Quintillion is a fund administration company that provides services to financial fund-holding organisations, in large part crunching numbers to validate market activity. It is based in Dublin, where it has two datacentres – a primary site and disaster recovery site – and is a subsidiary of the US Bancorp company.
Its IT infrastructure is 90% virtualised on VMware on Dell and Fujitsu-Siemens hardware plus some servers dedicated to running its core accounting app, Geneva by Advent Software. Its storage assets comprise a Dell Compellent SAN at each of its datacentres.
The move to deploy PernixData resulted from increased volumes and complexity of business that placed greater demands on SQL Server databases which provide business intelligence, workflow and dashboard functionality to the core Geneva application.
Quintillion CTO Charles Gillanders said: “We had gone from 5,000 to 10,000 transactions a day for our entire client base to a situation where one client was doing 50,000 a day. This created a general sense of everything slowing down and anecdotal user complaints.”
Gillanders said the problem was fairly quickly isolated to the SQL Server workload and to writes in particular. “The storage was using most of the available IOPS for writes and this was also impacting reads,” he said.
At the time, the company was looking to upgrade its existing SAN but came across PernixData, then in stealth, and approached the caching software company about running a beta of its FVP product at Quintillion.
More on cache and caching
- Definition: Cache algorithm
- Data loss concerns with write caching products
- Flash caching software market roundup
- Martin: Using SSD as cache
- Comparing network, server and storage-based flash caching
- Considerations when using server-side caching in virtual environments
- Flash hypervisors, caching engines similar, but not the same
PernixData FVP is deployed in the VMware hypervisor. From there, it intercepts database writes and sends them to flash storage, which in Quintillion's case comprises two duplicated instances on Intel flash drives on separate servers.
The data can then be used by the database from flash, while the data is also written to the storage array. FVP ensures hot data is retained in cache for rapid access.
PernixData’s product creates a form of write-back cache, where data is acknowledged to the host before it is written to permanent storage. This can risk data loss in cases where a server fails, but with PernixData FVP, customers can specify that data is written to more than one cache to mitigate against host failure.
Deploying PernixData has enabled Quintillion to avoid a €60,000 upgrade to its Dell Compellent SAN.
Gillanders said: “It has given a new lease of life to our storage arrays and allowed us to take on new, more complex business without disruption and achieve the same or better performance.”
He added: “We are keen to see what will be in version 2 of the product, in particular its ability to use RAM as cache.”