EMC vs NetApp: NAS TCO Tussle
Meta Group report says TCO of EMC's NAS beats NetApp's. Is the data legit?
September 12, 2003
A report from technology research firm Meta Group Inc. comparing the total cost of ownership of EMC Corp. (NYSE: EMC) and Network Appliance Inc. (Nasdaq: NTAP) NAS storage systems running databases says EMC's NAS has better TCO than NetApp's does -- but NetApp says Meta's analysis is fundamentally flawed.
The Meta report, published on July 29, compares EMC's Celerra NS600 with high-availability configurations of NetApp's high-end FAS940c and midrange F825c systems, with the goal of determining which platform would be the best option for hosting database applications.
The firm based its findings on published benchmark results for both vendors on Standard Performance Evaluation Corp. (SPEC)'s System File Server (SFS) benchmark. Meta used those in conjunction with Meta's own estimates on the total costs of buying and maintaining those systems over three- and five-year periods to arrive at its total cost of ownership (TCO) figures.
"Although Network Appliance has dominated the nascent NAS-based database market, we believe EMC's new high-availability Celerra system (NS600) will prove competitive in both the price/capacity and price/performance metrics," writes Meta analyst Rob Shafer in the report.
According to Meta, the EMC NS600 "compares favorably" in both areas with either NetApp box. The TCO figures are based on premium hardware and software support options.Table 1: NAS TCO Comparison
EMC NS600 | NetApp F825c | NetApp FAS940c | |
Base system cost (3 Tbytes, high-availability configuration) | $202,680 | $193,000 | $248,000 |
Software | $34,000 | $49,000 | $90,000 |
Installation | $7,830 | $5,200 | $6,400 |
Total 3-year ownership | $281,350 | $297,600 | $424,400 |
SPEC Mark Results (HA configuration) | 25,656 ops/sec | 13,822 ops/sec | 33,340 ops/sec |
3-year price/performance | $10.97/ops/sec | $21.53/ops/sec | $12.73/ops/sec |
3-year price/Gbyte | $91.76/Gbyte | $97.06/Gbyte | $138.42/Gbyte |
If HA weren't part of the mix, NetApp's systems would handily beat the EMC NS600. But Meta's Shafer says high-availability features are "non-optional" in the context of hosting databases: "Whereas in its traditional file-serving area of competence, high availability was rarely an absolute NAS requirement, it is clearly a fundamental and critical prerequisite for most users database-centric applications."
Shafer notes that neither vendor sponsored the report, although EMC has subsequently acquired reprint rights and is offering downloads of Meta's report on its Website. It's also worth noting that EMC just announced a partnership with Oracle Corp. (Nasdaq: ORCL) to provide database-on-NAS services (see EMC, Oracle Prep NAS Nest).
Indeed, EMC sees the Meta report as validating its approach for environments that need redundant failover capability for high availability. "One of the attributes of Celerra NS600 is its integrated high availability... so you don't have to buy two NS600s in order to be able to failover during an outage," says EMC's A.J. Ragosta. "NetApp requires customers to buy two of their systems and cluster them together in order to get HA."
However, NetApp spokeswoman Kris Newton says the system configurations Meta used to estimate price have no relationship to the systems used to achieve the performance numbers."Using SPEC SFS, you cannot create an apples-to-apples comparison of system price to performance," she says.
NetApp points out that Meta priced out 3-Tbyte configurations for the EMC NS600 and NetApp's F825c and FAS940c. However, the SPEC performance numbers used different configurations: The NS600 had 9.9 Tbytes of storage; the F825c had 3.6 Tbytes; and the FAS940c had an 8.4-Tbyte configuration. "A 3-Tbyte Celerra will not perform at the same level as a 9.9-Tbyte system," says Newton. "Essentially what Meta has done is compare the performance of a Ferrari to the price of a Honda."
Meta's Shafer does acknowledge that the performance numbers he cites "do not necessarily reflect the configurations specified in our table and were not performed in a database environment." But he insists that the benchmark figures and pricing can be used as a "reasonable initial relative reference" for IT buyers choosing between EMC and NetApp to host databases on NAS.
"I don't argue with NetApp's point," says Shafer. "It's legitimate -- these are not real-world pricing and performance numbers. But the point is to give users a gross idea that both of these vendors are in the same ballpark."
Just how gross the idea is, though, depends on whether you're looking at it from EMC's or NetApp's perspective.— Todd Spangler, US Editor, Byte and Switch
You May Also Like