Because our techops guys like to sleep well at night!

Seriously though, 2 reasons - first (and most importantly) is reliability : UNIX (we use flavours of Linux and BSD Unix) is REALLY reliable. We literally have servers that have have their uptime measured in years (1070 days continuous uptime on one machine recently).

And while there will be lots of people who can cite similar stories with Windows® machines, it hasn't been our experience in general. It’s a view that is supported when you consider that something like 80% of the servers deployed on the Internet today run one form of Unix or another.

The second is cost - we have around 80 servers making up the evolve™ system, all of which are at least dual processor (many of which are then multi-core). Around 50 of these machines run a database engine (remember that evolve™ doesn't run on just one machine, but is clustered with the whole system running in parallell over lots of machines).

As a rough calculation, a switch to Windows Server 2003® and SQL 2007® (enterprise, multi-core licenses) would run to about £600,000 (just for the licenses!); a cost which would ultimately result in a higher service fee - all for no practical benefit and (probably) less reliability to our users. You can see why our competitors all run their systems on single machines (and there goes the reliability...).

Absolutely, but hardware RAID has its limitations (and risks) so have systems and infrastructure beyond that to guarantee your data.

All of our storage subsystems are RAID Level 6 (Mirrored and Striped), based around an enterprise class NetApp Metro Cluster Fibre Channel Storage Area Network (SAN) featuring over 12 terrabytes of storage.

Our platform architecture is then designed so that each individual SAN is in turn mirrored in real time to a secondary (identical) SAN storage array.

This provides an additional level of real-time replication over and above the RAID 6 offered by the NetApp filers. Which means that even in the even of a complete subsystem failure (a hardware failure of the RAID controller in the subsystem), your data integrity is guaranteed.

This is serious kit - all told about £250,000's worth just for storing your data!

So, when our competitors tell you 'We use RAID' to guarantee you data.. ask them how they are doing it.. and unless they tell you that they have a SAN infrastructure (one supplied by a by a substantial enterprise vendor..like Network Appliance (NetApp), EMC or Hitachi Data, rather than something they've just cobbled together using a couple of Dell servers and a RAID card), then all they are really doing is relying on the kind of low-cost setup that frankly you could buy yourself for a few hundred quid... is that really what you should be getting from a proper SaaS supplier?

Good question!

When we were testing our current hardware platform we ran out of local bandwidth and capacity (that’s on the system that was acting to simulate the users) at around 12,000 users. We simply could't push any more data over the network to test any more than this (i.e the limitation wasn't the evolve™ platform!)

Right now, we are using around 10% of that capacity and when we do get close, we can simply increase the number of machines we have (the beauty of a multi-tenant architecture - Wikipedia) to accommodate.

In theory this is unlimited (and certainly for all practical purposes) and we have evolve™ customers with hundreds of thousands of documents ( more than 300,000 CVs) and millions of records. Across our entire customer base, we have around 4 million CVs and 20 million documents stored.

For new customers, we do have an AUP (which basically means that we do reserve the right to review data before agreeing to an initial import – we need to be fair to everyone here), but unless you have over 100,000 records or 50,000 documents, you can take it as read that we won’t want to review it.

Background Fact :Average Document Sizes

In terms of individual document size, we ran a program to calculate an average across the entire dataset we hold for our customers (around 30 million documents) and worked our that the average size for a CV or document was around 60K (kilobytes). From this, we set the individual file size limit to 1.5Mb which is around 25x the average.