Gear6's focus was initially on expanding the capabilities of Memcached and its ability to quickly onboard and manage dynamic data. In their research of modern web architectures, Joaquin Ruiz said Gear6 had found many web developers who wanted to persist their mountains of unstructured data in their natural form, instead of force fitting it into an RDBMS. Persistence was especially important with ad-driven sites, said Ruiz. "They were storing hundreds of millions of cookies and they didn't want to instantiate a SQL-driven data structure behind that," said Ruiz
Now Gear6 has shifted its focus towards accelerating structured MySQL data sets in their Memcached server and deploying a NoSQL architecture. Today's release of Gear6 Web Cache integrates their Memcached interface with operational query capabilities and a NoSQL backend. Gear6 decided to use the key-value store Redis because it already had traction among NoSQL users. Mark Atwood said that Gear6 chose a key-value store (as opposed to a document-store or graph database) because it is simple work with. "Cassandra and CouchDB are very usable and useful, but they're not as straightforward as a key-value store," said Atwood. Gear6 wanted a technology that developers from many different backgrounds could pick up and work with.
Ruiz said, "Web 2.0 domains are looking at putting this scalable web framework around their existing software architectures, and [using Gear6] they have a choice of where they want to put data according to their needs." According to the IDC, non-relational data sets are growing 3 to 4 times faster than relational sets, and in-memory data manipulation is important for providing meaningful access to that data. Web Cache users will have the ability to put tabular data in an RDBMS and put less-structured key-value data into Redis. Currently, Redis can be used to build, develop, and deploy Dynamic Data Services. The Redis integration will include advanced data operations.
Operational Query Capabilties
Advanced Memcached query capabilities are another major addition to today's release. In addition to standard key queries, Gear6 Memcached now includes native parallel queries on regex key and value matches. Gear6's new persistence capabilities also provide the ability to snapshot, import, and export cache data, which reduces the time required for cache warm-up and recovery. Caches can also be queried at DRAM speed and queries can be issued specific keys or regular expression matches on keys or values. Gear6 has also added synchronization for WAN and N-Way, which enables Memcached data replication across multiple Memcached servers and clusters in clouds or datacenters.
Memcached's dynamic services support has also been enhanced with elastic sizing, which lets you shrink an active Memcached poolcache without losing cache data. Regular Memcached can't do this because it must run rehashing functions. Gear6 rounds out this release with support for:
- Memcached binary protocol
Ubuntu and CentOS Support
In a separate announcement, Gear6 said that their Memcached distro is now available as a DEB package for Ubuntu and Debian, or as an RPM format for Red Hat Enterprise Linux and CentOS. These packages support:
- Ubuntu 10.04 LTS (Lucid Lynx)
- Ubuntu 8.04 LTS (Hardy Heron)
- Ubuntu 9.10 (Karmic Koala)
- Debian 5.0
- CentOS 5.3/5.4
- RHEL 5.3/5.4
Gear6 sees a lot of replication going on in the cloud, not just as an availability mechanism, but also for distributing data. For its upcoming releases, Gear6 is working on a dynamic data bus for more flexible data migration services between different clouds and datacenters. Ruiz says they're seeing a lot of hybrid operations in web enterprises right now.
Since their last release, Gear 6 has announced dozens of new enterprise web customers for their 'Memcached in the Cloud' Web Cache service. There are hundreds more who have tried Gear6's free services. The Gear6 Amazon pricing model can be found in this article.