When you are developing a storage solution your biggest worry is data loss. As an Open vStorage platform can lose a server or even a complete data center without actual data loss, we are pretty sure we have that base covered. The next challenge is to make sure that safely stored data can be quickly accessed when needed. In this blog section we already discussed a lot of the performance improvements we made over the past releases. We introduced the Edge component for guaranteed performance, the accelerated ALBA as read cache, multiple proxies per volume driver and various performance tuning options.
Today it is time to introduce the latest performance improvement: High Performance Read Mesh (HPRM). This HPRM is an optimization of the read path and allows the compute host to directly fetch the data from the drives where the data is located. Earlier the read path always had to go through the Volume Driver before the data was fetched from the ASD. This newly introduced short read path can only be taken in case the Edge has the necessary metadata of where (SCO, fragment, disk) each LBA’s data is stored. In case the Edge doesn’t have the needed metadata, for example because the cached metadata is outdated, the slow path is taken through the Volume Driver. For the write path nothing is changed as all writes go through the Volume Driver.
The short read path which bypasses the Volume Driver has 2 direct advantages: lower latency on reads and less network traffic as data only goes once over the network. Next, the introduction of the HPRM also allows for a cost reduction on the hardware front. Since the hosts running the Volume Driver are no longer in the read path in many cases, they are freed up and can focus on processing incoming writes. This means the ratio between compute hosts running the Edge and the Volume Driver can be increased. Since the Volume Driver hosts are typically beefy servers with expensive NVMe devices for the write buffer and the distributed databases, a significant change in the Compute/Volume Driver ratio means a significant reduction of the hardware cost.
HPRM, the technical details
Let’s have a look under the hood on how the HPRM works. First we will have a look at the write path. The application, f.e. the hypervisor, writes to the block device exposed by the Edge client. The Edge client will connect to its server part which in its turn, writes the data to the write buffer of the Volume Driver. Once enough writes are accumulated in the buffer, a SCO (Storage Container Object) is created and dispatched to the ALBA backend through the proxy. The proxy makes sure the data is spread across different ASDs according to the specified ALBA preset. Which ASDs contain the fragments of the SCO is stored in a manifest.
Once a read comes for the LBA, the Edge client will check its local metadata cache for the SCO info and manifest of the SCO. If the info is available the Edge will get the LBA data through the PRACC (Partial Read ACCelerator) client which can directly fetch the data from the ASDs. If the info isn’t available in the cache or if it is outdated, the manifest and SCO info are retrieved by the Edge client from the Volume Driver and stored in the Edge metadata cache.
The Edge also pushes the IO statistics to the Volume Driver so these can be queried by the Framework or the monitoring components. Gathering IO statistics is done by the Edge as it is the only component that has a view on both the fast path, through the PRACC, and the slow path through the Volume Driver.
Note that the High Performance Read Mesh is part of the Open vStorage Enterprise Edition. Contact us for more info on the Open vStorage Enterprise Edition.
When I present the new Open vStorage architecture for Fargo, I almost always receive the following Edge question:
What is the Edge and why did you develop it?
What is the Edge about?
The Edge is a lightweight software component which can be installed on a Linux host. It exposes a block device API and connects to the Storage Router across the network (TCP/IP or RDMA). Basically the applications believes it talks to a local block device (the Edge) while the volume actually runs on another host (Storage Router).
Why did we develop the Edge?
The reason why we have developed the Edge is quite simple: componentization. With Open vStorage we are mainly dealing with large, multi-petabyte deployments and having this Edge component gives additional benefits in large environments:
In large environments you want to be able to scale the compute and storage part independently. In case you run Open vStorage hyper-converged, as advised with earlier versions, this isn’t possible. This has as consequence that if you need more RAM or CPU to run VMs, you had to also invest in more SSDs. With the Edge you can scale compute and storage independent.
With Eugene the Volume Driver, the high performance distributed block layer, was running on the compute host together with the VMs. This results in the VMs and the Volume Driver fighting for the same CPU and RAM resources. This is a typical issue with hyper-converged solutions. The Edge component avoids this problem as it runs on the compute hosts (and requires only a small amount of resources) and the Volume Drivers runs on dedicated nodes and hence provides a predictable and consistent amount of IOPS to the VMs.
Limit the Impact of Updates
Storage software updates are a (storage) administrator’s worst nightmare. In previous Open vStorage versions an update of the Volume Driver required all VMs on that node to be migrated or brought down.With the Edge the Volume Driver can be updated in the background as each Edge/compute host has HA features and can automatically connect to another Volume Driver on request without the need of a VM migration.
Open vStorage is designed to be rock solid and survive failures. These failures can come in many forms and shapes: nodes might die, network connections might get interrupted, … Let’s give an overview of the different tactics that are used by Open vStorage when disaster strikes by going over some possible use cases where the new edge plays a role.
Use case 1: A hypervisor fails
In case the hypervisor fails, the hypervisor management (OpenStack, vCenter, …) will detect the failure and restart the VM on another hypervisor. Since the VM is started on another hypervisor, the VM will talk to the edge client on the new hypervisor. The edge client will connect to a volume driver in the vPool and enquire which volume driver owns the disks of the VM. The volume driver responds who is the owner and the edge connects to the volume driver owning the volume. This all happens almost instantaneously and in the background so the the IO of the VM isn’t affected.
