweled net is of course infinite. However, returning to the realm of the material, finite and human-made, the principle of an assemblage of differing parts which contain within themselves resonances of other parts of the assemblage has found application in the world of computer systems engineering.

Like all systems engineering, IT systems administration has to deal with the fact that all made things are limited in their performance and their working lifespan – everything fails in the end. This provides twin challenges in the realm of disk storage. How to overcome the performance limitations of I/O speeds for current disk technology and how to engineer your systems such that disk failures do not lead to the irretrievable loss of the data stored upon them? The stupid answer would be to spend more money on making disks faster and more long-lasting. This is dumb for two reasons. First to throw money at an engineering problem to try to exceed the current tolerance levels quickly becomes a case of diminishing returns in cost-benefit terms. Secondly, you’re not solving the problem, simply putting it off.

The most common solution is what’s called a RAID array, originally standing for Redundant Array of Inexpensive Disks (the Inexpensive is now increasing replaced by Independant, as being more marketing-friendly). In this a number of disks are arranged in an array so as to appear as one virtual disk to the system. There are several different assemblages of these multiple disks which solve the twin engineering challenges of performance and recoverability in different trade-offs. They are designated by different numbers such as RAID 0, RAID 1 and RAID 5.

RAID 0 addresses the performance issue by setting multiple disks side-by-side and dividing a chunk of data written to the virtual disk into a number of sub-segments, as many as there are disks in the array, and writes or reads them in parallel as a "stripe". So, say we had a 4-disk RAID 0 array, we could read and write a given chunk of data in roughly one quarter of the time it would take us to read or write to a single disk.

However RAID 0 doesn’t help at all with the problem of disk failure. The loss of any disk in the array will lead to the loss of all data since the data on the remaining disks will be corrupt without the missing stripe. Worse, by multiplying the disks, you multiply the chance of a failure occuring in a given time period. RAID 0 is thus analagous to our chain assemblage.

RAID 1 is a simple solution for the problem of disk failure. You have two disks, one being the exact mirror of the other. So long as you have a separate disk controller for each disk, writing is not much slower than it would be for a single disk. Reading can be faster, because you can read in parallel from both disks, but the main advantage of RAID 1 is that when a disk dies, you still have a copy of your data. You can replace the failed disk and rebuild the mirror from the survivor.

The big disadvantages of RAID 1 is that it doesn’t give you great performance and its the most expensive way of getting failure-proofing, called redundancy in systems engineering after the most common means of achieving it (think parachutes). For every unit of storage of your virtual disk, you need the double in physical disk storage.

RAID 5 is an attempt to combine the performance gains of striping, like RAID 0, with redundancy like RAID 1. In RAID 5 a virtual disk is made up of a multiple of disks (NB not 5 disks, the RAID levels have nothing to do with the number of disks involved) which must be at least 3. Similarly to RAID 0 a chunk of data to be written to the virtual disk is divided up into chunks to be striped across the disks, in addition a parity stripe is calculated by combining all the data stripes by binary arithmatic. The data segments plus parity are then written to the RAID stripe. Which disk gets the parity stripe is taken in turn. When a disk failure occurs the missing segments of the stripes can be recalculated from the remaining data segments and the parity segment for that stripe. The failed disk can then be swapped out of the array for a new replacement and rebuilt using the same mechanism.

Clearly the performance of RAID 5 is lower than RAID 0 and the recuperation of the data from a lost drive takes more work than having a full second copy as in RAID 1. But it has redundancy for much less disk resources than mirroring (e.g. a 4 disk RAID 5 array would make 75% of physical storage available as virtual storage, as opposed to just 50% for RAID 1) and performs multiple times faster than a single physical disk due to striping. It’s a relatively cost-effective engineering compromise that is still in widespread use in the industry, even though SAN and network storage technologies are slowly making other RAID configurations (e.g. RAID 1 + 0, a.k.a RAID 10 which creates stripes of mirrored disks) and more complex dynamic solutions more common.

