C LO U D STAC K Clayton Weise firstname.lastname@example.orgIRC: iswcin #cloudstack on freenode
LICENSING This presentation and it’s contents unless otherwise noted are released under a Creative Commons Attributions, Share-Alike 3.0 unported license.
H I STO RY Original company formed - 2008 (VMOps) Project open sourced as CloudStack – May 2010 Acquired by Citrix – July 2011 Dropped open core – August 2011 Release of Acton (3.0) – Real soon now
W H AT I S C LO U D STAC K ?Open Source Infrastructure as a Service platform thatsupports multiple hypervisors, complex network, firewall,load balancer and VPN configurations, high availability, in amulti-tenant environment.
W H AT D O ES I T R EA L LY D O ? Provide separation for the varied tenants Allocate compute resources in a deterministic manner Expose to the end user the ability to provision various computing services in a controlled manner (VLAN allocation, firewall rules, load balancer deployment, VM creation, etc) Manage High Availability Massively Scalable Permit the placement of resource limits to be applied Measuring usage over time
M U LT I P L E H Y P E RV I S O R S U P P O RT KVM XenServer Xen Cloud Platform VMware (via vCenter) Oracle VM Bare Metal
M U LT I - T E N A N T S E PA R AT I O N Largely built around abstraction from an end-user POV No interaction with hypervisor directly No knowledge of underlying storage Networking separation Every account has at least one dedicated/isolated VLAN (Tagged Networking) Layer 3 isolation aka Security Groups for untagged networking Option to use dedicated hardware
N E T WO R K I N G CloudStack has a number of network models They are generally broken down by: Method of isolation (VLAN, Security Groups) Physical hardware or virtual CloudStack largely manages network infrastructure
N E T WO R K I N G Services managed by CloudStack DHCP VLAN allocation Firewall NAT/Port forwarding Routing VPN Load Balancing
N E T WO R K I N G CloudStack can also manage physical network hardware (or the virtualized alternatives) F5-Big IP NetScaler Juniper SRX Additionally you can ‘mix and match’ some network elements as service offerings.
S EC U R I T Y G RO U P S Traditional isolation has been via VLAN VLANs isolate well, but have some problems scaling Standard has a hard limit of 4096 VLANs Hardware that can actually keep up with 4096 VLANs is VERY expensive. Regardless people tend to not like having arbitrary limits on what they can do. Amazon and others use layer 3 isolation (Security Groups)
S EC U R I T Y G RO U P S Assumption of a quasi-trusted Layer 2 network Typically will only have hypervisors directly connected to that network. Filtering/isolation occurs at the bridge device (from a Linux perspective – think ebtables) Deny by default
H I G H AVA I L A B I L I T Y RFMTTR – but apparently HA looks better in marketing slicks and is used that way across the virtualization industry. CloudStack is not a magical solution for HA – but might be a useful tool in the process to increase availability. CloudStack will watch for HA-enabled VMs to ensure that they are up, and that the hypervisor it’s on is up – and will restart on another hypervisor if it goes down. Redundant router
A L LO C AT I O N A LG O R I T H M S How do you place VMs?, allocate storage, etc. CloudStack ships with a number of options: First Fit Fill first Disperse Create your own Tags OS Preference
U SAG E Not billing per se – but does give you something to bill against. Usage stats show VM count, CPU usage, disk allocation and usage, network usage; all over time. Lots of integration and howto’s - from Excel spreadsheets to Ubersmith, Amysta, and Cloud Portal.
S ECO N DA RY STO R AG E Used for storing templates and snapshots Historically NFS – just added the option of object storage Technically Swift, but Caringo, GlusterFS and others should work. Managed by Secondary Storage VM – manages moving templates and snapshots from/to primary storage, aging snapshots out, etc.
P R I M A RY STO R AG E In the UI we support NFS, iSCSI, and CLVM. We can also make use of local storage No HA, no live migration, etc. Shared mountpoint Anything that all the hypervisors can mount and write to.
R ES O U RC E D I V I S I O N We have some somewhat arbitrary divisions of resources within CloudStack Zones • Pods – Clusters
ZO N E In general practice this is used to designate a specific geographic location. Shares secondary storage resource across the entire zone Single network model for the entire zone
POD In general practice – this is used to refer a rack of machines or a row of racks. Shares guest network
C LU ST E R This is typically a max of 8-15 machines per cluster and homogenity is enforced: Same hypervisor (and same version of the hypervisor) Same CPUs Same networking (i.e. /dev/eth0 is connected to the same network across all machines) Primary storage is cluster specific
P L E T H O R A O F N E T WO R KS Management Network: Where the hypervisors and management server communicate Private Network: Default network for system VMs. (virtual router, secondary storage VM, Console proxy VM) Public Network: The public (often internet-facing network) Guest Network: The network that VMs are provisioned on. Link-local network: The RFC 3927 network used for communication between hypervisor and system VMs.
M A N AG E M E N T S E RV E R UI/API pieces are stateless (state is stored in a MySQL database. All UI functionality is an API call
API RESTful API interface Unauthenticated API interace on 8096 (for localhost) Authenticated API interface natively on port 8080 Responses in XML or JSON http://demo4.cloudstack.org/client/api?apikey=ZRFLiXIkm AHqgRmZzdiXMfaROyK35P_dXxS517WSa9Tmy1Hg&comm and=deployVirtualMachine&serviceofferingid=1&template id=291&zoneid=1&signature=eXW%2fxfqx%2fhu%2frMreF ksVsp3cT4M%3d