Use case 2: A Storage Router fails
In case a Storage Router and hence the volume driver on it die, the edge client automatically detects that the connection to the volume driver is lost. Luckily the edge keeps a list of volume drivers which also serve the vPool and it connects to one of the remaining volume drivers in the vPool. It is clear that the edge prefers to fail-over to a volume driver which is close-by f.e. within the same datacenter. The new volume driver to which the edge connects detects that it isn’t the owner of the volume. As the old volume driver is no longer online, the new volume driver steals the ownership of the VMs volume. Stealing is allowed in this case as the old volume driver is down. Once the new volume driver becomes the owner of the volumes, the edge client can start serving IO. This whole process process happens in the background and halts the IO of the VM for a fraction of a second.
Use case 3: Network issues
In some exceptional cases it isn’t the hypervisor or the storage router that fails but the network in between. This is an administrator’s worst nightmare as it might lead to split brain scenarios. Even in this case the edge is able to outlive the disaster. As the network connection between the edge and the volume driver is lost, the edge will assume the volume driver is dead. Hence, as in use case 2 the edge connects to another volume driver in the same vPool. The volume driver first tries to contact the old volume driver.
Now there are 2 options:
- The new volume driver can contact the old volume driver. After some IO is exchanged the new volume driver asks the old volume driver to hand over the volume. This handover doesn’t impact the edge.
- The new volume driver can also not contact the old volume driver. In that case old volume driver steals the volume from the old volume driver. It does this by updating the ownership of the volume in the distributed DB and by uploading a new key to the backend. As the ALBA backend uses a conditional write approach, it only writes the IO to disks of the backend if the accompanying key is valid, it can ensure only the new volume driver is allowed to write to the backend. If the old volume driver would still be online (split brain) and try to update the backend, the write would fail as it is using an outdated key.
With the Fargo release of Open vStorage we are focussing even more on the Open vStorage sweet spot: multi-petabyte, multi-datacenter storage clusters which offer super-fast block storage.
In order to achieve this we had to significantly change the architecture for the Fargo release. Eugene, the version before Fargo, already had the Shared Memory Server (SHM) in its code base but its wasn’t activated by default. The Fargo release now primarily uses the SHM approach. To make even more use of it, we created the Open vStorage Edge. The Edge is a lightweight block storage driver which can be installed on Linux servers (hosts running the hypervisor or inside the VM) and talks across the network to the Shared Memory of a remote Volume Driver. Both TCP/IP and the low latency RDMA protocol can be used to connect the Edge with the Volume Driver. Northbound the Edge has an iSCSI, Blktap and QEMU interface. Additional interfaces such as iSER and FCoE are planned. Next to the new Edge interface, the slower Virtual Machine interface which exposes a Virtual File System (NFS, FUSE), is still supported.
The Volume Driver has also been optimized for performance. The locks in the write path have been revised in order to minimize their impact. More radical is the decision to remove the deduplication functionality from the Volume Driver in order to keep the size of the metadata of the volumes to a strict minimum. By removing the bytes reserved for the hash, we are capable of keeping all the metadata in RAM and push the performance across 1 million IOPS per host on decent hardware. For those who absolutely need deduplication there is still a version available of the Volume Driver which has support for deduplication.
With the breakthrough of RDMA, the network bottleneck is removed and network latency is brought down to a couple of microseconds. Open vStorage makes use of the possibilities RDMA offers to implement a shared cache layer. To achieve this it is now possible to create an ALBA backend out of NVMe or SSD devices. This layer acts as a local, within a single datacenter, cache layer in front of an SATA ALBA backend, the capacity tier, which is spread across multiple datacenters.
This means all SSDs in a single datacenter devise a shared cache for the data of that datacenter. This minimizes the impact of an SSD failure and removes the cold cache effect when moving a volume between hosts. In order to minimize the impact of a single disk failure we introduced the NC-ECC (Network and Clustered Error Correction Codes) algorithm. This algorithm can be compared with solving a Sudoku puzzle. Each SCO, a collection of consecutive writes, is chopped up in chunks. All these chunks are distributed across all the nodes and datacenters in the cluster. The total amount of chunks can be configured but allows for example to recover from a multi node failure or a complete datacenter loss. A failure, whether it is a disk, node or datacenter will cross out some numbers from the complete Sudoku puzzle but as long as you have enough numbers left, you can still solve the puzzle. The same goes for data stored with Open vStorage: as long as you have enough chunks (disk, nodes or datacenters) left, you can always recover the data. The NC-ECC algorithm is based on forward error correction codes and is further optimized for usage within a multi-datacenter approach. When there is a disk or node failure, additional chunks will be created using only data from within the same datacenter. This ensures the bandwidth between datacenters isn’t stressed in case of a simple disk failure.
By splitting up the Edge, the Volume Driver, the cache layer and the capacity tier, you have the ultimate flexibility to build the storage cluster of your needs. You can run everything on the same server, hyperconverged, or you can install each component on a dedicated server to maximize scalability and performance.
The first alpha version of Fargo is now available on the repo.