RAID 5 is then like our five-footed friend the starfish. The removal of a single component will not destroy either the starfish or the RAID 5 array. On another level RAID 5 takes redundancy beyond that of the starfish type – similar to RAID 1 – that can only regenerate a single pattern. RAID 5 has more of the nature a Lankavatara net of jewels. Each disk in the assemblage contains reflections of the diverse contents of all the others within in it, thanks to the parity segments (and the power of binary encoding which makes parity simple).

So what implication does this have for strategies for breaking up assemblages? Clearly if the assemblage has self-repairing features like a starfish or RAID 5 array, then the strategy of striking at the weakest link will fail entirely.

This has serious political implications depending on whether the complementary institutions of capitalist society – the state, private property, exchange and the wage – are a brittle, mutually-dependant assemblage of the chain or RAID 0 type, or a self-reinforcing, self-regenerating assemblage of the starfish or RAID 5 type.

In his "Critique of the Gotha Programme", Marx made his opinion on this question clear – to him the removal of the institution of private property meant the retention of the money and wage form as well as the state, unproblematic. Both the latter institutions, although they still had the appearance of their capitalist fore-runners, "content and form are changed" by the prohibition of private property in the means of production.

For Marx then, the relations of capitalism are a chain assemblage, where the removal of the private property link will mean that the "lower stage of communism", even though it still contains exchange, the wage and the state, will inevitably proceed to develop the forces of production to the level where distribution will be according to needs rather than contribution and the state, already become the mere "administration of things", will wither away.

In contrast the classical anarchists moved from Proudhon’s initial promotion of a society that eliminated the state while retaining both private property and exchange, to a collectivist position that put the means of production into the hands of the producers. But it was not the controversy between Bakunin and Marx which led to the initial split in the International, that developed anarchist politics, but the "Public Services" debate between the Belgian syndicalist César de Paepe, and the Italo-Swiss Bakuninists. In the course of this debate de Paepe argued that public services such as health and education should be provided at a general social level rather than exchanges between individual collectives, and should be freely provided whether the recipient was currently capable of productive work or not. For the provision of these services, members of the producers collectives would have to be prepared for deductions from their earnings to pay for it. Naturally a social body should be responsible for collecting these deductions and organising the provision of these public services, that body should be the Worker’s State, in de Paepe’s view.

De Paepe’s challenge provoked responses from Bakuninists such as the Swiss collectivist Adhémar Schwitzguébel which was as long as attacking the Worker’s State for the evils of statehood, as it was short on actually answering the public services question.

It was this inability of the collectivists to answer de Paepe’s challenge – that to abolish private property but retain exchange, necessarily implied the retention of some form of state – that pushed the classical era anarchists into re-examining the possibility of communism rather than collectivism as a possible solution to the conundrum. In the end they concluded that de Paepe was basically right – even with the abolition of private property of the means of production, to retain distribution in proportion to contribution to production, could only lead to the regeneration of the state. The relations of capitalism were not a chain, but a starfish assemblage.

So,is capitalism RAID 1 or RAID 5? It’s a question that is of prime relevance today. Despite the historical evidence of the 20th century. Namely, that abolishing private property in the means of production while retaining the state and the wage, leads to a transition not from capitalism to communism, but from feudalism to capitalism, as demonstrated by both Russia and China, there are still those today who claim that these outcomes are the results of purely contingent political factors. Further there are many who still appear to claim, along with Schwitzguébel, that there are no inherent contradictions in a new world free from both capital and state, but retaining the wage work ethic, that cannot be ignored if you try hard enough.

For libertarian communists though, the answer is clear and the lessons are that we must examine the specific inter-relations of capitalist social institutions with care and build a strategic direction that is aware of the necessity to not only fight but to break through on multiple fronts if we are to make a new world that transcends the ability of capitalist social relations to re-compose their net of coprolites.