Virtual Infrastructure

Virtual Infrastructure

Challenge

Create an open storage infrastructure with the performance and features demanded by large-scale VDI efforts, without sacrificing integrity, efficiency, or excessive cost.

Solution

Leveraging open source software with enterprise-class hardware, WARP Mechanics appliances bring unparalleled performance and capacity in a small package that integrates closely with enterprise virtualization applications. In addition to standard storage features such as thin- provisioning, deduplication, snapshots, and replication, the advanced data protection features ensure consistent availability for virtual desktop systems.

Results

WARP Mechanics provides the complete open storage solution to support tens of thousands of virtual desktop machines while retaining the reliability and flexibility needed for long term data growth and future applications. WARPware features compliment and extend the scalability and reliability of the most advanced virtualization engines in the world.

VDI & Virtual Servers

For some time, storage and system virtualization has been transforming the enterprise to reduce floor tile footprint, electric and cooling load, and management costs. As the success of this has built up over time, businesses have been inspired to “virtualize everything” from servers to desktops.

But this has actually increased cost and complexity. High scale virtual systems come with their own costs: proprietary licensing, equipment sprawl, performance overhead, and recovery issues, to name a few. Admins are constantly on watch for new ways to expand their infrastructure to accommodate new data, retain old data longer, and meet the performance needs of new applications… without breaking the bank.

Enter open storage and virtualization solutions. The WARP Mechanics line of unified storage appliances provide the features and performance that enterprise storage needs coupled with an attractive pricing model. Each appliance supports a mix of HDD, SSD, and NVRAM modules, and numerous high-speed network interface and protocol options. Using NFS or iSCSI, they support the backend datastore of popular virtualization systems like RedHat Enterprise Virtualization (RHEV) and VMware.

As an example, take RHEV. This is a top-end, open, enterprise hypervisor that has achieved high marks in performance and features. It currently holds 19 of the 27 published SPECvirt_sc2010 performance benchmarks for 2-socket and 4-socket hosts, and the only published results for 8-socket systems. It can sustain 550+ VMs per host running enterprise workloads. Individual VMs scale up to 160 logical CPUs and 2TB of memory. So if WARP’s storage can keep up with RHEV, it can handle anything.

Performance & Scalability

With such potent VM scalability, RHEV systems need a storage subsystem that can keep up at every stage of deployment. The WARP appliances have the power to match whether the environment is a handful of small business VMs or a large scale VDI deployment for Fortune 500 companies.

WARP Mechanics’ WARPware OS is based on the same Linux underpinnings as RHEV, and is an enhanced version of the leading supercomputing operating system deployed at Lawrence Livermore National Lab, Sandia National Lab, Los Alamos National Lab, and many other top-ranked HPC facilities. Then WARPware adds all the performance and scalability benefits of ZFS, the most advanced filesystem in the world.

For instance, WARPware supports effectively unlimited file system sizes. It is practical to scale the appliance beyond a petabyte in a single filesystem with current disk

densities. This scale compliments the best practices for VDI, allowing capacity for multiple VM clones and snapshots per user. And, with RHEV’s powerful SPICE remote protocol, VDI encroaches more into traditional “fat” client bastions of rich media production and distribution, making multi-petabyte VM datastores a reality in large organizations.

RHEV solutions require data to be updated frequently, so the file system must handle different versions and update them gracefully. WARPware includes standard enterprise storage features such as thin-provisioning, deduplication, replication, and virtually unlimited filesystem snapshots. Many legacy solutions are limited e.g. to 255 snapshots, but WARPware can accommodate 2 x 1048. Run hourly snapshots if desired, to support extremely granular recover points.

The hardware in each WARP appliance scales to support its filesystem. A single ultra-dense chassis can start with as little as 40TB and grow to 240TB just by adding drives. By adding expansion shelves, the appliances grow to over a petabyte per rack without downtime. Network connectivity options start with a few 1GbE/10GbE interfaces, up to 24x 10GbE ports and 8x 40GbE or 56Gb (FDR) Infiniband ports.

The workload for virtual systems, particularly VDI, is high in IOPS. Some environments can approach database-style IO patterns of many small writes pounding the datastore – the nightmare scenario for legacy storage subsystems. The WARP appliances solve this problem with large DRAM pools integrated with high-performance flash and NVRAM modules. Each module is capable of >500MBps and 100,000+ IOPS, allowing the appliances to easily scale to over 1 Million IOPS per shelf.

Data Integrity

Virtual desktop solutions require a high level of integrity, availability, and reliability so employees can continue working regardless if a host or datastore goes offline. Even though virtualization engines like RHEV have many features to ensure VDI reliability, they can still suffer the problems of large datastores protected by legacy RAID. It doesn’t do much good to roll back a VM snapshot or clone it if the original data is corrupt. Such things can – and do – happen in large datasets

where data can be lost after it’s written due to silent corruption, misdirected IO, and firmware bugs. Legacy storage verifies parity information to detect some kinds of errors, but cannot protect against several other problems like phantom writes (a drive claims to have written data but never did), misdirected reads or writes (a drive read or wrote a data block with matching parity but in the wrong location) stripe corruption or “write holes” (caused by RAID5 read-modify- write), and more.

To solve these problems, WARPware adds further protection with copy-on-write, multi-level checksums, scrubbing, single-, double-, and triple-parity, and ditto blocks. The file system never modifies data in place leaving the original data unchanged. New data is placed in an unused storage location and the pointer to the data is updated, leaving the original intact. All writes are atomic in nature: writes complete properly or not at all, eliminating corruption due to interruptions like power loss. If a bit flip occurs at any stage for any reason, checksum verification will detect the error and correct it. Checksums for each block are stored in physically separate locations with pointers from “parent” to “child” to form a self- validating structure for the entire I/O path, both on the fly and in the background.

Perfect Match

When choosing a datastore provider for a VDI or virtual server environment, scalability, performance and cost are key criteria. WARP Mechanics appliances match and extend the capabilities in virtual infrastructure systems like RedHat Enterprise Virtualization to ensure the growth and reliability that businesses need for long-term success. Whether you deploy it as a stand-alone virtual storage system, or as part of a comprehensive WARP Mechanics data center architecture, WARP appliances provide unmatched density, power, and cost in a turnkey package.

DOWNLOAD WHITEPAPER