public override void checkFairness_blades(hostDB db, disposingList <lockableBladeSpec> blades)
        {
            // If a blade owner is under its quota, then promote it in any queues where the current owner is over-quota.
            currentOwnerStat[] stats  = db.getFairnessStats(blades);
            string[]           owners = stats.Where(x => x.ownerName != "vmserver").Select(x => x.ownerName).ToArray();
            if (owners.Length == 0)
            {
                return;
            }
            float fairShare = (float)db.getAllBladeIP().Length / (float)owners.Length;

            currentOwnerStat[]      ownersOverQuota  = stats.Where(x => x.allocatedBlades > fairShare).ToArray();
            List <currentOwnerStat> ownersUnderQuota = stats.Where(x => x.allocatedBlades < fairShare).ToList();

            foreach (currentOwnerStat migrateTo in ownersUnderQuota)
            {
                var migratory = blades.Where(x =>
                                             (
                                                 // Migrate if the dest is currently owned by someone over-quota
                                                 (ownersOverQuota.Count(y => y.ownerName == x.spec.currentOwner) > 0) ||
                                                 // Or if it is a VM server, and currently holds VMs that are _all_ allocated to over-quota users
                                                 (
                                                     x.spec.currentOwner == "vmserver" &&

                                                     db.getVMByVMServerIP_nolocking(x.spec.bladeIP).All(vm =>
                                                                                                        (ownersOverQuota.Count(overQuotaUser => overQuotaUser.ownerName == vm.currentOwner) > 0)
                                                                                                        )
                                                 )
                                             )
                                             &&
                                             x.spec.nextOwner == migrateTo.ownerName &&
                                             (x.spec.state == bladeStatus.inUse || x.spec.state == bladeStatus.inUseByDirector)).ToList();
                {
                    if (migratory.Count == 0)
                    {
                        // There is nowhere to migrate this owner from. Try another owner.
                        continue;
                    }

                    // Since migration cannot fail, we just take the first potential.
                    // TODO: should we prefer non VM-servers here?
                    lockableBladeSpec newHost = migratory.First();

                    if (newHost.spec.currentlyBeingAVMServer)
                    {
                        // It's a VM server. Migrate all the VMs off it (ie, request them to be destroyed).
                        newHost.spec.nextOwner = migrateTo.ownerName;
                        using (disposingList <lockableVMSpec> childVMs = db.getVMByVMServerIP(newHost,
                                                                                              bladeLockType.lockNone, bladeLockType.lockOwnership))
                        {
                            foreach (lockableVMSpec VM in childVMs)
                            {
                                Debug.WriteLine("Requesting release for VM " + VM.spec.VMIP);
                                VM.spec.state = bladeStatus.releaseRequested;
                            }
                        }
                        newHost.spec.nextOwner = migrateTo.ownerName;
                        newHost.spec.state     = bladeStatus.releaseRequested;
                    }
                    else
                    {
                        // It's a physical server. Just mark it as .releaseRequested.
                        Debug.WriteLine("Requesting release for blade " + newHost.spec.bladeIP);
                        newHost.spec.nextOwner = migrateTo.ownerName;
                        newHost.spec.state     = bladeStatus.releaseRequested;
                    }
                }
            }
        }